Energy storage system communication method and communication system

Abstract

The present application relates to the technical field of energy storage system communication, and particularly relates to an energy storage system communication method and a communication system. The method comprises the following steps: collecting working condition data of a connection point of an energy storage system through a piezoelectric micro-vibration sensor; extracting vibration-communication coupling features of the working condition data to obtain vibration-signal feature spectrum of the connection point; collecting signal reflection characteristics of a communication line under vibration conditions according to the vibration-signal feature spectrum of the connection point, then performing non-intrusive time domain reflection analysis to obtain a reflection characteristic change map; performing vibration response mapping processing on the communication line according to the reflection characteristic change map to obtain a communication quality dynamic mapping table; and performing dynamic impedance calculation and calibration according to the communication quality dynamic mapping table and the reflection characteristic change map to obtain dynamic impedance characteristic data. The present application can improve the reliability and safety of the energy storage system by adaptively optimizing the communication path and actively adjusting the communication route before predicting a failure of the connection point.

Description

Technical Field

[0001] The present invention relates to the field of energy storage system communication technology, and in particular to an energy storage system communication method and a communication system. Background Art

[0002] Intermittent connection failures in existing energy storage systems are difficult to diagnose. Traditional monitoring technologies rely primarily on electrical parameter detection with fixed thresholds, and are unable to capture transient disconnections or high-resistance states that only occur briefly under specific operating conditions (such as vibration, temperature changes, or during charging and discharging). This causes such problems to be ignored by the system until they accumulate into permanent failures. The predictive capability is insufficient. The threshold judgment method used by existing technologies can only identify connection points that have clearly deteriorated, lacks sensitivity to subtle trends, and is unable to identify signs of connection degradation at an early stage. As a result, maintenance is often triggered only after a failure is about to occur or has already occurred, missing the optimal maintenance opportunity. Static measurement limitations. Traditional impedance measurement methods are usually performed when the system is stationary and require suspension of normal operation. This not only affects system availability, but also fails to capture connection problems that only manifest themselves under dynamic operating conditions (such as charging and discharging processes, temperature fluctuations, or vibration conditions), resulting in a large number of potential failure points being missed.

[0003] In summary, the problem of the inability to effectively diagnose and predict intermittent connection failures of energy storage systems in existing technologies needs to be solved urgently. Summary of the Invention

[0004] Based on this, it is necessary to provide an energy storage system communication method and a communication system to solve at least one of the above technical problems.

[0005] To achieve the above object, a communication method for an energy storage system includes the following steps:

[0006] Step S1: collecting operating data of the energy storage system connection point through a piezoelectric micro-vibration sensor; extracting vibration-communication coupling features from the operating data to obtain a vibration-communication characteristic spectrum of the connection point;

[0007] Step S2: Based on the vibration-signal characteristic spectrum of the connection point, the signal reflection characteristics of the communication line under vibration conditions are collected, and then non-intrusive time domain reflection analysis is performed to obtain a reflection characteristic change spectrum; vibration response mapping processing is performed on the communication line based on the reflection characteristic change spectrum to obtain a communication quality dynamic mapping table; dynamic impedance calculation and calibration are performed based on the communication quality dynamic mapping table and the reflection characteristic change spectrum to obtain dynamic impedance characteristic data; impedance-vibration sensitivity analysis is performed on the dynamic impedance characteristic data to obtain a connection health matrix;

[0008] Step S3: Perform time series data accumulation and denoising on the connection health matrix to obtain denoised health data; perform micro-change pattern recognition on the denoised health data to obtain a micro-change feature set; perform degradation pattern evolution analysis on the micro-change feature set to obtain a connection reliability prediction map;

[0009] Step S4: Perform communication path transfer analysis based on the connection reliability prediction map to obtain a path transfer strategy plan; execute the path transfer strategy plan and calculate the communication resilience index.

[0010] The application can obtain original correlation data of vibration and communication performance of the connection point in actual dynamic environment (including various vibration modes, charging and discharging states, temperature changes, etc.) by deploying a piezoelectric micro-vibration sensor at the energy storage system connection point and synchronously collecting signal quality parameters of the communication link under different operating conditions. Further preprocessing and time / frequency domain feature extraction are performed on the data, and the coupling relationship between vibration features and communication quality parameters is constructed to generate a connection point vibration-signal feature spectrum, which enables quantification of the influence of the connection point on communication performance under different vibration conditions and operating conditions, providing an accurate input basis for subsequent deeper dynamic health state assessment and fault prediction, and breaking through the limitations of static monitoring that cannot capture dynamic coupling effects. Using the connection point vibration-signal feature spectrum generated in step S1, this step can periodically collect signal reflection characteristics of the communication line during operation of the energy storage system through non-intrusive time domain reflection analysis technology, avoiding interrupting system operation for traditional impedance measurement. By correlating the reflection characteristic change map with the communication quality parameters and establishing a vibration response mapping, the specific influence of vibration on communication quality can be assessed in real time. On this basis, dynamic impedance is calculated and calibrated, and the influence of vibration, current and electrochemical state is considered, which can more accurately reflect the real electrical connection state of the connection point under complex operating conditions. Through impedance-vibration sensitivity analysis, the sensitivity and tolerance of the connection point impedance to different vibration frequency bands, currents and temperatures are quantified, and finally a connection health degree matrix is constructed, providing refined and dynamic health state assessment results for subsequent fault prediction and reliability management. After time series accumulation and denoising processing of the connection health degree matrix, this step can accurately capture the small change trend, fluctuation characteristics and correlation with operating conditions, temperature and cycle period of the connection point health degree over time, and identify the mutation point, overcoming the problem that traditional threshold judgment methods cannot detect early signs of degradation. According to the identified micro-change feature set, the degradation mode of the connection point can be classified, and a matching prediction model can be selected to accurately extrapolate the degradation trajectory of the denoised health degree data, predict the future health degree evolution of the connection point, and calculate the probability of failure at different time points in the future to generate a connection reliability prediction map, providing prospective and quantitative risk warning information for taking intervention measures before failure occurs. Using the connection reliability prediction map generated in step S3, this step can perform reliability assessment on the communication topology of the entire energy storage system to identify potential single-point faults and weak links. Combined with communication flow characteristic analysis and backup path resource assessment, it can determine which high-risk connection points carry critical services and which available backup communication paths exist in the network and their capabilities.On this basis, a refined path transfer strategy is formulated, including the identification of high-risk points, matching of backup paths, rationality of traffic distribution, planning of transfer timing, and assessment of potential electrical impacts, to ensure that when degradation or abnormality of the connection point is detected, the critical communication traffic can be switched to a more reliable backup path in a timely and intelligent manner. By executing the strategy and performing adaptive verification and tuning, the transfer effect can be continuously optimized, the communication interruption time and performance degradation can be minimized, and finally the communication resilience index can be calculated to quantify the overall ability of the system to resist and respond to communication failures, significantly improving the reliability and availability of energy storage system communications. Therefore, the present invention provides a communication method for an energy storage system, which realizes real-time monitoring of the dynamic health status of the connection point by combining micro-vibration sensing and communication signal analysis technology; uses non-invasive time domain reflection technology to measure dynamic impedance changes without interrupting system operation; introduces pattern recognition algorithms to analyze the slight change trend of health parameters, breaking through the limitations of traditional threshold detection; and finally, through adaptive communication path optimization, actively adjusts the communication route before predicting a failure at the connection point, significantly improving the reliability and safety of the energy storage system.

[0011] Preferably, the present invention further provides a communication system for an energy storage system, for executing the energy storage system communication method described above, wherein the communication system for the energy storage system comprises:

[0012] The correlation feature extraction module is used to collect the operating condition data of the energy storage system connection point through a piezoelectric micro-vibration sensor; extract the vibration-communication coupling features of the operating condition data to obtain the vibration-communication characteristic spectrum of the connection point;

[0013] The dynamic impedance mapping module is used to collect the signal reflection characteristics of the communication line under vibration conditions based on the vibration-signal characteristic spectrum of the connection point, and then perform non-invasive time-domain reflection analysis to obtain a reflection characteristic change spectrum; perform vibration response mapping processing on the communication line based on the reflection characteristic change spectrum to obtain a communication quality dynamic mapping table; perform dynamic impedance calculation and calibration based on the communication quality dynamic mapping table and the reflection characteristic change spectrum to obtain dynamic impedance characteristic data; and perform impedance-vibration sensitivity analysis on the dynamic impedance characteristic data to obtain a connection health matrix;

[0014] The reliability prediction module is used to accumulate and denoise the time series data of the connection health matrix to obtain denoised health data; perform micro-change pattern recognition on the denoised health data to obtain a micro-change feature set; and perform degradation pattern evolution analysis on the micro-change feature set to obtain a connection reliability prediction map.

[0015] The communication path optimization module is used to perform communication path transfer analysis based on the connection reliability prediction map to obtain a path transfer strategy plan; execute the path transfer strategy plan and calculate the communication resilience index.

[0016] The energy storage system's communication system, through its integrated modules for correlation feature extraction, dynamic impedance mapping, reliability prediction, and communication path optimization, enables comprehensive monitoring and intelligent management of the dynamic health of the energy storage system's communication connection points. The system accurately captures the vibration-communication coupling characteristics of connection points under complex real-world operating conditions (including vibration, current, and temperature). Using non-intrusive time-domain reflectometry, it dynamically assesses their impedance characteristics and communication quality, overcoming the limitations of traditional static monitoring, which cannot diagnose intermittent connection failures. By identifying subtle changes in connection health data and analyzing degradation evolution, the system sensitively detects early signs of connection point degradation and predicts future reliability, providing proactive early warning capabilities and supporting preventive maintenance. Ultimately, based on the reliability prediction results, the system intelligently evaluates the communication topology and formulates an optimized path migration strategy. This proactively mitigates potential failures before they occur, maximizing the continuity and reliability of critical communications and significantly improving the communication resilience of the energy storage system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The figure is a flowchart of a communication method for an energy storage system.

[0018] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0019] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0020] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0021] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0022] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a flow chart of the steps of the energy storage system communication method of the present invention. In this example, the energy storage system communication method includes the following steps:

[0023] Step S1: collecting operating data of the energy storage system connection point through a piezoelectric micro-vibration sensor; extracting vibration-communication coupling features from the operating data to obtain a vibration-communication characteristic spectrum of the connection point;

[0024] In an embodiment of the present invention, a piezoelectric micro-vibration sensor is deployed at the connection point of the energy storage system and communication quality parameters are collected synchronously to form an original vibration-communication data set. The data set is preprocessed by filtering, alignment, outlier processing, and classification by working conditions to obtain a preprocessed working condition data packet. The time domain statistical characteristics and frequency domain vibration characteristics of the vibration and communication quality are extracted from the data packet to form a time domain feature vector and a frequency domain feature spectrum. On this basis, by calculating the correlation and regression coefficient between the vibration characteristics and the communication quality parameters, a coupling feature set that quantifies the impact of vibration on communication is constructed. Finally, the coupling feature sets under different working conditions are integrated to generate a connection point vibration-communication feature spectrum that describes the vibration-communication coupling characteristics of the connection point under various vibrations and working conditions.

[0025] Step S2: Based on the vibration-signal characteristic spectrum of the connection point, the signal reflection characteristics of the communication line under vibration conditions are collected, and then non-intrusive time domain reflection analysis is performed to obtain a reflection characteristic change spectrum; vibration response mapping processing is performed on the communication line based on the reflection characteristic change spectrum to obtain a communication quality dynamic mapping table; dynamic impedance calculation and calibration are performed based on the communication quality dynamic mapping table and the reflection characteristic change spectrum to obtain dynamic impedance characteristic data; impedance-vibration sensitivity analysis is performed on the dynamic impedance characteristic data to obtain a connection health matrix;

[0026] In an embodiment of the present invention, based on the vibration-signal characteristic spectrum of the connection point, a non-invasive TDR is periodically used to collect the signal reflection characteristics of the communication line under vibration, obtaining a reflection characteristic change spectrum. Combined with the collected key communication quality parameters, a mapping model is established between the TDR spectrum characteristics and the communication quality parameters to form reflection communication mapping data. Simultaneously, a set of response functions of the key communication quality parameters to the vibration characteristics is constructed, and the effects of charge and discharge current and temperature on these vibration response functions are further analyzed to obtain current response data and temperature effect data. The reflection communication mapping data, current response data, and temperature effect data are correlated and analyzed to generate a dynamic communication quality mapping table that can dynamically predict communication quality parameters. Based on the TDR spectrum and the dynamic communication quality mapping table, the reflection characteristics are converted into a raw impedance dataset. Vibration correlation analysis, current effect compensation, and electrochemical state adjustment are then performed to obtain electrochemically adjusted impedance data. Static-dynamic impedance calculation is performed on this data to obtain impedance stability data, and this information is ultimately integrated to form dynamic impedance characteristic data. Frequency band sensitivity is calculated for dynamic impedance characteristic data, critical vibration thresholds are determined, the interaction between current and temperature is analyzed, three-dimensional response data is constructed, and vibration tolerance is evaluated to generate an impedance sensitivity feature set. Combining the impedance sensitivity feature set with the dynamic impedance characteristic data, the health score or status of each connection point is calculated to construct a connection health matrix.

[0027] Step S3: Perform time series data accumulation and denoising on the connection health matrix to obtain denoised health data; perform micro-change pattern recognition on the denoised health data to obtain a micro-change feature set; perform degradation pattern evolution analysis on the micro-change feature set to obtain a connection reliability prediction map;

[0028] In this embodiment of the present invention, the health time series of each connection point in the connection health matrix is ​​accumulated and de-noised using filtering, smoothing, and other methods to generate denoised health data. A sliding window is applied to the denoised health data to calculate health trend characteristics (such as slope and standard deviation) within the window, generating trend analysis results. The temporal evolution of trend characteristics (fluctuation) is analyzed and correlated with operating conditions and temperature to quantify operating condition sensitivity and temperature sensitivity data. Simultaneously, a change point detection algorithm is applied to the health time series and trend / fluctuation data to identify and extract mutation point feature data. The temperature sensitivity data and mutation point feature data are correlated with the system operating cycle to generate cyclic correlation data. Finally, the mutation point feature data and cyclic correlation data are integrated to generate a micro-change feature set that quantifies the subtle change patterns in the health of the connection point. Using the micro-change feature set, a pre-trained classification model is used to identify the current degradation pattern of the connection point. Based on the identified degradation pattern, a corresponding mathematical degradation model is selected, and the model parameters are fitted and calibrated using the historical denoised health data of the connection point. The model parameters are further modified to account for the impact of the energy storage unit's aging state and expected usage patterns on the degradation rate. Using the calibrated and modified degradation model, starting from the current health value, the mean health value and uncertainty range of the connection point are extrapolated forward for a period of time to generate degradation trajectory prediction data. Based on the degradation trajectory prediction data and its uncertainty, combined with the preset failure threshold, the probability of failure of the connection point at each future time point is calculated, generating failure probability time series data. The failure probability time series data and the degradation trajectory prediction data are combined to generate a connection reliability prediction map.

[0029] Step S4: performing communication path transfer analysis based on the connection reliability prediction map to obtain a path transfer strategy; executing the path transfer strategy and calculating the communication resilience index;

[0030] In this embodiment of the present invention, based on the connection reliability prediction map obtained in step S3, the predicted failure probability of each communication connection point within a specific future time window is obtained. The energy storage system communication topology is modeled as a graph, and the predicted failure probability is used as the unreliability of the edge. Topological reliability assessments, such as critical path reliability calculation and cut point / cut edge identification, are performed to generate a topology reliability map. Traffic data of communication links is acquired through real-time network monitoring, and the distribution, size, priority, and requirements of different service flows are analyzed to generate a traffic distribution characteristics table. Combining the traffic distribution characteristics table with the topology reliability map, a list of high-risk connection points that carry critical services and have a high predicted failure probability is identified. For these high-risk connection points, alternative communication paths with high reliability and sufficient capabilities are searched in the network topology. Their available bandwidth and latency are evaluated to form a backup resource capacity table. Based on the list of high-risk connection points and the backup resource capacity table, one or more optimal backup paths are matched for each high-risk connection point to form a path matching plan. Based on the traffic distribution characteristics table, a plan is used to distribute traffic on the high-risk links to the matched backup paths to generate a traffic distribution plan. Based on the risk level and traffic type, combined with real-time monitoring indicators, trigger conditions and priorities are planned for each traffic transfer operation to form a transfer time plan. The potential electrical impact of the path switching operations involved in the path matching solution is assessed to produce an electrical impact assessment report. Finally, based on the electrical impact assessment report, the transfer time plan and traffic allocation plan are adjusted and integrated to form a path transfer strategy plan with different priorities and trigger mechanisms. This strategy plan is deployed and executed, and the switch execution log is recorded during each execution. The switch execution log and network performance data before and after the transfer are analyzed to evaluate the actual effectiveness of the strategy and potential issues, and an optimization effect evaluation report is produced. Based on the quantitative indicators in the optimization effect evaluation report (such as event avoidance rate, transfer success rate, performance impact, etc.), the communication resilience index of the energy storage system is calculated.

[0031] Preferably, the step S1 includes:

[0032] Step S11: deploying piezoelectric micro-vibration sensors at the module connections of the energy storage system to synchronously collect signal quality parameters of the communication link under different working conditions to form an original vibration-communication data set;

[0033] Step S12: performing working condition correlation data preprocessing on the original vibration-communication data set to obtain a preprocessed working condition data packet;

[0034] Step S13: extracting a time domain feature vector from the pre-processed working condition data packet;

[0035] Step S14: performing frequency domain feature analysis on the pre-processed working condition data packet to obtain a frequency domain feature spectrum;

[0036] Step S15: constructing vibration-communication coupling features based on the time domain feature vector and the frequency domain feature spectrum to obtain a coupling feature set;

[0037] Step S16: Generate a connection point vibration-signal characteristic spectrum according to the coupling characteristic set.

[0038] In an embodiment of the present invention, miniature piezoelectric vibration sensors are closely mounted at the electrical connection points and communication cable connectors between each battery module in the energy storage system. For example, a PZT-5H piezoelectric ceramic sensor, measuring 5 mm × 5 mm × 1 mm, is used and fixed near the connection points using epoxy resin. Simultaneously, signal quality data for the communication link at the corresponding connection point is collected synchronously under different operating conditions via the energy storage system's integrated communication interface (e.g., a communication management unit based on the CAN bus or Ethernet). This data includes, but is not limited to, signal strength (RSSI), packet loss rate, communication latency, and bit error rate (BER). System operating conditions are captured by the battery management system (BMS), covering the charging phase (e.g., constant current charging, constant voltage charging), the discharging phase (e.g., constant current discharge, constant power discharge), the idle state, various ambient temperatures (e.g., -10°C, 25°C, 40°C), and internal system vibration levels (e.g., micro-vibrations generated when the cooling fan is activated). All collected vibration data and communication quality data are time-stamped with high precision to ensure data synchronization, forming a raw vibration-communication dataset that is stored in a local data storage unit, such as a solid-state drive array.

[0039] The raw vibration-communication data set is preprocessed. First, the vibration sensor data is bandpass filtered (for example, from 0.1 Hz to 1000 Hz) to remove ambient noise and DC offset. Outlier detection is performed on the communication quality data, for example, using an interquartile range (IQR)-based method to identify and mark data points outside the normal range. Then, the vibration and communication quality data streams are precisely aligned based on their timestamps. Missing data points due to sampling rate differences or brief data interruptions are handled using linear interpolation or nearest neighbor interpolation. Finally, the aligned data is divided into different operating condition data segments based on the operating condition information recorded by the BMS. For example, all data from the "constant current charging, temperature 25°C, fan running" operating condition is combined into one data packet; data from the "constant power discharging, temperature 40°C, idle vibration" operating condition is combined into another data packet. These data segments, divided by operating condition, constitute the preprocessed operating condition data packets.

[0040] A time-domain feature vector is extracted from each preprocessed operating condition data packet. A fixed-length sliding window (e.g., a window length of 1 second and a step size of 0.5 seconds) is applied to the vibration time series data in each operating condition data packet. Within each window, the time-domain statistical features of the vibration signal are calculated, including the root mean square (RMS) value, peak value, variance, skewness, and kurtosis. Simultaneously, statistical features such as the average, maximum, minimum, and standard deviation are calculated for communication quality parameters (RSSI, packet loss rate, latency, and BER) within the same time window. The vibration time-domain features and communication quality time-domain features calculated within the same window are combined into a high-dimensional vector, namely the time-domain feature vector. For example, a time-domain feature vector may contain [RMS vibration, peak vibration, variance vibration, average RSSI, maximum packet loss rate, and average latency]. After applying sliding window processing to the entire operating condition data packet, a collection of time-domain feature vectors is obtained.

[0041] Frequency domain feature analysis is performed on the vibration time series data in each preprocessed operating condition data packet. A fast Fourier transform (FFT) is applied to the vibration data to calculate its power spectral density (PSD). By analyzing the PSD, frequency domain features are extracted, such as the dominant frequency, the power of the dominant frequency, the total power within a specific frequency band (e.g., 10Hz-50Hz, 100Hz-200Hz), and the spectral centroid. These frequency domain features reflect the energy distribution of the vibration signal across different frequency components. These frequency domain features are organized into a frequency domain feature map. For example, this could be a matrix with rows representing time windows and columns representing extracted frequency domain feature values, or a two-dimensional map showing the variation of the power spectral density over time (i.e., a spectrogram). While communication signal quality itself is typically not subjected to traditional frequency domain analysis, analysis of its fluctuation frequency or periodicity can be used as supplementary frequency domain information, such as calculating the periodicity of packet loss rate fluctuations.

[0042] Based on the time-domain feature vectors and frequency-domain feature maps, coupling features are constructed to reflect the mutual influence between vibration and communication. For each time window or operating data segment, the correlation between vibration time-domain features (such as RMS vibration) and communication quality time-domain features (such as average packet loss rate) is analyzed. For example, the Pearson correlation coefficient can be calculated. The relationship between specific vibration frequency-domain features (such as the power of a dominant frequency) and communication quality parameters (such as RSSI) is analyzed. For example, how does RSSI change when the vibration power at a specific frequency increases? A regression model can be constructed to attempt to predict communication quality parameters (dependent variables) based on vibration features (independent variables). Extracted coupling features include, but are not limited to: the sensitivity of vibration intensity to packet loss rate (for example, packet loss rate / RMS vibration), the impact of specific vibration frequencies on communication latency, and the correlation between vibration energy and BER in key frequency bands. These features, which quantify how vibration affects communication, constitute the coupling feature set.

[0043] Based on the coupling feature set, a vibration-communication feature spectrum is generated for the connection point. Coupling feature sets extracted from the same connection point under different operating conditions (from different preprocessed operating condition data packets) are integrated and structured. A vibration-communication feature spectrum is not a simple list, but rather a comprehensive graph or model describing the vibration-communication coupling characteristics of the connection point under various vibration and operating conditions. For example, a multidimensional feature space can be constructed, with each dimension representing a key coupling feature (e.g., "the impact of 100Hz vibration on packet loss rate under charging conditions," "the impact of RMS vibration on latency under discharging conditions," etc.). These coupling feature values ​​calculated under different operating conditions serve as the coordinates of the connection point in this feature space. The vibration-communication feature spectrum can be a point cloud, a probability distribution model, or a predefined template in this feature space, populated with the specific coupling feature values ​​of the connection point under different operating conditions. It provides a comprehensive "portrait" of the communication behavior of the connection point under the influence of vibration.

[0044] Of particular importance is the construction of the vibration-communication coupling feature:

[0045] Calculate the relevant parameter table of time domain feature vector and frequency domain feature spectrum;

[0046] Perform spectrum correlation analysis on the relevant parameter table and frequency domain characteristic spectrum to obtain spectrum mapping data;

[0047] Calculating the sensitivity coefficients of the spectrum mapping data to obtain a sensitivity coefficient set;

[0048] Perform electrochemical state correlation on the sensitivity coefficient set to obtain electrochemical coupling data;

[0049] Perform current pulsation influence analysis on electrochemical coupling data to obtain current pulsation influence data;

[0050] Couple the current pulsation impact data with temperature factors to obtain temperature coupling data;

[0051] Perform coupling feature integration on temperature coupling data, current pulsation impact data and electrochemical coupling data to obtain a coupling feature set;

[0052] In this embodiment of the present invention, a data set containing time domain feature vectors (e.g., RMS vibration, peak vibration, average RSSI, average packet loss rate) and a data set containing frequency domain feature spectrum data (e.g., 100 Hz vibration energy, 200 Hz vibration peak) of a corresponding time window are obtained. For each connection point and each specific working condition data packet, the correlation between the time domain features and the frequency domain features is calculated. For example, the Pearson correlation coefficient between RMS vibration and average packet loss rate is calculated. , calculate the Pearson correlation coefficient between 100Hz vibration energy and communication delay . Linear regression analysis can also be performed, for example, the average bit error rate of the fitting model = + 200Hz vibration peak, record regression coefficient Calculate the correlation coefficient, regression coefficient, and determination coefficient for all time domain features of interest (vibration and communication quality) and all relevant frequency domain vibration feature pairs. These calculated parameter values ​​are organized into a table. The rows of the table can represent different feature pairs (for example, [RMS vibration, average packet loss rate], [100Hz energy, communication delay]), and the columns can contain correlation coefficients, regression coefficients, This table is the relevant parameter table. Using the relevant parameter table and frequency domain characteristic spectrum data, analyze the entries in the relevant parameter table that involve frequency domain characteristics and associate specific vibration frequencies or frequency bands with changes in communication quality parameters. For example, if the relevant parameter table shows that under certain working conditions, there is a high regression coefficient between 100Hz vibration energy and communication delay , and there is a high regression coefficient between the 200Hz vibration peak and the average bit error rate , then association rules such as "100Hz vibration is related to communication delay" and "200Hz vibration is related to average bit error rate" are identified. For each main frequency component or preset frequency band in the frequency domain feature map, the degree of influence of the vibration of this frequency / frequency band on different communication quality parameters is quantified according to the information in the relevant parameter table. This can be achieved by creating a mapping structure, such as a dictionary or lookup table: {vibration frequency / frequency band: {communication parameter 1: influence degree 1, communication parameter 2: influence degree 2,...}}. The degree of influence can be represented by a correlation coefficient, a regression coefficient or its absolute value. This mapping structure constitutes the spectrum mapping data, which describes how the vibration spectrum is specifically mapped to changes in communication quality. Utilize the spectrum mapping data. Quantify the "degree of influence" in the spectrum mapping data as a standardized sensitivity coefficient. For example, if the spectrum mapping data records that every increase of 1 unit (for example, dB / Hz) in the 100Hz vibration energy will result in an increase of m milliseconds in communication delay (m is the regression coefficient ), then m is defined as the sensitivity coefficient of 100Hz frequency band vibration to communication delay. For all vibration frequencies / frequency bands identified in the spectrum mapping data and the association with communication quality parameters, the corresponding sensitivity coefficient is calculated. The sensitivity coefficient can be defined as the unit change in the communication quality parameter caused by a unit vibration amount (e.g., unit RMS value, unit energy density). For example, =ΔCommunication parameters / Δ vibration characteristics , where i represents the vibration characteristics (such as 100Hz energy) and j represents the communication parameters (such as delay). Calculate the The calculated sensitivity coefficient values ​​constitute the sensitivity coefficient set, which is a quantitative indicator that directly reflects the sensitivity of the connection point communication quality to different vibration frequency components. Using the sensitivity coefficient set and the synchronously collected electrochemical state data of the energy storage system (for example, SOC, SOH), we analyze how the sensitivity coefficient changes with the change of the electrochemical state. For example, we focus on the sensitivity coefficient of 100Hz vibration to communication delay. Collect the values ​​of this coefficient calculated at different SOC and SOH levels. Build a model to describe this relationship, for example, using a multiple regression model: S =p +p ·SOC+p ·SOH+p ·SOC·SOH. The coefficient p is determined by fitting historical data ,p ,p ,p For all coefficients in the sensitivity coefficient set, correlation analysis and modeling with SOC and SOH are performed. These relationship models or parameter sets that describe how the sensitivity coefficients are modulated by the electrochemical state constitute the electrochemical coupling data. For example, the electrochemical coupling data includes: {sensitivity coefficient S Model:[p ,p ,p ,p ], sensitivity coefficient S Model:[q ,q ,q ,q ],...}. Using the electrochemical coupling data and the synchronously collected charge and discharge current data, especially the current pulsation characteristics (e.g., the RMS value and dominant frequency of the current ripple), analyze how the current pulsation affects the relationship between the sensitivity coefficient and the electrochemical state. For example, in step S15, S =p +p ·SOC+p ·SOH+p ·SOC·SOH model. Now, the current ripple characteristic I_ripple is introduced into the model as another influencing factor: S =p +p ·SOC+p ·SOH+p ·SOC·SOH+ p I_ripple+p SOC I_ripple. Quantify the direct effect of current ripple on the sensitivity coefficients and their interaction with the electrochemical state by fitting a model that includes current ripple terms and interaction terms. Repeat this process for all sensitivity coefficients. These relationship models or their parameter sets that describe how the sensitivity coefficients are modulated by current ripple (through interaction with the electrochemical state) constitute the current ripple impact data. For example, the current ripple impact data includes: {sensitivity coefficient S Model extension parameters: [p ,p ], sensitivity coefficient S 23 Model extension parameters: [q ,q ],...}.

[0053] Using the current pulsation impact data and the temperature data collected simultaneously, we analyze how temperature affects the relationship between the sensitivity coefficient and the electrochemical state and current pulsation. For example, in step S15, we establish a sensitivity model that includes SOC, SOH, and current pulsation terms. Now, we introduce temperature T as an influencing factor: S =p +...+p ·SOC·I_ripple+p ·T+ p ·T·SOC+p ·T·I_ripple+p T·SOC·I_ripple. Quantify the direct effect of temperature on the sensitivity coefficients and their interactions with the electrochemical state and current ripple by fitting a model that includes the temperature term and all relevant interaction terms. Repeat this process for all sensitivity coefficients. These relationship models or their parameter sets that describe how the sensitivity coefficients are modulated by temperature (through interactions with the electrochemical state and current ripple) constitute the temperature-coupled data. For example, the temperature-coupled data includes: {sensitivity coefficient S The final parameters of the model: [p ,...,p ],...}. Electrochemical coupling data, current pulsation impact data, and temperature coupling data are integrated. Together, these data describe how the sensitivity of the connection point's communication quality to vibration (at different frequency bands) is affected by the combined effects of electrochemical state, current pulsation, and temperature. A coupling feature set is a comprehensive, structured data set that includes all model parameters or key indicators that describe these complex relationships. For example, for each connection point, the coupling feature set can be a vector or data structure containing the following elements: [predicted values ​​of all sensitivity coefficients under nominal operating conditions (e.g., 25°C, 50% SOC, low current pulsation), the rate of change of all sensitivity coefficients with SOC, the rate of change of all sensitivity coefficients with current pulsation, the rate of change of all sensitivity coefficients with temperature, the coefficient of the interaction term between SOC and temperature for a key sensitivity,...]. This integrated feature vector comprehensively quantifies the vibration-communication coupling characteristics of the connection point under various operating environments, thus constituting the coupling feature set.

[0054] Preferably, the vibration response mapping process in step S2 is specifically as follows:

[0055] Collect key quality parameters of the communication link;

[0056] Construct vibration-communication response functions for key quality parameters and obtain a vibration response function set;

[0057] A reflection-communication mapping relationship is established based on key quality parameters and reflection characteristic change maps to obtain reflection-communication mapping data;

[0058] Analyze the charging and discharging current changes of the vibration response function set to obtain current response data;

[0059] Quantify the temperature effect of the vibration response function set to obtain temperature effect data;

[0060] A three-dimensional correlation data analysis is performed on the reflection communication mapping data, current response data and temperature effect data to obtain a communication quality dynamic mapping table.

[0061] In this embodiment of the present invention, during the operation of the energy storage system, a monitoring unit deployed on the communication module or communication cable continuously collects key quality parameters of the communication link. These parameters include, but are not limited to, signal strength (RSSI, in dBm), packet loss rate (Packet Loss Rate, in %), communication latency (Latency, in milliseconds), and bit error rate (BER). The collection frequency is set to 100 times per second, synchronized with the vibration data collection frequency to ensure data time consistency. The collected raw data stream is accurately timestamped and stored in the same local storage system as the vibration data.

[0062] Collect and preprocess vibration data (e.g., RMS vibration value, vibration energy in a specific frequency band) and simultaneously collect key communication quality parameters. Use these paired data as training samples. For each key communication quality parameter, construct one or more mathematical models to describe how it changes with changes in vibration characteristics. For example, for packet loss rate, a nonlinear regression model can be constructed: packet loss rate = f(RMS_vibration, vibration_100Hz_energy), where f represents a mapping function, such as a polynomial function or exponential function. For communication delay, another model can be constructed: delay = g(RMS_vibration, vibration_dominant frequency). Using the least squares method or other optimization algorithm, fit the coefficients in the model using the collected data to obtain the optimal response function. Once constructed, the set of mathematical models established for different key quality parameters (packet loss rate, delay, BER, etc.) constitutes the vibration response function set. For example, this function set includes: packet loss rate = 0.01 × RMS_vibration +0.5×vibration_100Hz_energy+0.02 and delay=1.2×RMS_vibration+0.1×vibration_dominant_frequency+5.

[0063] At the connection point of a communication link, a non-intrusive time domain reflectometer (TDR) is used to periodically measure the reflection characteristics of the communication cable, generating a reflection characteristic variation profile. This profile typically displays a waveform showing the reflection coefficient varying over time (corresponding to cable distance). This waveform is analyzed to extract features reflecting the connection point status, such as the amplitude, width, or waveform distortion of the reflection peak at the time delay corresponding to the connection point. These TDR features are correlated with key communication quality parameters collected at the same time or under similar operating conditions. For example, an increase in the reflection peak amplitude at a connection point is associated with an increase in packet loss. A mapping model is developed that takes the TDR profile features as input and predicts the corresponding key communication quality parameter values ​​as output. This can be trained using machine learning models such as support vector regression (SVR) or shallow neural networks. The training data consists of TDR profile features and corresponding communication quality parameters under different connection states (either simulated or actual degradation conditions). The trained model can predict the key communication quality parameters of the communication link based on real-time TDR profiles. These predictions constitute the reflection communication mapping data.

[0064] The vibration response function set is used to further analyze the impact of charge and discharge current on these functional relationships. The real-time charge and discharge current data of the battery module where the connection point is located is obtained from the energy storage management system (BMS). The construction process of the vibration response function set is refined, considering the current as another input variable. For example, the data is grouped according to different current ranges (such as: charging current <10A, 10A≤charging current <50A, discharging current <10A, 10A≤discharging current <50A, etc.), and the vibration response function is fitted separately in each current group. Alternatively, a joint response function that includes the current term is constructed, for example: packet loss rate = a×RMS_vibration +b×vibration_100Hz_energy+c×|current|+d×RMS_vibration×|current|+e. By fitting a model that includes the current term, we quantify how current changes affect vibration's impact on communication quality (i.e., the change in the function coefficient). The resulting current response data describes how communication quality's sensitivity to vibration changes at different charge and discharge currents.

[0065] Similar to analyzing current changes, analyze the impact of ambient temperature on the vibration response function set. Obtain real-time temperature data from a temperature sensor at the connection point or its surroundings. Introduce temperature as another input variable when constructing the vibration response function. For example, group the data by temperature range (e.g., temperature < 0°C, 0°C ≤ temperature < 20°C, 20°C ≤ temperature < 40°C, temperature ≥ 40°C) and fit a vibration response function within each temperature group. Alternatively, construct a joint response function that includes a temperature term, such as: delay = f × RMS_vibration + g × vibration_dominant frequency + h × temperature + i × RMS_vibration × temperature + j. By fitting a model that includes a temperature term, quantify how temperature changes alter the impact of vibration on communication quality. The analyzed temperature effect data describes how the sensitivity of communication quality to vibration varies at different temperatures.

[0066] The acquired reflective communication mapping data (communication quality predicted by TDR), current response data (describing how current modulates the vibration-communication relationship), and temperature effect data (describing how temperature modulates the vibration-communication relationship) are integrated. The goal is to create a comprehensive model or lookup table that can dynamically predict key quality parameters of the communication link (such as packet loss rate and latency) given TDR characteristics, current vibration state, current charge and discharge current, and current temperature. This data is used as multi-dimensional input and the actual collected communication quality parameters as output to train a multi-input, single-output, or multi-output model. For example, a multivariate regression model or a feedforward neural network can be constructed, with the input layer receiving TDR feature vectors, vibration feature vectors, current values, and temperature values, and the output layer predicting packet loss rate, latency, and other parameters. By training on data collected under a large number of different operating conditions, the model learns the complex nonlinear relationship between these input variables and communication quality. The trained model is the dynamic communication quality mapping table (or its logical representation), which dynamically outputs predicted values ​​for key quality parameters critical to the health of the communication link based on real-time environmental and TDR monitoring data.

[0067] Preferably, the dynamic impedance calculation and calibration in step S2 is specifically as follows:

[0068] Perform basic impedance conversion on the reflection characteristic change map to obtain the original impedance data set;

[0069] Perform vibration correlation analysis on the original impedance data set to obtain vibration-impedance relationship data;

[0070] Perform current influence compensation on the vibration-impedance relationship data to obtain corrected impedance data;

[0071] performing electrochemical state adjustment on the corrected impedance data to obtain electrochemically adjusted impedance data;

[0072] Perform static-dynamic impedance calculation on the electrochemically adjusted impedance data to obtain impedance stability data;

[0073] The electrochemically adjusted impedance data and the impedance stability data are subjected to dynamic impedance characteristic integration to obtain dynamic impedance characteristic data.

[0074] In this embodiment of the present invention, a non-invasive time domain reflectometer (TDR) is used to measure the communication cable and obtain a reflection characteristic change spectrum at the connection point. This spectrum is a function of the reflection coefficient ρ over time τ (corresponding to the cable length), namely ρ(τ). According to transmission line theory, the instantaneous impedance Z(τ) at the connection point is proportional to the characteristic impedance Z of the cable. There is a relationship between Z(τ) and the reflection coefficient ρ(τ): Z(τ)=Z (1+ρ(τ)) / (1-ρ(τ)). Assume that the characteristic impedance of the communication cable is Z It is known that, for example, it is 50Ω or 75Ω. By traversing each time point τ in the TDR spectrum (corresponding to a specific position at the connection point), measure the reflection coefficient ρ(τ ), and then use the above formula to calculate the impedance Z(τ The TDR waveform within a specific length range near the connection point is processed to calculate a series of impedance values ​​that vary over time. These calculated instantaneous impedance values ​​constitute the raw impedance dataset, reflecting the basic impedance characteristics of the connection point at the time of measurement.

[0075] The vibration data collected synchronously with the TDR measurement (for example, the RMS vibration value at the connection point location, the power at a specific vibration frequency, etc.) is time-aligned with the calculated raw impedance data set. The amplitude or pattern of change in the raw impedance of the connection point at different vibration levels and vibration frequencies is analyzed. For example, the deviation ΔZ between the instantaneous value of the raw impedance and a certain reference impedance value (for example, the average impedance at rest without vibration) is calculated. Then, a mathematical relationship model is established between ΔZ and the vibration characteristics at the same moment. This can be achieved using a linear regression model: ΔZ = , where V Represents different vibration characteristics (such as RMS vibration, 100Hz vibration power), are the coefficients to be fitted, Is the error term. The coefficients are obtained by fitting the collected multiple sets of vibration-impedance data using the least squares method. The values ​​of are used to quantify the influence of different vibration characteristics on the impedance change of the connection point. These fitted model coefficients, correlation indicators (such as the correlation coefficient), and the functional relationship describing the influence of vibration on impedance constitute the vibration-impedance relationship data.

[0076] Obtain real-time charge and discharge current data for the battery module at the connection point from the energy storage management system (BMS) at the time of TDR measurement. Analyze the vibration-independent changes in impedance at the connection point when different currents flow through it (for example, increased contact resistance due to Joule heating or changes in contact pressure due to electromagnetic forces). Develop a mathematical model between current and impedance changes, for example: ΔZ_Current = ,in is the current value, is the coefficient to be fitted. The model is fitted using impedance data collected under different currents (preferably static or low vibration conditions). Then, for each data point Z_raw in the original impedance data set collected under vibration conditions, the current value at the same moment is calculated. Calculate the estimated impedance change due to current, ΔZ_current. Subtract this estimated value from the original impedance to obtain the corrected impedance value: Z_corrected = Z_original - ΔZ_current. The resulting Z_corrected data is the corrected impedance data, which, to a certain extent, eliminates the direct effect of current on impedance and better reflects impedance changes caused by other factors such as vibration.

[0077] Obtain the electrochemical state parameters of the battery module where the connection point is located, such as state of charge (SOC) and state of health (SOH). Analyze how the impedance of the connection point (especially the connection directly connected to the battery terminal) changes with changes in SOC and SOH. For example, the contact resistance of the battery terminal is affected by the surface oxide layer or the state of the active material, which are related to the electrochemical state. Develop a mathematical model to describe the relationship between electrochemical state and impedance change: ΔZ_electrochemical = + ·SOC+ ·SOH+ ·SOC·SOH, where are the coefficients to be fitted. The model is fitted using impedance data collected at different SOC and SOH conditions (preferably static, low-vibration, and low-current conditions). Then, for each data point Z_correction in the corrected impedance dataset, an estimate of the impedance change due to the electrochemical state, ΔZ_electrochemical, is calculated based on the SOC and SOH values ​​at that moment. This estimate is subtracted from the corrected impedance to obtain the electrochemically adjusted impedance value: Z_electrochemicaladjustment = Z_correction - ΔZ_electrochemical. The resulting Z_electrochemicallyadjusted data is the electrochemically adjusted impedance data, which further eliminates the systematic effects of the electrochemical state on the junction impedance.

[0078] Analyze the time series data in the electrochemically tuned impedance dataset. Calculate the average value of this time series data and use it as the "static" or baseline impedance value, Z_static, for the junction under that operating condition. Calculate the standard deviation, σ_Z, of this time series data to reflect the degree of impedance fluctuation around the average value. Calculate the peak-to-peak value (the difference between the maximum and minimum values) or a specific percentile range (for example, the difference between the 95th and 5th percentiles) of the impedance value as an indicator of the magnitude of the dynamic impedance change, ΔZ_dynamic. These indicators, including Z_static, σ_Z, and ΔZ_dynamic, collectively describe the impedance stability and dynamic characteristics of the junction under specific operating conditions. The resulting set of values ​​constitutes the impedance stability data.

[0079] Integrate the electrochemically adjusted impedance data (i.e., the time-series impedance values ​​adjusted for current and electrochemical state) with the impedance stability data (including indicators such as Z_static, σ_Z, and ΔZ_dynamic). Dynamic impedance characteristic data is a structured data set that not only contains the average impedance (Z_static) of the connection point under specific working conditions, but more importantly, it also contains its changing characteristics under dynamic conditions. For example, dynamic impedance characteristic data can be a vector or a data structure containing: [Z_static, σ_Z, ΔZ_dynamic, ]. Among them, Z_static, σ_Z, ΔZ_dynamic are directly derived from the impedance stability data; and the coefficient The dynamic impedance characteristic data is derived from the vibration correlation analysis, current impact model, and electrochemical state model previously described. They quantify the impedance sensitivity to vibration, current, and electrochemical state. This integrated dynamic impedance characteristic data comprehensively describes the impedance behavior of the connection point under actual operating conditions, accounting for multiple influencing factors and their interactions.

[0080] Preferably, the impedance-vibration sensitivity analysis in step S2 is specifically as follows:

[0081] Perform frequency band sensitivity calculation on the dynamic impedance characteristic data to obtain frequency band sensitivity data;

[0082] Determine the critical vibration threshold value of the frequency band sensitivity data to obtain vibration critical threshold data;

[0083] Conduct current correlation analysis on the energy storage system based on the vibration critical threshold data to obtain current sensitivity data;

[0084] Perform temperature interaction effect analysis on current sensitivity data and frequency band sensitivity data to obtain the temperature impact matrix;

[0085] Construct three-dimensional response data based on frequency band sensitivity data, current sensitivity data and temperature impact matrix;

[0086] Performing vibration tolerance assessment on the three-dimensional response data and the vibration critical threshold data to obtain vibration tolerance data;

[0087] Generate an impedance sensitivity feature set based on frequency band sensitivity data and vibration tolerance data;

[0088] A connection health matrix is ​​constructed based on the impedance sensitivity feature set and dynamic impedance characteristic data.

[0089] In the embodiment of the present invention, dynamic impedance characteristic data is used, which includes a sequence of connection point impedance changes over time and coefficients describing the relationship between impedance and vibration characteristics (for example, the coefficients aᵢ obtained in dynamic impedance calculation and calibration, which relate the energy and impedance changes in a specific vibration frequency band). The absolute values ​​of these coefficients are analyzed. The larger the absolute value of the coefficient, the more significant the impact of the vibration frequency band on the connection point impedance. For example, if the dynamic impedance characteristic data contains coefficient a1 that relates the vibration energy and impedance changes in the 10Hz-50Hz frequency band, the coefficient aᵢ is used to describe the relationship between the impedance and the vibration characteristics. The vibration energy and impedance changes in the 100Hz-200Hz frequency band are correlated, and by comparing | | and | | determines the vibration frequency band to which the connection point is most sensitive. Impedance sensitivity values ​​are calculated for multiple preset vibration frequency bands (e.g., 0-50Hz, 50-150Hz, 150-300Hz, and bands near the dominant frequency). Sensitivity values ​​can be defined as the impedance change per unit vibration amount (e.g., unit energy or RMS value) or directly using the coefficients of the fitted model. The collection of sensitivity values ​​calculated for each frequency band constitutes the frequency band sensitivity data.

[0090] Based on the energy storage system's communication protocol requirements or historical fault data, a maximum tolerable threshold ΔZ_max for connection point impedance change is set. For example, the connection point impedance cannot deviate more than 5Ω from its steady-state state. Using frequency band sensitivity data, for each specific vibration frequency band, based on its sensitivity value S_band, the vibration level V_band_critical required to cause the impedance change to reach ΔZ_max is calculated. If sensitivity S_band is defined as ΔZ / V_band, then the critical vibration level V_band_critical = ΔZ_max / S_band. If the vibration-impedance relationship is a more complex function, the function must be solved to determine the vibration level at which ΔZ_max is reached. For example, if the relationship is a quadratic function, a quadratic equation is solved. The critical vibration energy or critical vibration RMS value for each frequency band of interest is calculated. These calculated critical vibration levels constitute the vibration critical threshold data, which indicates the upper limit of vibration at which the connection point impedance fails under vibration in a specific frequency band.

[0091] The modulation coefficient of the current-to-impedance or vibration-impedance relationship included in the dynamic impedance characteristic data (for example, the coefficient obtained in the dynamic impedance calculation and calibration in step S2) is used. Correlating current with impedance change, with coefficient d relating current to vibration-impedance interaction). Analyze how the critical vibration threshold determined in step S2 changes under different charge and discharge currents. For example, high current causes the connection point to heat up or generates additional electromagnetic forces, which changes its mechanical properties or contact resistance, making the connection point more sensitive to vibration, thus lowering the critical vibration threshold. Build a model to describe the relationship between the current value I and the critical vibration threshold V_critical: V_critical_adjusted = V_critical_base - k × |I|, where k is a coefficient obtained by analyzing and fitting the vibration and impedance data collected at different currents. Calculate the rate or functional relationship of the critical threshold as a function of current for each vibration frequency band. These data describing how the critical vibration threshold is affected by current constitute the current sensitivity data.

[0092] Utilize the modulation coefficient of temperature on impedance or vibration-impedance relationship contained in the dynamic impedance characteristic data (for example, the coefficient h obtained in the dynamic impedance calculation and calibration in step S2 relates to the temperature and impedance change, and the coefficient i relates to the temperature and vibration-impedance interaction). Analyze how temperature affects both the frequency band sensitivity data and the current sensitivity data. For example, low temperature makes the material brittle and increases the sensitivity to high-frequency vibration; high temperature causes the material to soften or expand, changing the contact pressure change caused by the current, thereby affecting the modulation effect of the current on the vibration sensitivity. Construct a model to quantify the effect of temperature T on the specific frequency band sensitivity S_band and current sensitivity C_current. For example, S_band_adjusted = S_band_base×(1+m×T), C_current_adjusted = C_current_base×(1+n×T), where m and n are fitting coefficients. More complex interaction effects can be represented by a matrix. For example, construct a matrix M whose rows represent different vibration frequency bands, columns represent different current ranges, and the matrix elements Indicates that at a specific temperature, the frequency band and current range Combine the impact factors on impedance or adjust the critical threshold. By collecting data under different temperature, current and vibration conditions and performing multivariate regression analysis, the quantitative values ​​of these interaction effects are determined to form a temperature impact matrix.

[0093] The frequency band sensitivity data, current sensitivity data, and temperature influence matrix are integrated into a multidimensional data structure. This structure describes the response characteristics of the connection point impedance or its critical vibration threshold under different vibration frequency bands, different charge and discharge currents, and different ambient temperature combinations. For example, a three-dimensional lookup table or a multidimensional function model can be constructed, whose input is {vibration frequency band, current value, temperature value}, and the output is the impedance sensitivity value or critical vibration threshold under the combined conditions. For example, a data point indicates that at 100Hz vibration, 50A charging current, and 40°C ambient temperature, the connection point impedance's sensitivity to vibration is a certain value, or its critical 100Hz vibration threshold is a certain acceleration value. This structured data set is the three-dimensional response data, which comprehensively reflects the joint sensitivity of the connection point impedance to major environmental and operating factors.

[0094] Utilizing three-dimensional response data and critical vibration threshold data, for typical operating conditions of the energy storage system (i.e., different current and temperature conditions), the critical vibration thresholds for each vibration frequency band under these conditions are found or calculated from the three-dimensional response data. Vibration tolerance refers to the maximum vibration level that a connection point can withstand under specific operating conditions without experiencing impedance failure. This can be a comprehensive indicator, such as the maximum overall RMS vibration that a connection point can withstand under specific current and temperature, or the maximum allowable vibration energy in a specific frequency band. The evaluation process involves combining the critical thresholds for each frequency band, taking into account that actual vibration often contains multiple frequency components. For example, based on the vibration cumulative damage model, the "equivalent" vibration load under a specific operating condition is calculated and compared with the critical threshold for that condition. A set of vibration tolerance indicators under different typical operating conditions is calculated to form the vibration tolerance data.

[0095] Key features are extracted from the frequency band sensitivity data and vibration tolerance data to form a compact feature vector. These features should comprehensively summarize the connection point impedance's sensitivity to vibration and its tolerance under different operating conditions. For example, the feature set includes: the most sensitive vibration frequency band, the highest sensitivity value in that frequency band, the overall RMS vibration tolerance under nominal operating conditions (e.g., 25°C, 0A current), the lowest overall RMS vibration tolerance under the worst-case operating conditions (e.g., highest temperature, maximum current), and the gradient of sensitivity with current or temperature. This carefully selected, representative set of values ​​constitutes the impedance sensitivity feature set.

[0096] Combine the impedance sensitivity feature set (reflecting the intrinsic sensitivity of the connection point to vibration) with the dynamic impedance characteristic data (reflecting the connection point's current static impedance, dynamic fluctuations, and response to current and electrochemical conditions). Using this combined information, calculate or assess the overall health of the connection point. This can be achieved using a pre-trained health assessment model, such as a classifier or regressor based on a support vector machine (SVM) or neural network. The model inputs are key values ​​from the impedance sensitivity feature set and the dynamic impedance characteristic data (e.g., Z_static, ΔZ_dynamic, σ_Z, sensitivity in the most sensitive frequency band, and minimum vibration tolerance). The output is a quantitative health score (e.g., from 0 to 100, with higher scores indicating better health) or a health status category (e.g., excellent, good, concern, and warning). Repeat this process for all communication connection points in the energy storage system, recording the health score or status of each connection point in a matrix. The rows of this matrix correspond to different connection points, and the columns contain the health score or health status as well as supporting indicators (e.g., static impedance value, dynamic fluctuation amplitude, and minimum vibration tolerance level). This matrix is ​​referred to as the connection health matrix.

[0097] Preferably, the micro-change pattern recognition in step S3 is specifically:

[0098] Determine the sliding window of denoised health data and perform trend feature statistics within the window to obtain trend analysis results;

[0099] Perform fluctuation characteristic evolution processing on trend analysis results to obtain fluctuation evolution data;

[0100] Perform working condition difference response analysis on the fluctuation evolution data to obtain working condition sensitivity data;

[0101] Conduct temperature sensitivity tracking on the working condition sensitivity data to obtain temperature sensitivity data;

[0102] Accurately identify mutation points based on trend analysis results and fluctuation evolution data to obtain mutation point feature data;

[0103] Performing cyclic correlation analysis on temperature sensitivity data and mutation point characteristic data to obtain cyclic correlation data;

[0104] A micro-change feature set is generated based on the cyclic correlation data and mutation point feature data.

[0105] In this embodiment of the present invention, a sliding window is applied to the denoised health data (which is a sequence of health scores for each connection point that varies over time, for example, one score per day). The window length is set to 7 days, with a step size of 1 day. For each connection point, the first 7 days of data from the start of its health time series are intercepted as a window. Within this window, a linear regression analysis is performed on these 7 health scores, using the fitted model: health = m × days + c, where days is the number of days from the start of the window, m is the slope, and c is the intercept. The slope m obtained from the fitting is used as the health trend feature within the window. Simultaneously, the mean, standard deviation, and head-to-tail difference (last day score minus first day score) of the health scores within the window are calculated as supplementary trend features. The window is then slid backward by 1 day, and the above calculations are repeated until the entire denoised health data sequence is covered. For each connection point, a series of trend feature values ​​that vary over time are obtained (for example, the slope, mean, standard deviation, and head-to-tail difference are calculated daily), which constitute the trend analysis results.

[0106] Using trend analysis results, analyze how the trend characteristics themselves change over time. For example, consider the time series of the health trend slope (m). Calculate the local standard deviation of the slope time series and process the slope series using a longer sliding window (for example, a window length of 30 days and a step size of 1 day). The slope standard deviation within the window reflects the volatility of the health trend. Calculate the time series of this volatility. Simultaneously, analyze the volatility of the health score standard deviation series (derived from the trend analysis results), for example by calculating its moving average or exponentially smoothed value and analyzing its rate of change. You can also calculate the variance of the rate of change of the trend slope or its mutation frequency. These indicators describing the temporal changes in the health trend volatility constitute the volatility evolution data.

[0107] Correlate the fluctuation evolution data with the energy storage system's operating condition information (e.g., charging, discharging, idle, different current levels, different load types) during the corresponding time period. Analyze how the health fluctuation characteristics differ under different operating conditions. For example, calculate the average standard deviation of the health trend slope under charging conditions and compare it with the average under discharging conditions. Quantify this difference, for example, by calculating the difference or ratio of the fluctuation indicators under different operating conditions. Build a model (e.g., analysis of variance ANOVA) to determine whether the impact of different operating conditions on the health fluctuation characteristics is statistically significant. These data, which quantify the responsiveness of the health fluctuation characteristics to different operating conditions, constitute the operating condition sensitivity data. For example, the data obtained show that the health fluctuation standard deviation under charging conditions is 15% higher than under discharging conditions.

[0108] Correlate the operating condition sensitivity data with the temperature data of the connection point or its surrounding environment. Analyze how temperature affects the sensitivity of the health fluctuation characteristics to different operating conditions. For example, in a low-temperature environment, is the health fluctuation under charging conditions more severe than in a high-temperature environment? Build a model to describe how temperature modulates the operating condition sensitivity. For example, if the operating condition sensitivity is expressed as the ratio of the standard deviation of the charge / discharge fluctuations, then establish a model ratio = p × temperature + q. Quantify the degree to which this sensitivity changes with temperature through regression analysis of the fitting coefficient p. Continuously track and record the changes in these temperature-related sensitivity coefficients or indicators over time. These data describing how the operating condition sensitivity is affected by temperature constitute the temperature sensitivity data.

[0109] Apply a change point detection algorithm to the denoised health time series, trend analysis results (such as slope series), and volatility evolution data (such as volatility standard deviation series). For example, use a penalized maximum likelihood method (such as the PELT algorithm) to detect time points in the series where the mean, variance, or slope changes significantly. Set a confidence level for change detection (e.g., 95%). The algorithm outputs a series of timestamps for the detected change points. For each detected change point, analyze the data characteristics before and after it to extract the mutation point characteristics. For example, record the time of the mutation, the average health value before and after the mutation, the mutation amplitude (the amount of change in health), the standard deviation of the health value before and after the mutation, and the trend slope before and after the mutation. For all detected mutation points, summarize their characteristic information to form mutation point feature data. For example, a mutation point feature data item can be: [timestamp: 2023-10-26 10:00, health before mutation: 85, health after mutation: 80, amplitude: -5, standard deviation before mutation: 1.2, standard deviation after mutation: 3.5, slope before mutation: -0.1, slope after mutation: -0.5].

[0110] Correlate temperature sensitivity data and mutation point characteristic data with the energy storage system's operating cycle information (e.g., number of complete charge / discharge cycles, daily charge / discharge time periods). Analyze whether changes in temperature sensitivity or the occurrence of mutation points are associated with specific operating cycles or cycle phases. For example, calculate the distribution of mutation points across the charge / discharge cycle to determine whether they tend to occur at the end of charge or discharge. Calculate the correlation between temperature sensitivity indicators (e.g., the coefficient p of temperature sensitivity to operating conditions) and the cumulative number of charge / discharge cycles. Analyze whether temperature sensitivity exhibits cyclical variations with seasonal temperature cycles. Use cyclic autocorrelation analysis or cross-correlation analysis to quantify the frequency of mutation points or the strength and phase of the correlation between temperature sensitivity and the operating cycle. These metrics, which quantify the relationship between micro-variations (including sensitivity changes and mutation points) and the system's operating cycle, constitute cyclic correlation data. For example, data shows that 80% of health mutations occur within 2 hours of the end of a discharge cycle, and that temperature sensitivity is linearly positively correlated with the cumulative number of cycles.

[0111] The cycle-related data and mutation point feature data are integrated to form a comprehensive feature vector describing the micro-change pattern of the connection points. Key statistics are extracted from the mutation point feature data, such as the total number of mutation points per unit time, the average mutation amplitude, the maximum mutation amplitude, and the proportion of health-degrading mutation points. Key correlation indicators are extracted from the cycle-related data, such as the correlation coefficient between the mutation point and the charge / discharge cycle stage, the correlation coefficient between temperature sensitivity and the cumulative number of cycles, and the presence of significant periodic changes. These selected, representative statistics and correlation indicators are combined into a feature vector. For example, the micro-change feature set can be a vector containing the following elements: [average number of mutation points per month, average health-degradation amplitude, strength of correlation with end of discharge, coefficient p for temperature sensitivity to operating conditions, correlation between cumulative number of cycles and temperature sensitivity]. This feature vector, known as the micro-change feature set, comprehensively summarizes the pattern of micro-changes in the health of the connection points and their relationship to the environment, operating conditions, and operating cycle.

[0112] Preferably, the degradation mode evolution analysis in step S3 is specifically as follows:

[0113] Perform degradation pattern classification on the micro-change feature set to obtain degradation pattern recognition results;

[0114] According to the degradation pattern recognition results, the degradation trajectory of the denoised health data is deduced to obtain the degradation trajectory prediction data;

[0115] Calculate the failure probability of the degradation trajectory prediction data to obtain the failure probability time series data;

[0116] Generate a connection reliability prediction map based on the failure probability time series data and degradation trajectory prediction data.

[0117] In this embodiment of the present invention, a micro-change feature set is used as input. This feature set contains a quantitative description of minor changes in the health of a connection point, such as mutation frequency, average mutation amplitude, sensitivity of fluctuation to operating conditions and temperature, and correlation with operating cycles. A multi-class classification model, such as a trained support vector machine (SVM) classifier, is constructed. This classifier is trained on historical data containing a large number of micro-change feature sets exhibited by connection points in actual operation and corresponding actual degradation patterns (for example, contact wear, bolt loosening, electrochemical corrosion, and transient disconnection caused by stress fatigue, as determined through post-inspection or fault analysis). During training, the classifier learns association rules between micro-change features and specific degradation patterns. The micro-change feature set of the connection point to be analyzed is input into the trained SVM classifier. The classifier outputs a classification result, indicating the most likely degradation pattern category for the connection point. For example, the output result may be the string "contact wear" or the integer code "1" representing bolt loosening. This classification result is the degradation pattern recognition result.

[0118] Utilize the degradation pattern identification results (e.g., "contact wear") and the past denoised health time series data of the connection points. Based on the identified degradation pattern, select or adjust a specific degradation prediction model. Different degradation patterns exhibit different evolution patterns; for example, "contact wear" exhibits a slow but continuous acceleration in health, while "bolt loosening" causes a step-like decrease in health. A mathematical model matching the identified pattern is selected to predict the health trajectory. For example, if "contact wear" is identified, an exponential decay model H(t) = A·exp(-λt)+B is used, where H(t) is the predicted health value at time t in the future, and A, B, and λ are model parameters. Using the denoised health data of the connection points over a recent period (e.g., the past 90 days), the model parameters A, B, and λ are fitted using the least squares method or other curve fitting algorithm to ensure that the model curve closely matches the historical data. The fitted model is extended forward to predict the health value series for a period of time in the future (e.g., the next 180 days). This predicted health value series constitutes the degradation trajectory prediction data.

[0119] Define a health threshold H_failure for a connection point to fail. For example, a health score below 20 is considered a connection point communication failure. Use degradation trajectory prediction data, which includes the predicted health value μ(t) for each time point t within a future period, as well as the prediction uncertainty (e.g., the standard deviation σ(t), where σ(t) typically increases with longer prediction times). Assume that at each time point t, the actual health value H(t) follows a normal distribution with mean μ(t) and standard deviation σ(t). Calculate the probability of a failure at each future time point t, P(failure|t), that is, the probability that the health value falls below the threshold H_failure. Based on the normal distribution's cumulative distribution function (CDF), Φ, this probability is P(failure|t) = Φ((H_failure - μ(t)) / σ(t)). Where Φ(z) is the CDF of the standard normal distribution. Perform this calculation for each time point within the prediction timeframe, resulting in a series of time-varying failure probability values. These probability value sequences constitute the failure probability time series data.

[0120] Integrate and visualize the failure probability time series data and degradation trajectory prediction data. Generate a two-dimensional graph with the horizontal axis representing future time (e.g., starting from the current date and extending backward) and the vertical axis representing the health score of the connection point (e.g., 0 to 100). Plot the degradation trajectory prediction data on the graph, displaying the predicted health mean curve μ(t). Also, plot the uncertainty range of the prediction, for example, by plotting the upper and lower bounds of μ(t) ± 2σ(t), forming a prediction interval band. Overlay the failure probability time series data on the graph, using a separate vertical axis or color depth to represent the magnitude of the failure probability. For example, draw a failure probability curve below the graph, or use a gradient color within the prediction interval band, with darker colors indicating higher failure probability. Clearly label the failure threshold line H_failure. This graph intuitively displays the future evolution of the connection point's health, the prediction uncertainty, and the time-varying failure risk. This graph is the connection reliability prediction graph.

[0121] What is particularly important is that the degradation trajectory deduction is specifically as follows:

[0122] Determine the degradation model parameter set according to the degradation pattern recognition result;

[0123] Perform short-term change prediction based on the degradation model parameter set and denoised health data to obtain short-term prediction data;

[0124] Calculating the impact of usage patterns based on the degradation model parameter set to obtain usage pattern correction data;

[0125] Perform energy storage unit aging coupling analysis based on the degradation model parameter set to obtain aging coupling data;

[0126] Generate degradation trajectory prediction data based on short-term prediction data, model correction data and aging coupling data;

[0127] In an embodiment of the present invention, the degradation pattern recognition result is used, for example, to identify that the degradation pattern of the connection point is "contact wear". For each predefined degradation pattern, a corresponding mathematical degradation model library is maintained. For example, the "contact wear" pattern corresponds to an exponential decay model: health H(t) = A × exp(-λ × t) + B, where t is time, A represents the initial health decline, B represents the stable health level, and λ represents the decay rate. The "bolt loosening" pattern corresponds to a step decline model plus slow linear decay: health H(t) = ,in is the initial health, is the step-down amplitude, is the time when loosening occurs, is the additional decay rate after loosening, is a unit step function. Based on the identified degradation pattern, the corresponding model structure is selected from the model library, and the typical or initial model parameter values ​​of the connection point under this pattern are loaded. These typical parameter values ​​can be obtained based on historical data statistics or accelerated life test results. For example, if the identified pattern is "contact wear", the typical initial values ​​of A, B, and λ under this pattern are loaded to form a degradation model parameter set, such as {A:15, B:70, λ:0.005}. Utilize the denoised health data (the historical health score time series of the connection point) and the degradation model parameter set (the loaded typical or initial model parameters). These initial parameter sets are universal and need to be calibrated based on the specific historical data of the current connection point to more accurately predict its short-term trend. Use the denoised health data of the connection point in the recent period (for example, the past 90 days) as training data. The degradation model selected in step S3 (e.g., the exponential decay model H(t) = A × exp(-λ × t) + B) is applied to these historical data. Using a nonlinear least-squares fitting algorithm, the model parameters A, B, and λ are adjusted to minimize the sum of squared residuals between the model curve and the historical health data points. The resulting optimized parameter values ​​(e.g., A*: 12, B*: 72, λ*: 0.006) better reflect the current degradation rate of the connection point than the initial parameters. Using these optimized parameters, the model is extrapolated forward to predict a series of health values ​​for a future period (e.g., the next 30 days). The prediction results include the predicted health mean and variance at each time point (reflecting the uncertainty of the prediction, which increases over time). This series of health means and variances for the next 30 days constitutes the short-term forecast data. This is done using the degradation model parameter set (here, the optimized parameters A*, B*, and λ*) and information about the expected future usage patterns of the energy storage system. Different usage patterns (e.g., high-frequency charge-discharge cycles, prolonged high-current operation, and continuous exposure to high vibration) can have varying impacts on the degradation rate of the connection point. Based on historical data analysis or physical models, a quantitative relationship is established between usage pattern characteristics (e.g., the expected number of charge-discharge cycles and the expected duration of high vibration over the next 180 days) and degradation model parameters (particularly the rate parameter λ). For example, the relationship is: λ_correction = λ* + α × (expected number of cycles / 1000) + β × (expected duration of high vibration / 100 hours), where α and β are coefficients derived from fitting historical data. Based on the energy storage system's planned or predicted usage patterns, the usage pattern characteristic values ​​for a future period (e.g., the next 180 days) are calculated and substituted into the above relationship to calculate the correction Δλ_usage_pattern to the degradation rate parameter λ due to the usage pattern. This correction Δλ_usage_pattern, or simply using the corrected λ value, constitutes the usage pattern correction data. Utilize the degradation model parameter set (using the optimized parameters A*, B*, λ*) and the current aging state (e.g., SOH value) of the energy storage unit (e.g., battery module) to which the connection point belongs.The degradation rate of a connection point is coupled to the aging of the connected battery module. For example, internal module expansion increases connection stress, accelerating mechanical wear; increased module internal resistance leads to uneven current distribution at the connection, increasing thermal stress. A quantitative relationship is established between the energy storage unit's aging state (e.g., SOH) and the parameters of the connection point degradation model (specifically, the rate parameter λ or the stable health level B). For example, the relationship is: λ_SOH correction = γ × (100% - SOH), where γ is a coefficient obtained by fitting historical data. Based on the currently monitored or predicted SOH value of the energy storage unit, a correction Δλ_SOH to the degradation rate parameter λ due to unit aging is calculated. This correction Δλ_SOH, or simply using the corrected λ value, constitutes the aging-coupled data. Short-term forecast data (the mean and variance of the health forecast for the next 30 days), usage pattern correction data (Δλ_usage pattern), and aging-coupled data (Δλ_SOH) are integrated to generate complete degradation trajectory prediction data. First, the predicted health value and model parameters at the end of the short-term forecast (e.g., day 30) are obtained from the short-term forecast data. Then, the usage pattern correction Δλ_usage pattern and the aging coupling correction Δλ_SOH are applied to the degradation rate parameter λ* at the end of the short-term forecast: λ_final = λ* + Δλ_usage pattern + Δλ_SOH. Using this finalized degradation rate parameter λ_final and the health value at the end of the short-term forecast as the new starting point, a selected degradation model (e.g., an exponential decay model) is applied to perform a forward extrapolation to predict the health value series from the end of the short-term forecast to the desired future forecast endpoint (e.g., 180 days into the future or until the health value falls below the failure threshold). When performing long-term forecasts, the cumulative forecast uncertainty is taken into account, and the variance of the predicted health value at each future time point is calculated. This variance typically grows nonlinearly with the forecast time. The short-term forecast data and the long-term extrapolated data are integrated to form a complete time series containing the mean and variance of the predicted health value at each time point in the future. This time series is the degradation trajectory forecast data.

[0128] Preferably, step S4 includes the following steps:

[0129] Step S41: performing communication topology reliability assessment based on the connection reliability prediction map to obtain a topology reliability map;

[0130] Step S42: Analyze the communication traffic characteristics based on the topology reliability map to obtain a traffic distribution characteristics table;

[0131] Step S43: Evaluate the backup path resources of the energy storage system based on the traffic distribution characteristic table and the topology reliability map to obtain a backup resource capacity table;

[0132] Step S44: formulating a path transfer strategy based on the topology reliability map, the traffic distribution characteristics table, and the backup resource capability table;

[0133] Step S45: Perform a switching execution mechanism on the path transfer strategy to obtain a switching execution log;

[0134] Step S46: Perform adaptability verification and optimization on the switch execution log to obtain an optimization effect evaluation report;

[0135] Step S47: Calculate the communication resilience index of the energy storage system according to the optimization effect evaluation report.

[0136] In an embodiment of the present invention, a connection reliability prediction graph is used, which displays the predicted health trajectory, uncertainty range, and corresponding failure probability time series data of each communication connection point in the energy storage system over a period of time in the future (for example, the next 30 days). The communication network of the energy storage system is represented as a directed graph G=(V,E), where V represents a communication node (for example, a main control unit, a battery management module, an energy conversion unit, etc.) and E represents a communication link or connection point. According to the connection reliability prediction graph, for each edge e∈E in the graph G, its "unreliability" or predicted failure probability P(e,t) is set at a specific time point t in the future or within a certain time period (for example, the next 7 days). Based on this weighted graph, reliability analysis at the topological level is performed. For example, critical communication paths are identified, such as the control signal path between the main control unit and each battery module. The end-to-end reliability of these critical paths is calculated, assuming that the failures of each connection point on the link are independent of each other, and the path reliability R_path=∏ ∈path(1-P(e Identify network cutpoints and cutedges—those whose failure could cause a network partition or disrupt communication between critical nodes. Visualize these analysis results, for example, by using color shading to indicate link unreliability on a network topology map, highlighting cutpoints and cutedges with special markers, and generating a report listing the reliability values ​​of critical paths and identified single-point failure risks. This visualization and report constitute a topology reliability map.

[0137] Using a topology reliability map, it identifies potential weak links in the network. Simultaneously, using energy storage system communication network monitoring tools, traffic data on each communication link is collected and recorded in real time. This traffic data is analyzed to quantify the distribution and characteristics of different types of communication (e.g., control commands, status collection, balancing data, and safety alerts) within the network. The paths along which critical service traffic (e.g., heartbeat signals between the BMS master and slaves, emergency shutdown commands) flows are tracked, and the average and peak bandwidth usage, packet transmission latency, and priority requirements of each link along these paths are determined. These analysis results are structured, for example, by creating a table where each row represents a communication link and columns include the link's identifier, predicted failure probability, average traffic (e.g., KB / s), peak traffic, the types of critical services carried, the priority of these services, and the corresponding latency and bandwidth requirements. This table, known as the traffic distribution characteristics table, describes the flow and importance of communication traffic within the current topology and is associated with link reliability information.

[0138] Combined with the traffic distribution characteristics table and the topology reliability map from step S41, the backup path resources of the energy storage system's communication network are evaluated. For communication links L listed in the traffic distribution characteristics table as carrying critical services and indicated by a high predicted failure probability in the topology reliability map, alternative paths for L are searched in the network graph G. Using a graph search algorithm (e.g., a modified Dijkstra algorithm that considers link reliability as a weight), a highly reliable alternative path connecting the nodes at both ends of link L or other relevant key nodes is searched, excluding L itself. For each found alternative path P_alt, its carrying capacity is evaluated. This involves calculating the available bandwidth of each link on path P_alt (the total bandwidth minus the current traffic flow, as determined in step S42). The minimum value is taken as the bottleneck bandwidth of the path. The total delay of path P_alt (the sum of the delays of each link) is evaluated to see if it meets the delay requirements for critical services. The analysis results are recorded in the backup resource capacity table. For example, the table includes: the original high-risk link ID, predicted failure probability, alternative path ID (consisting of a series of link IDs), calculated reliability of the alternative path (based on S41), available bandwidth of the alternative path, total latency of the alternative path, and whether it can carry critical services on the original link (yes / no). This table quantifies the available backup communication paths in the network that can replace the high-risk primary path and their capabilities.

[0139] Based on the topology reliability map from step S41, the traffic distribution characteristics table from step S42, and the backup resource capability table from step S43, a communication path migration strategy is developed. A list of communication links with predicted failure probabilities exceeding a preset threshold (e.g., failure probability >15% within the next 7 days) is identified. For these high-risk links, the traffic distribution characteristics table is referenced to distinguish the types of services they carry (critical or non-critical). A migration strategy is prioritized for high-risk links carrying critical services. For each high-risk link L requiring traffic migration, the backup resource capability table is consulted to identify the alternative path P_alt with the best matching capabilities (e.g., highest reliability, sufficient available bandwidth, and latency requirements). Trigger conditions for path migration are defined (e.g., real-time monitoring of the packet loss rate on link L exceeding 5%, or communication latency exceeding 100ms, or a predicted failure probability reaching 20%). A migration action is defined: when the trigger condition is met, specific service traffic or all traffic originally flowing through L is switched to P_alt. These trigger conditions and migration actions are combined to form a policy rule set. For example, a rule could be: "IFLink_ABC predicted 7-day failure probability > 20% OR Link_ABC real-time packet loss rate > 5% THEN REROUTE Critical_Traffic_Type_1 via Path_AD-DE-EC." This policy rule set is the path transfer policy solution.

[0140] Execute the path transfer strategy developed in step S44. A policy execution module is deployed in the energy storage system's communication network. This module continuously monitors the real-time status of communication links (for example, by receiving metrics such as packet loss rate and latency reported by the communication modules) and obtains the latest connection reliability prediction map. When the monitored real-time status or predicted failure probability meets a trigger condition defined in the policy scheme, the policy execution module immediately executes the corresponding transfer action. This transfer action involves sending configuration instructions to relevant communication nodes (for example, routers, switches, or intelligent communication modules) to modify their routing tables or traffic forwarding rules, redirecting the affected traffic to a designated alternative path. For example, if a policy trigger requires that traffic from module A to module C be transferred from link ABC to path AD-DE-EC, the policy execution module will send the corresponding configuration command to modules A, D, E, and C. Each time a path transfer is triggered and executed, detailed event information is recorded: trigger time, trigger condition, original links involved, selected alternative path, executed configuration command, command execution result (success / failure), and actual traffic transfer completion time. These records constitute the transfer execution log.

[0141] Analyze the switch execution log generated in step S45 and, combined with real-time performance data from the communication network before and after the path switch, evaluate the adaptability and effectiveness of the path switch strategy. For each successful switch event recorded in the log, analyze whether the actual performance of traffic on the alternative path after the switch (e.g., packet loss rate, latency, and throughput) meets expectations (compared to the evaluation results in the backup resource capability table in step S43). Analyze whether the switch operation adversely affects other unswitched traffic in the network (e.g., causing congestion or increased latency on other links). For failed switches recorded in the log, analyze the cause of the failure (e.g., configuration command failure, problems with the alternative path itself). Identify deficiencies in the strategy, such as improper trigger threshold settings (switching too late or too frequently), suboptimal alternative paths, or flaws in the switch mechanism. Based on the analysis results, propose optimization recommendations for the strategy (S44) or its input data generation process (S41-S43). For example, it may be recommended to increase the predicted failure threshold for a critical link or to consider real-time load when selecting an alternative path. These analysis, evaluation results, and optimization recommendations are summarized in an optimization effectiveness evaluation report.

[0142] The communication resilience index of the energy storage system is calculated based on the optimization effect evaluation report and the overall operating data of the system during the evaluation period. The communication resilience index is a comprehensive indicator that quantifies the anti-interference, self-adaptation and recovery capabilities of the system communication when facing connection degradation or potential failures. The calculation of this index can take into account multiple factors: for example, during the evaluation period, among the predicted high-risk communication interruption events, what proportion of the impact was successfully avoided or mitigated through path transfer (for example, the proportion of critical business interruption time reduction); the success rate of path transfer; the average performance degradation caused by the transfer process; the average time from problem detection to completion of transfer; and the time required for communication to return to normal levels after transfer. Define a weighted calculation formula to combine the above quantitative indicators. For example, the communication resilience index = ×(Total number of successful avoidance / prediction events) + ×(Path transfer success rate)- ×(Average performance degradation)- ×(average transfer time), where This is a weighting factor determined by system importance. Using the specific values ​​extracted from the assessment report, we substituting them into a formula to calculate a quantitative communication resilience index. This value reflects the energy storage system's comprehensive ability to resist and cope with degradation under current communication methods and strategies.

[0143] Preferably, the path transfer strategy scheme formulated in step S44 is specifically:

[0144] Identify high-risk connection points on the topology reliability map based on the traffic distribution characteristic table and obtain a list of risky connection points;

[0145] Perform backup path matching on the risk connection point list and backup resource capacity table to obtain a path matching solution;

[0146] Perform traffic distribution calculation on the path matching plan to obtain a traffic distribution plan;

[0147] Plan the transfer time based on the risk connection point list and traffic distribution plan to obtain a transfer time plan;

[0148] Conduct electrical impact assessment on the path matching solution and obtain an electrical impact assessment report;

[0149] According to the electrical impact assessment report, the transfer time plan and traffic distribution plan are integrated into a hierarchical transfer plan to obtain a path transfer strategy plan.

[0150] In this embodiment of the present invention, a traffic distribution characteristics table is used, which details the traffic type, volume, priority, latency, and bandwidth requirements of each link in the energy storage system's communication topology. A topology reliability map is also used, which provides the predicted failure probability of each link within a specific future time window (e.g., the next seven days). These two factors are combined to identify high-risk connection points. A risk assessment rule is defined: if a link's predicted failure probability exceeds a preset threshold (e.g., greater than 10%), and the link carries high-priority critical service traffic, or its peak traffic consistently exceeds 80% of the link's rated capacity, the connection point is marked as high-risk. Each link in the traffic distribution characteristics table is traversed, its corresponding predicted failure probability and traffic characteristics are queried, and the risk assessment rule is applied. All connection points marked as high-risk are compiled into a list, along with their associated link ID, risk level (e.g., high, medium), and primary risk factor (e.g., high failure probability, critical service, high traffic). This list is referred to as the risky connection point list. For example, a list item may be: [{connection point ID: "CN_101", associated link: "Link_A_B", risk level: "high", reason: "predicted high probability of failure and carries critical control signals"}, {connection point ID: "CN_105", associated link: "Link_C_D", risk level: "high", reason: "peak traffic continues to be overloaded"}].

[0151] Utilizing a list of high-risk connection points and a backup resource capability table, which lists backup paths for potential high-risk links and their capability assessments, for each high-risk connection point in the risk connection point list (and its associated link L), the backup resource capability table is searched for all backup path entries whose "Original High-Risk Link ID" matches L. The resulting backup path sets are filtered and sorted. The filtering criteria are based on whether the backup path's capabilities meet the critical service requirements carried by the high-risk link (derived from the traffic distribution characteristics table in step S42). For example, the backup path's "Alternative Path Available Bandwidth" must be greater than or equal to L's peak traffic, and its "Alternative Path Total Delay" must be less than or equal to the maximum allowable delay for the critical service carried by L. Among the backup paths that meet these criteria, they are sorted in descending order based on "Alternative Path Calculated Reliability," prioritizing the path with the highest reliability. If the reliabilities are equal, the path with the highest available bandwidth is prioritized. One or more optimal backup path sets are determined for each high-risk connection point. The correspondence between the high-risk connection point and the identified backup path set is recorded. For example, the record is: {high-risk connection point ID: [optimal backup path ID_1, suboptimal backup path ID_2, ...]}. This set of corresponding relationships constitutes a path matching solution.

[0152] Utilizing the path matching plan and traffic distribution characteristics table, for each high-risk connection point (and its associated link L) in the path matching plan and its corresponding set of backup paths, a plan is developed to distribute this traffic to the backup paths based on the traffic characteristics carried by L (from the traffic distribution characteristics table). If L carries a single type of critical traffic, all of this traffic is directed to the optimal matched backup path. If L carries multiple traffic types or high volumes of traffic and is matched to multiple backup paths, traffic allocation is calculated based on the available bandwidth and reliability of each backup path. For example, a multi-objective optimization algorithm can be used, with the goal of maximizing the overall reliability of the traffic after diversion while minimizing load imbalance on the backup paths. The inputs are the total amount / type of traffic to be diverted, the available bandwidth and reliability of each backup path, and the output is the traffic proportion or a list of specific traffic types that each backup path should carry. For example, the calculation result is: all critical control signal traffic on Link_A_B is diverted to Path_A_C_B, 60% of the data collection traffic on Link_C_D is diverted to Path_C_E_D, and the remaining 40% is diverted to Path_C_F_D. These specific traffic allocation rules and ratios constitute the traffic allocation plan.

[0153] Utilize a list of risky connection points (including risk levels and causes) and a traffic allocation plan. Based on the risk level and traffic type, determine the priority and trigger timing for each migration action. For connection points marked as "high" risk and carrying critical services, traffic migration on their associated links is prioritized. Dynamic trigger conditions based on real-time monitoring metrics (such as packet loss rate and latency) are set to require immediate migration upon detection of an anomaly. For example, a trigger condition could be set as: the packet loss rate on the associated link exceeds 5% or the latency exceeds 100ms. For connection points with a "medium" risk level or carrying only non-critical services, planned migration can be scheduled during periods of low system load (such as nighttime idle time) or scheduled maintenance windows. For low-risk connection points where the predicted failure probability will not reach the threshold for a significant period of time (for example, more than 30 days), only an early warning can be issued, with no immediate migration planned, or the migration can be scheduled during the next routine maintenance. These migration trigger conditions or scheduled times for different risk levels and traffic types are combined to form a migration schedule.

[0154] Utilize a path matching plan that specifies which links' traffic will be diverted and to which backup paths. Conduct an electrical impact assessment for all path switching operations involved in the plan. Analyze the communication interface modules (e.g., CAN transceivers, Ethernet PHYs), network switch chips, or communication processing units on the control board involved in the switching process. Assess whether modifying routing tables or activating backup interfaces will cause brief power or signal integrity disturbances. For example, by reviewing hardware design documentation or conducting lab tests, evaluate the current consumption changes or crosstalk observed during configuration switching for specific communication module models. Identify any instances of shared power supplies, shared clocks, or tightly coupled signal lines that could cause the switching of one communication path to impact other nearby critical circuits (e.g., cell balancing control signals, temperature sampling signals). Quantify the potential impact, such as estimating the duration of communication interruption (e.g., microseconds or milliseconds), the magnitude of voltage transients, and the level of interference to nearby signals. Summarize these assessment results into an electrical impact assessment report detailing the electrical impact and potential risks of each planned migration operation.

[0155] Leverage the electrical impact assessment report, migration timeline, and traffic allocation plan. Final adjustments and integration of the migration timeline and traffic allocation plan are made to form a tiered path migration strategy. If, based on the electrical impact assessment report, a planned migration action is expected to result in significant electrical impact or prolonged communication disruption, and is categorized as "immediate migration" or "critical service migration," its necessity and timing need to be reassessed. The migration timeline may need to be adjusted to postpone it to a time when the system is less susceptible to disruption. Alternatively, the traffic allocation plan may need to be adjusted to temporarily maintain some of the most critical traffic on the original link (if a complete failure has not occurred) or to implement a more conservative, phased migration approach. All identified migration actions are categorized by priority and trigger method, such as "emergency migration" (triggered in real time, with minimal impact), "planned migration" (predetermined, with some acceptable impact), and "background migration" (for low-priority traffic, with no system impact). For each migration level, detailed trigger conditions (time and event), involved traffic (from the traffic allocation plan), backup paths to be used (from the path matching plan), expected electrical impact (from the assessment report), execution steps, and rollback mechanisms are defined. This structured, hierarchical set of transfer strategies is the path transfer strategy solution.

[0156] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced within the present invention.

[0157] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A communication method for an energy storage system, characterized in that: The following steps are involved: Step S1: Collecting operating data of the energy storage system connection point using a piezoelectric micro-vibration sensor; extracting vibration-communication coupling features from the operating data to obtain a connection point vibration-communication characteristic spectrum, wherein the connection point vibration-communication characteristic spectrum is a graph or model used to describe the vibration-communication coupling characteristics of the connection point under various vibration and operating conditions; Step S2: Based on the vibration-signal characteristic spectrum of the connection point, the signal reflection characteristics of the communication line under vibration conditions are collected, and then non-intrusive time domain reflection analysis is performed to obtain a reflection characteristic change spectrum; based on the reflection characteristic change spectrum, vibration response mapping processing is performed on the communication line to obtain a communication quality dynamic mapping table, wherein the vibration response mapping processing is specifically as follows: Collect key quality parameters of the communication link; construct vibration-communication response functions for the key quality parameters to obtain a vibration response function set; establish a reflection-communication mapping relationship based on the key quality parameters and the reflection characteristic change spectrum to obtain reflection communication mapping data; analyze the charging and discharging current changes of the vibration response function set to obtain current response data; quantify the temperature effect of the vibration response function set to obtain temperature effect data; perform three-dimensional correlation data analysis on the reflection communication mapping data, current response data, and temperature effect data to obtain a communication quality dynamic mapping table; Dynamic impedance calculation and calibration are performed based on the communication quality dynamic mapping table and the reflection characteristic change spectrum to obtain dynamic impedance characteristic data; impedance-vibration sensitivity analysis is performed on the dynamic impedance characteristic data to obtain a connection health matrix; Step S3: performing time series data accumulation and denoising on the connection health matrix to obtain denoised health data; performing micro-change pattern recognition on the denoised health data to obtain a micro-change feature set; The degradation pattern evolution of the micro-change feature set is analyzed to obtain the connection reliability prediction map; Step S4: Perform communication path transfer analysis based on the connection reliability prediction map to obtain a path transfer strategy plan; execute the path transfer strategy plan and calculate the communication resilience index.

2. The energy storage system communication method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: deploying piezoelectric micro-vibration sensors at the module connections of the energy storage system to synchronously collect signal quality parameters of the communication link under different working conditions to form an original vibration-communication data set; Step S12: performing working condition correlation data preprocessing on the original vibration-communication data set to obtain a preprocessed working condition data packet; Step S13: extracting a time domain feature vector from the pre-processed working condition data packet; Step S14: performing frequency domain feature analysis on the pre-processed working condition data packet to obtain a frequency domain feature spectrum; Step S15: constructing vibration-communication coupling features based on the time domain feature vector and the frequency domain feature spectrum to obtain a coupling feature set; Step S16: Generate a connection point vibration-signal characteristic spectrum according to the coupling characteristic set.

3. The energy storage system communication method according to claim 1, characterized in that: The dynamic impedance calculation and calibration in step S2 are specifically as follows: Perform basic impedance conversion on the reflection characteristic change map to obtain the original impedance data set; Perform vibration correlation analysis on the original impedance data set to obtain vibration-impedance relationship data; Perform current influence compensation on the vibration-impedance relationship data to obtain corrected impedance data; performing electrochemical state adjustment on the corrected impedance data to obtain electrochemically adjusted impedance data; Perform static-dynamic impedance calculation on the electrochemically adjusted impedance data to obtain impedance stability data; The electrochemically adjusted impedance data and the impedance stability data are subjected to dynamic impedance characteristic integration to obtain dynamic impedance characteristic data.

4. The energy storage system communication method according to claim 1, characterized in that: The impedance-vibration sensitivity analysis in step S2 is specifically as follows: Perform frequency band sensitivity calculation on the dynamic impedance characteristic data to obtain frequency band sensitivity data; Determine the critical vibration threshold value of the frequency band sensitivity data to obtain vibration critical threshold data; Conduct current correlation analysis on the energy storage system based on the vibration critical threshold data to obtain current sensitivity data; Perform temperature interaction effect analysis on current sensitivity data and frequency band sensitivity data to obtain the temperature impact matrix; Construct three-dimensional response data based on frequency band sensitivity data, current sensitivity data and temperature impact matrix; Performing vibration tolerance assessment on the three-dimensional response data and the vibration critical threshold data to obtain vibration tolerance data; Generate an impedance sensitivity feature set based on frequency band sensitivity data and vibration tolerance data; A connection health matrix is ​​constructed based on the impedance sensitivity feature set and dynamic impedance characteristic data.

5. The energy storage system communication method according to claim 1, characterized in that: The micro-change pattern recognition in step S3 is specifically as follows: Determine the sliding window of denoised health data and perform trend feature statistics within the window to obtain trend analysis results; Perform fluctuation characteristic evolution processing on trend analysis results to obtain fluctuation evolution data; Perform working condition difference response analysis on the fluctuation evolution data to obtain working condition sensitivity data; Conduct temperature sensitivity tracking on the working condition sensitivity data to obtain temperature sensitivity data; Accurately identify mutation points based on trend analysis results and fluctuation evolution data to obtain mutation point feature data; Performing cyclic correlation analysis on temperature sensitivity data and mutation point characteristic data to obtain cyclic correlation data; A micro-change feature set is generated based on the cyclic correlation data and mutation point feature data.

6. The energy storage system communication method according to claim 1, characterized in that: The degradation mode evolution analysis in step S3 is specifically as follows: Perform degradation pattern classification on the micro-change feature set to obtain degradation pattern recognition results; According to the degradation pattern recognition results, the degradation trajectory of the denoised health data is deduced to obtain the degradation trajectory prediction data; Calculate the failure probability of the degradation trajectory prediction data to obtain the failure probability time series data; Generate a connection reliability prediction map based on the failure probability time series data and degradation trajectory prediction data.

7. The energy storage system communication method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing communication topology reliability assessment based on the connection reliability prediction map to obtain a topology reliability map; Step S42: Analyze the communication traffic characteristics based on the topology reliability map to obtain a traffic distribution characteristics table; Step S43: Evaluate the backup path resources of the energy storage system based on the traffic distribution characteristic table and the topology reliability map to obtain a backup resource capacity table; Step S44: formulating a path transfer strategy based on the topology reliability map, the traffic distribution characteristics table, and the backup resource capability table; Step S45: Perform a switching execution mechanism on the path transfer strategy to obtain a switching execution log; Step S46: Perform adaptability verification and optimization on the switch execution log to obtain an optimization effect evaluation report; Step S47: Calculate the communication resilience index of the energy storage system according to the optimization effect evaluation report.

8. The energy storage system communication method according to claim 7, characterized in that: The path transfer strategy scheme formulated in step S44 is specifically as follows: Identify high-risk connection points on the topology reliability map based on the traffic distribution characteristic table and obtain a list of risky connection points; Perform backup path matching on the risk connection point list and backup resource capacity table to obtain a path matching solution; Perform traffic distribution calculation on the path matching plan to obtain a traffic distribution plan; Plan the transfer time based on the risk connection point list and traffic distribution plan to obtain a transfer time plan; Conduct electrical impact assessment on the path matching solution and obtain an electrical impact assessment report; According to the electrical impact assessment report, the transfer time plan and traffic distribution plan are integrated into a hierarchical transfer plan to obtain a path transfer strategy plan.

9. A communication system for an energy storage system, characterized in that: For executing the energy storage system communication method according to claim 1, the communication system of the energy storage system comprises: The correlation feature extraction module is used to collect the operating condition data of the energy storage system connection point through a piezoelectric micro-vibration sensor; extract the vibration-communication coupling features of the operating condition data to obtain the vibration-communication characteristic spectrum of the connection point; The dynamic impedance mapping module is used to collect the signal reflection characteristics of the communication line under vibration conditions based on the vibration-signal characteristic spectrum of the connection point, and then perform non-invasive time-domain reflection analysis to obtain a reflection characteristic change spectrum; perform vibration response mapping processing on the communication line based on the reflection characteristic change spectrum to obtain a communication quality dynamic mapping table; perform dynamic impedance calculation and calibration based on the communication quality dynamic mapping table and the reflection characteristic change spectrum to obtain dynamic impedance characteristic data; and perform impedance-vibration sensitivity analysis on the dynamic impedance characteristic data to obtain a connection health matrix; The reliability prediction module is used to accumulate and denoise the time series data of the connection health matrix to obtain denoised health data; perform micro-change pattern recognition on the denoised health data to obtain a micro-change feature set; and perform degradation pattern evolution analysis on the micro-change feature set to obtain a connection reliability prediction map. The communication path optimization module is used to perform communication path transfer analysis based on the connection reliability prediction map to obtain a path transfer strategy plan; execute the path transfer strategy plan and calculate the communication resilience index.

Images