Storage battery abnormity early warning system based on intelligent battery gateway

Abstract

The invention discloses a storage battery abnormity early warning system based on an intelligent battery gateway, and relates to the technical field of battery monitoring. The system comprises a sensor module used for collecting voltage, internal resistance, temperature, thermal runaway state and floating charge voltage parameters of a single battery in real time; the intelligent battery gateway comprises a data processing module and an early warning module, and the data processing module carries out fusion analysis on sensor data through an intelligent algorithm, calculates the battery remaining capacity SOC and the state of health SOH and executes an intelligent equalization strategy; and the early warning module generates a multi-stage early warning signal according to the parameter threshold and the anomaly detection model. According to the invention, through a multi-dimensional fusion sensing technology and a dual-drive intelligent algorithm, the aging sign of the internal microstructure of the battery is accurately captured, and high-precision estimation of the residual electric quantity and the health state of the storage battery is realized; and in the aspect of early warning performance, a multi-dimensional linkage detection model and an intelligent early warning mechanism are established, so that the accuracy and timeliness of fault early warning are effectively improved, and the early warning time is greatly advanced.

Description

Technical Field

[0001] The present invention relates to the technical field of battery monitoring, and specifically to a storage battery abnormal warning system based on an intelligent battery gateway. Background Art

[0002] In the full life cycle management of storage batteries, traditional monitoring solutions face core problems: due to the adoption of a single-parameter monitoring and fixed-threshold alarm mechanism, it is difficult to accurately capture the internal microstructure aging of the battery, resulting in large estimation deviations of the state of charge (SOC) and state of health (SOH), and it is impossible to achieve effective early warning of faults. The specific manifestations are as follows: relying only on surface parameters such as voltage and temperature, it is impossible to sense internal aging such as SEI film growth and plate vulcanization; the early warning method based on a single-point threshold (such as alarm when the temperature > 60°C) cannot identify the abnormal parameter coupling before thermal runaway, and the early warning lead is insufficient; traditional algorithms are difficult to adapt to working condition changes such as high and low temperatures and different charge and discharge rates, and the estimation accuracy decays significantly during long-term operation, resulting in maintenance lags and resource waste. Summary of the Invention

[0003] The purpose of the present invention is to provide a storage battery abnormal warning system based on an intelligent battery gateway to solve the problems raised in the above background art.

[0004] To achieve the above purpose, the present invention provides the following technical solution: A storage battery abnormal warning system based on an intelligent battery gateway, including the following steps: A sensor module for real-time collection of parameters such as the voltage, internal resistance, temperature, thermal runaway state, and floating charge voltage of individual batteries; An intelligent battery gateway, including a data processing module and an early warning module. The data processing module performs fusion analysis on sensor data through an intelligent algorithm, calculates the state of charge SOC and state of health SOH of the battery, and executes an intelligent balancing strategy; the early warning module generates multi-level early warning signals according to parameter thresholds and an abnormal detection model; A communication module for local data interaction and remote platform data upload, supporting Bluetooth, serial port, and network communication; A human-computer interaction module for visual early warning of abnormal states and operation and maintenance management.

[0005] Preferably, the sensor module includes: A composite parameter measurement unit for synchronously collecting the voltage response curve and AC internal resistance of individual batteries by using the pulsed current injection method to obtain electrochemical impedance spectroscopy (EIS) data including charge transfer resistance , double-layer capacitance ; A thermal runaway detection unit for identifying the precursor of thermal runaway through a three-dimensional linkage criterion of temperature gradient, internal resistance mutation rate, and second derivative of voltage. The criterion formula is: ; In the formula, is the temperature change rate per unit time, is the change rate of the internal resistance relative to the initial value, is the curvature of the voltage curve.

[0006] Preferably, the sensor module further includes a self-calibration unit, and the self-calibration unit includes: A MEMS temperature and pressure compensation module for real-time monitoring of the ambient air pressure and humidity, and dynamically correcting the thermal resistance coefficient of the temperature sensor by means of a look-up table method, with the corrected temperature measurement accuracy of ±0.5 °C; A standard voltage source module for automatically calibrating the voltage acquisition channel every 24 hours, with the calibrated voltage drift error < 0.05%.

[0007] Preferably, the SOC calculation algorithm of the data processing module is an improved cubature Kalman filter algorithm, and the calculation of the remaining battery capacity SOC includes: Set a third-order RC equivalent circuit model to real-time identify the ohmic internal resistance , charge transfer resistance , double-layer capacitance and other model parameters; Construct a multi-input measurement equation, and fuse the voltage , internal resistance, temperature T to construct a state space equation: ; Among them, is the open-circuit voltage, is the ohmic voltage drop, is the current flowing through the battery, is the polarization voltage, is the change rate of the polarization voltage with time, is the change in the polarization voltage caused by the charging of the double-layer capacitance by the current, is the polarization voltage relaxation time constant; Set an anti-outlier adaptive mechanism, judge the measurement residual through the Mahalanobis distance, and switch to the square root cubature Kalman filter when it exceeds the limit, with the calculation accuracy of SOC of ±3%.

[0008] Preferably, the SOH prediction algorithm of the data processing module is a gated recurrent unit algorithm based on the attention mechanism, including: Construct a multi-feature input layer, extract 12-dimensional aging features such as the internal resistance growth rate and the capacity attenuation rate, and construct a time window sliding matrix; Construct an attention mechanism layer, calculate the contribution degree of historical data to the current SOH through weighted calculation, focus on the features in the accelerated aging stage, and the weight formula is: ; Among them, is the moment attention weight, is the hidden layer state, is the hidden layer state at the previous moment, is a scoring function used to measure the correlation degree between the hidden layer state at the previous moment and the hidden layer state at the current moment; Construct a physical constraint correction layer to correct the prediction result through the capacity-internal resistance coupling equation: ; Among them, represents the change ratio of the measured internal resistance relative to the initial internal resistance and the failure internal resistance. The coefficient 0.8 indicates that the internal resistance growth dominates the weight in SOH evaluation, is the capacity attenuation ratio. The coefficient 0.2 shows that the influence of capacity attenuation is relatively small, and the SOH prediction accuracy after correction is ±5%.

[0009] Preferably, the intelligent equalization strategy is dynamic grouping equalization based on impedance spectrum clustering, including: Construct a health state stratification, cluster the main components of the impedance spectrum through the K-means algorithm, and divide the batteries into a healthy group, a declining group, and a fault warning group; Set a time-varying threshold trigger mechanism, and set different equalization thresholds according to the charge and discharge conditions: ； During the equalization process, monitor the change rate of the internal resistance. When it automatically terminates.

[0010] Preferably, the warning module includes: A five-dimensional feature anomaly detection unit constructs a feature space including voltage, internal resistance, temperature, SOC, and SOH, and uses the isolation forest algorithm to detect the isolation degree of data points; A Bayesian network inference engine establishes a causal relationship knowledge base containing 56 fault modes, derives the root cause of the fault through posterior probability calculation, and outputs maintenance suggestions.

[0011] Preferably, the intelligent battery gateway adopts an edge computing architecture, and the intelligent battery gateway further includes: A data hierarchical processing module is used to upload abnormal data segments and 15-dimensional feature vectors, and process regular data locally; A federated learning update module with a lightweight AM-GRU model built-in is used to synchronize model parameter gradients through federated learning.

[0012] Preferably, the system further includes: A digital twin mapping module that builds a virtual mirror of the battery pack inside the intelligent battery gateway, synchronizes the parameters of individual batteries in real time, simulates the evolution law of fault scenario parameters through finite element simulation, predicts potential faults, and generates personalized maintenance work orders.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: Through multi-dimensional fusion perception technology and dual-drive intelligent algorithms, it accurately captures the signs of aging of the internal microstructure of the battery, realizes high-precision estimation of the remaining battery power and health status; in terms of early warning performance, a multi-dimensional linkage detection model and an intelligent early warning mechanism are established, effectively improving the accuracy and timeliness of fault early warning, and significantly advancing the early warning time; at the management and operation and maintenance level, the dynamic grouping intelligent balancing strategy and the fault inference engine not only improve the management efficiency of the battery pack, but also can achieve root cause diagnosis of faults, provide maintenance suggestions, and at the same time, the edge computing architecture optimizes data processing and protects user privacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic structural diagram of a battery abnormal early warning system based on an intelligent battery gateway provided by an embodiment of the present invention; Figure 2 It is an implementation step diagram of a battery abnormal early warning system based on an intelligent battery gateway provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0016] Please refer to Figure 1 , the present invention provides a battery abnormal early warning system based on an intelligent battery gateway, including: A sensor module 11 for real-time collection of parameters such as the voltage, internal resistance, temperature, thermal runaway state, and floating charge voltage of individual batteries; An intelligent battery gateway 12, including a data processing module and an early warning module. The data processing module performs fusion analysis on sensor data through intelligent algorithms, calculates the state of charge (SOC) and state of health (SOH) of the battery, and executes an intelligent balancing strategy; the early warning module generates multi-level early warning signals according to parameter thresholds and an abnormal detection model; A communication module 13 for local data interaction and remote platform data upload, supporting Bluetooth, serial port, and network communication; The human-machine interaction module 14 is used for visual warning of abnormal states and operation and maintenance management.

[0017] In an optional embodiment, the sensor module 11 further includes: A composite parameter measurement unit, which is used to synchronously collect the voltage response curve and the AC internal resistance of a single battery by using the pulsed current injection method, and obtain the electrochemical impedance spectrum (EIS) data including the charge transfer resistance , the double-layer capacitance .

[0018] Specifically, by injecting a pulsed current with a pulse width of 10 ms and an amplitude of 500 mA into the single battery, during this transient process, the voltage response curve of the battery is synchronously collected at a high sampling rate of 10 kHz. At the same time, using a sweep frequency technique of 10 mHz - 1 kHz, the AC internal resistance is accurately measured. In this way, the electrochemical impedance spectrum (EIS) data containing battery microstructure characteristics such as the charge transfer resistance ( ), the double-layer capacitance ( ) can be obtained. The traditional DC internal resistance measurement method has certain limitations and is difficult to capture the subtle changes inside the battery. However, the present invention can accurately identify internal resistance changes as low as 0.1 mΩ level through high sensitivity, so that it can effectively capture in advance when the signs of internal aging of the battery just emerge, providing key data support for subsequent maintenance and abnormal warning.

[0019] A thermal runaway detection unit, which is used to identify the precursor of thermal runaway through a three-dimensional linkage criterion of temperature gradient ( , unit: °C / min), internal resistance mutation rate ( , change rate relative to the initial value), and second derivative of voltage ( , reflecting the curve curvature). The criterion formula is: ; In the formula, is the temperature change rate per unit time, is the internal resistance change rate relative to the initial value, is the voltage curve curvature.

[0020] In the past, the single temperature threshold detection method often had problems of untimely warning or high false alarm rate. In contrast, the three-dimensional linkage criterion in this embodiment has achieved a significant improvement in the warning time. The warning lead time can reach 30 minutes, and at the same time, the false alarm rate is reduced by 60%, greatly improving the accuracy and reliability of the thermal runaway warning, laying a solid foundation for ensuring the safe and stable operation of the battery system.

[0021] In an optional embodiment, the sensor module 11 further includes: The MEMS temperature and pressure compensation module is used to monitor the ambient air pressure and humidity in real time, and dynamically correct the thermal resistance coefficient of the temperature sensor by means of a look-up table method. After correction, the temperature measurement accuracy is ±0.5°C; The standard voltage source module is used to automatically calibrate the voltage acquisition channel every 24 hours. After calibration, the voltage drift error is <0.05%.

[0022] Optionally, this embodiment provides a self-calibrating sensor array, which integrates a MEMS temperature and pressure compensation unit and can monitor the ambient air pressure in real time and with high precision, with an accuracy of up to ±0.1 kPa, and the relative humidity, with an accuracy of ±2%RH. By using the look-up table method and according to the ambient parameters monitored in real time, the thermal resistance coefficient of the temperature sensor is dynamically corrected. After correction, the temperature measurement accuracy can be stably controlled within ±0.5°C, effectively eliminating the interference of environmental factors on temperature measurement. In addition, the sensor array is built-in with a 16-bit ADC and a standard voltage source. The accuracy of the standard voltage source is as high as ±0.02%FS. Every 24 hours, the system will automatically start the calibration program to calibrate the voltage acquisition channel, which can effectively suppress the drift error generated during the long-term operation of the sensor, ensure that it always operates with high precision, and control the drift error within <0.05%, providing a reliable guarantee for the accurate acquisition of the battery state parameters.

[0023] In an optional embodiment, the SOC calculation algorithm of the data processing module is an improved cubature Kalman filter algorithm. The calculation of the state of charge (SOC) of the battery includes: Setting a third-order RC equivalent circuit model to identify the ohmic internal resistance in real time 、the charge transfer resistance 、the double-layer capacitance and other model parameters; Constructing a multi-input measurement equation to fuse the voltage 、the internal resistance, and the temperature T to construct a state space equation: ; Among them, is the open-circuit voltage, is the ohmic voltage drop, is the current flowing through the battery, is the polarization voltage, is the rate of change of the polarization voltage with time, is the change in the polarization voltage caused by the charging of the double-layer capacitance by the current, is the polarization voltage relaxation time constant; Setting an anti-outlier adaptive mechanism to judge the measurement residual by the Mahalanobis distance. When it exceeds the limit, switch to the square-root cubature Kalman filter, and the calculation accuracy of SOC is ±3%.

[0024] Understandably, in order to achieve a more accurate dynamic estimation of the battery's state of charge (SOC), we adopted an improved cubature Kalman filter (CV-KF) algorithm, whose core basis is to construct a third-order RC equivalent circuit model. In this model, a nonlinear polarization capacitance is introduced to comprehensively reflect the complex electrochemical characteristics inside the battery. During actual operation, the model parameters (such as the ohmic internal resistance , charge transfer resistance , and double-layer capacitance ) are not fixed, but through real-time identification, the model can keenly capture the dynamic changes in polarization characteristics in different SOC intervals. For example, at the initial stage of battery charging, the polarization characteristics are relatively weak. As the charging process progresses, in the medium and high SOC intervals, the polarization effect gradually strengthens. The real-time adjustment of the model parameters enables the model to always fit the true state of the battery.

[0025] Among them, the open-circuit voltage is closely related to the SOC of the battery. In order to accurately obtain the relationship between the two, a large amount of data from 300°C high-temperature accelerated aging tests was obtained. Through in-depth analysis and processing of the test data, the open-circuit voltage was finally fitted to a fifth-order polynomial, and the fitting degree . This high-precision fitting result provides a solid data basis for subsequent SOC estimation. In a complex and changing actual application environment, the battery will inevitably be affected by various interference factors, such as pulsed current interference, which may cause deviations or even divergence in the SOC estimation results. To effectively address this problem, this embodiment is equipped with an anti-interference adaptive mechanism. This mechanism dynamically evaluates the reliability of the estimation results by calculating the Mahalanobis distance of the measurement residuals in real time. Once the measurement residuals exceed 3σ, the system will automatically trigger a switching mechanism, quickly switching from the original cubature Kalman filter to the square root cubature Kalman filter (SCKF). This adaptive switching can greatly suppress the adverse effects of pulsed current interference on the estimation results and significantly improve the SOC estimation accuracy.

[0026] In an optional embodiment, the SOH prediction algorithm of the data processing module is a gated recurrent unit algorithm based on the attention mechanism, including constructing a multi-feature input layer, constructing an attention mechanism layer, and constructing a physical constraint correction layer.

[0027] Construct a multi-feature input layer, extract 12-dimensional aging features such as the internal resistance growth rate and capacity decay rate, and construct a time window sliding matrix; Among them, in the data collection stage, the key features during the battery aging process are deeply explored. By accurately extracting a total of 12 aging features such as the internal resistance growth rate, capacity attenuation rate, and self-discharge rate, the aging status of the battery is comprehensively reflected. To effectively utilize these time-series data, a time window sliding matrix is constructed. The setting of the window length is based on the designed battery life and is 1 / 100 of the designed battery life. For example, for a battery with a designed life of 10 years, the time window is selected as 36 days. This sliding window mechanism can dynamically capture the changes in the aging characteristics of the battery at different stages and provide a rich data basis for subsequent analysis.

[0028] Construct an attention mechanism layer to focus on the characteristics of the accelerated aging stage by calculating the contribution degree of historical data to the current SOH through weighted calculation. In traditional time-series analysis, all historical data are often treated equally, ignoring the differences in the influence degree of data at different time periods on the current state. In this embodiment, the attention mechanism is used for optimization. By calculating the key time periods that have the greatest impact on the current SOH in historical data through weighted calculation, effective screening and focusing of data are achieved. Factors such as the number of deep discharges and the duration of high-temperature operation, which have a significant impact on battery aging, can be focused on under this mechanism. The weight formula is: ; where, is t the attention weight at time is the hidden layer state, is the hidden layer state at the previous moment, is a scoring function used to measure the correlation degree between the hidden layer state at the previous moment and the hidden layer state at the current moment, normalizes the scores of all moments so that the sum of all weights is 1; in actual operation, this mechanism can keenly focus on the feature contributions in the battery accelerated aging stage, avoiding the problem of gradient disappearance that is prone to occur in traditional recurrent neural networks when processing long sequence data, so that the model can more accurately capture the key factors affecting SOH.

[0029] Construct a physical constraint correction layer. Considering that battery aging is not only a data-driven process but also follows certain electrochemical aging mechanisms, in this embodiment, a capacity-internal resistance coupling constraint equation is used to combine the physical model with the data-driven model and correct the prediction results through the capacity-internal resistance coupling equation: ; where, represents the change ratio of the measured internal resistance relative to the initial internal resistance and the failure internal resistance. The coefficient 0.8 indicates that the internal resistance growth dominates the weight in SOH evaluation, is the capacity attenuation ratio, and the coefficient 0.2 indicates that the impact of capacity attenuation is relatively small. The corrected SOH prediction accuracy is ±5%. In this equation, by comparing the real-time monitoring data of the battery internal resistance ( ) and capacity ( ) with the initial values and failure values, the impacts of capacity attenuation and internal resistance growth on SOH are comprehensively considered. This fusion method integrates the electrochemical aging mechanism into the data-driven results, significantly improving the accuracy of SOH prediction. After actual test verification, after adopting this solution, the SOH prediction accuracy can be improved to ±5%. Compared with the ±8% of the traditional solution, there is a qualitative leap in prediction accuracy, providing a more reliable basis for the abnormal warning of the storage battery.

[0030] In an optional embodiment, the intelligent balancing strategy is dynamic grouping balancing based on impedance spectrum clustering, including constructing a health state hierarchy and setting a time-varying threshold triggering mechanism.

[0031] Construct a health state hierarchy. By clustering the main components of the impedance spectrum through the K-means algorithm, the batteries are divided into a healthy group, a declining group, and a fault warning group.

[0032] Specifically, to achieve more precise management of the storage battery, in this embodiment, the K-means algorithm is used to perform clustering analysis on the main components (PC1-PC3) of the impedance spectrum of single cells. Through a large amount of experimental data and in-depth analysis of the battery characteristics, the batteries are clearly divided into different health levels. When PC1 < 0.3, the battery is classified into the "healthy group", which means its performance is good and no excessive balancing intervention is required; when 0.3 ≤ PC1 < 0.7, the battery enters the "declining group", and its performance has declined to a certain extent, and appropriate attention and adjustment are needed; when PC1 ≥ 0.7, the battery is classified into the "fault warning group", indicating that a fault may occur soon and urgent key treatment is required. This hierarchical method can effectively avoid unnecessary and ineffective balancing operations on healthy batteries, greatly improving the utilization efficiency of balancing resources.

[0033] Set a time-varying threshold triggering mechanism. Considering that there are significant differences in the working conditions of the battery during charge and discharge, this embodiment gives a flexible time-varying threshold triggering mechanism. Specifically, the balancing threshold will be dynamically adjusted according to the charge and discharge working conditions as follows: ； During the balancing process, monitor the change rate of the internal resistance. When , it will automatically terminate.

[0034] Specifically, in the above formula, represents the change amount of the State of Charge of the battery, It represents the change in the State of Health of the battery. During charging, when exceeds 8% and exceeds 10%, the system determines that the balancing condition is triggered; during discharging, the triggering condition is more stringent, it needs to be greater than 12% and greater than 15%. In addition, during the balancing process, the system will monitor the change rate of the battery internal resistance in real time. When occurs, the system will automatically terminate the balancing operation. Through actual test verification, this strategy has significantly increased the balancing efficiency by 40% and reduced the energy consumption by 30% at the same time, effectively saving energy while improving the battery performance.

[0035] In an optional embodiment, the warning module includes: A five-dimensional feature anomaly detection unit that constructs a feature space including voltage, internal resistance, temperature, SOC, and SOH, and uses the Isolation Forest algorithm to detect the isolation degree of data points.

[0036] Specifically, during the operation of the storage battery, in order to accurately capture the early fault signs, this system innovatively constructs a five-dimensional feature vector space. This vector space covers the parameters of five key dimensions: voltage (V), internal resistance (R), temperature (T), state of charge (SOC), and state of health (SOH). These five parameters are like the "barometer" of the storage battery's health status, and their changes can intuitively reflect the physical and chemical changes inside the storage battery.

[0037] To effectively mine the information contained in this five-dimensional feature vector, this embodiment uses the Isolation Forest algorithm. The core principle of this algorithm is to calculate the "isolation degree" of data points. Simply put, for a data point, if it is far from other data points in the data space, that is, in a relatively isolated position, then its "isolation degree" is high. In the storage battery fault detection scenario, early faults often cause some data points to deviate from the data distribution under normal operating conditions, and these abnormal points can be sensitively identified through the Isolation Forest algorithm.

[0038] For example, when the self-discharge is abnormal ( / 24h > 5%), the change in the SOC parameter will form a point deviating from the normal trajectory in the five-dimensional feature space, and the Isolation Forest algorithm can quickly capture this change. Another example is the plate sulfation fault ( ), and the abnormal changes in the internal resistance R and the state of health SOH will also be accurately identified by the algorithm. Compared with the traditional method that simply relies on threshold judgment, this unsupervised detection method based on the five-dimensional feature space and the Isolation Forest algorithm has successfully increased the warning accuracy rate from the previous low level to 98% significantly, providing strong technical support for timely detecting the early faults of the storage battery.

[0039] The Bayesian network inference engine establishes a knowledge base of causal relationships containing 56 failure modes, derives the root cause of the failure through posterior probability calculation, and outputs maintenance suggestions.

[0040] Specifically, taking the positive plate corrosion failure as an example, when the system detects that the three conditions of "float charge voltage > 2.45V + internal resistance growth rate > 20% / month + temperature > 50°C" are simultaneously met, it is very likely to trigger the positive plate corrosion failure. An overly high float charge voltage will accelerate the oxidation reaction of the positive plate. An overly fast internal resistance growth rate indicates that the internal structure of the plate may have changed, and an overly high temperature will further accelerate the rate of the chemical reaction. The interaction of the three greatly increases the risk of positive plate corrosion.

[0041] Looking at the electrolyte dry-out failure again, its triggering conditions are "voltage fluctuation > 0.1V + internal resistance mutation rate > 30% + ambient humidity < 20%RH". Abnormal voltage fluctuation reflects the instability of the electrochemical reaction inside the battery. The internal resistance mutation may be due to the reduction of the electrolyte, which leads to hindered ion conduction, and an overly low ambient humidity will accelerate the evaporation of the electrolyte. The combined effect of these factors indicates the possibility of electrolyte dry-out failure. When the system detects an abnormal situation, it will quickly call the Bayesian network fault inference engine. Based on the Bayesian principle, through complex and accurate posterior probability calculation, this engine can derive the most likely root cause of the current abnormality from numerous possible causes, and the confidence level of the derivation result is over 90%. Once the cause of the failure is determined, the system will immediately output detailed and targeted maintenance suggestions, such as "It is recommended to detect the electrolyte specific gravity of battery No. 15". This leap from simple "threshold alarm" to in-depth "root cause diagnosis" greatly improves the efficiency and accuracy of fault handling, providing a reliable guarantee for ensuring the stable operation of the battery system.

[0042] In an optional embodiment, the intelligent battery gateway 12 adopts an edge computing architecture. The intelligent battery gateway 12 further includes: A data hierarchical processing module, which is used to upload abnormal data segments and 15-dimensional feature vectors, and process regular data locally; A federated learning update module, which has a lightweight AM-GRU model built in and is used to synchronize model parameter gradients through federated learning.

[0043] Optionally, at the local edge layer, up to 95% of the regular data is processed in real time. Through precise data screening algorithms, data in the normal operating state is quickly identified and the processing process is completed locally. Only when an abnormal situation is detected, specific data, that is, abnormal data segments, are uploaded to the cloud. This segment contains the detailed waveforms 10 minutes before and after the occurrence of the abnormality, so as to deeply analyze the development trend and change characteristics of the abnormality in the follow-up. At the same time, 15-dimensional feature vectors obtained through calculation and refinement are also uploaded. These feature vectors condense the key information of the battery operating state. After this optimization, the data transmission volume is reduced by 70% compared with the traditional mode, greatly alleviating the network transmission pressure. In addition, the local edge layer is set with a function of resuming data transmission after network interruption. Through a built-in 16MB cache, this cache can store abnormal data for up to 7 days, ensuring that the unuploaded abnormal data can be transmitted to the cloud completely and correctly after the network is restored, guaranteeing the continuity and integrity of the data.

[0044] In addition, a lightweight AM-GRU model is built into each gateway. This model is specifically optimized for battery data processing and has efficient and accurate feature extraction and state prediction capabilities. Through federated learning technology, the model parameter gradients can be periodically synchronized between gateways, rather than the original data. This method plays a key role in protecting user privacy and avoids the risk of privacy leakage caused by the aggregation of original data from different sites. At the same time, the cross-site synchronization of model parameter gradients enables the model to comprehensively learn the battery aging characteristics of different sites and achieve global optimization of the battery aging characteristics. Compared with the traditional centralized model, it effectively solves the overfitting problem caused by data concentration in a single center, and significantly improves the generalization ability and prediction accuracy of the model in complex and changing actual scenarios.

[0045] In an optional embodiment, the system further includes: A digital twin mapping module. The digital twin mapping module builds a virtual mirror of the battery pack inside the intelligent battery gateway 12, synchronizes the parameters of individual batteries in real time, simulates the evolution law of fault scenario parameters through finite element simulation, predicts potential faults and generates personalized maintenance work orders.

[0046] Specifically, an advanced digital twin model of the battery pack is deployed inside the gateway to build a precise virtual image for each single battery, realizing real-time synchronous mapping of battery electrical parameters and thermal parameters. This model efficiently converts various collected data into an intuitive presentation in the virtual space, enabling maintenance personnel to clearly understand the operating status of the battery. At the same time, with the help of finite element simulation technology, it deeply simulates the evolution laws of battery parameters under common fault scenarios such as pole post loosening and separator perforation. After data training and verification, the system can keenly capture potential fault signs 72 hours in advance and generate highly personalized maintenance work orders according to the prediction results. For example, when the system monitors that the 8th single battery in the 3rd battery pack is about to have an abnormality, it will accurately generate a work order, clearly stating "It is recommended to replace the 8th single battery in the 3rd battery pack, and the expected failure time is July 15th", providing a strong basis for maintenance work and greatly improving the pertinence and timeliness of maintenance work.

[0047] In this embodiment, through multi-dimensional fusion perception technology and dual-drive intelligent algorithms, it accurately captures the signs of internal microstructure aging of the battery, realizing high-precision estimation of the remaining battery power and health status; in terms of early warning performance, a multi-dimensional linkage detection model and an intelligent early warning mechanism are established to effectively improve the accuracy and timeliness of fault early warning and significantly advance the early warning time; at the management and maintenance level, dynamic grouping intelligent balancing strategies and fault inference engines not only improve the management efficiency of the battery pack but also enable root cause diagnosis of faults, provide maintenance suggestions, while the edge computing architecture optimizes data processing and protects user privacy.

[0048] Based on the above embodiments, as Figure 2 shown, the present invention also provides an exemplary implementation step of a storage battery abnormal early warning system based on an intelligent battery gateway, and the exemplary implementation step includes: Step 100, hardware deployment and connection: Install sensors, deploy an intelligent gateway, and connect a Bluetooth / network communication module to ensure the stable operation of the hardware.

[0049] Specifically, step 100 includes: Step 110, sensor installation: Integrate a composite parameter measurement unit composed of a voltage probe, an NTC temperature sensor, an impedance excitation coil, etc. through a flexible printed circuit board (FPCB), and closely attach it to the surface and electrode connection of the single battery, ensuring that the thickness does not exceed 0.5 mm to adapt to the installation in the battery gap; deploy a thermal runaway detection unit at key positions such as the heat dissipation channel of the battery pack, and after completion, connect the sensor module 11 to the intelligent battery gateway 12 through a cable; Step 120, Gateway and Communication Device Configuration: Connect the intelligent battery gateway 12 to the field network, configure Bluetooth, serial port or network communication parameters to enable it to support data interaction with the local mobile APP and the remote monitoring platform; Connect the local alarm indicator light to ensure the normal operation of the human-machine interaction module 14.

[0050] Step 200, Software Configuration and Algorithm Initialization: Load the SOC / SOH algorithm, set the warning threshold and the intelligent balancing strategy, and complete the parameter initialization.

[0051] Specifically, Step 200 includes: Step 210, Algorithm Loading for the Data Processing Module: In the data processing module of the intelligent battery gateway 12, load the improved cubature Kalman filter (CV-KF) algorithm for SOC estimation and the attention mechanism-based gated recurrent unit (AM-GRU) algorithm for SOH prediction; According to the type and specifications of the battery, initialize the relevant parameters in the algorithm, such as the initial parameters of the third-order RC equivalent circuit model, the dimension of the feature vector of the AM-GRU algorithm, etc.; Conduct preliminary debugging and optimization of the algorithm through the battery data collected on-site to ensure that the algorithm can accurately calculate SOC and SOH; Step 220, Parameter Setting for the Warning Module: In the warning module, set the safety thresholds for each parameter, such as the upper and lower limits of the single-cell voltage (e.g., <2.0V or >2.4V), the internal resistance increase threshold (e.g., >30%), the temperature threshold (e.g., >60°C), etc.; Configure the trigger conditions and methods for multi-level warnings, determine the color and blinking frequency of the local alarm indicator light, the content and method of push notifications of the mobile APP, as well as the alarm prompt and processing process of the remote monitoring platform; Conduct simulation tests on the warning module to verify the accuracy and timeliness of warnings in different abnormal situations; Step 230, Intelligent Balancing Strategy Configuration: Based on the K-means algorithm, cluster the principal components (PC1-PC3) of the impedance spectrum of the battery pack to divide the batteries into a healthy group, a degraded group, and a fault warning group; According to the charge and discharge conditions, set different intelligent balancing thresholds, such as trigger balancing under the charging state, and trigger balancing under the discharging state; During the balancing process, continuously monitor the internal resistance change rate, and automatically terminate the balancing operation when ; After the configuration is completed, conduct actual balancing tests to observe the balancing effect.

[0052] Step 300, System Debugging and Performance Optimization: Test the data collection, state estimation, warning, and balancing functions, and optimize the algorithm and communication configuration.

[0053] Specifically, Step 300 includes: Step 310, Function Test: Start the system, check whether the sensor module 11 can collect various parameters of the single battery in real time and accurately, and transmit the data to the intelligent battery gateway 12; verify the calculation results of the data processing module, compare the actual measurement data with the SOC and SOH values estimated by the algorithm, and evaluate the calculation accuracy; test the response of the early warning module under different abnormal conditions to ensure that multi-level early warning signals can be sent out in a timely and accurate manner; check the execution of the intelligent equalization strategy and observe the parameter changes of the battery pack during the equalization process; Step 320, Performance Optimization: According to the results of the function test, optimize the performance of the system. If the estimation accuracy of SOC or SOH does not meet the standard, adjust the algorithm parameters or improve the algorithm model; if data loss or delay occurs during communication, optimize the communication protocol and network configuration; for the intelligent equalization strategy, further optimize the clustering algorithm and threshold setting to improve the equalization efficiency and the consistency of the battery pack; conduct long-term operation tests on the system, collect actual operation data, and continuously optimize the system performance.

[0054] Step 400, System Application and Operation and Maintenance Management: Connect to the actual scenario for real-time monitoring, generate maintenance work orders based on digital twins, and perform regular maintenance and upgrades.

[0055] Specifically, Step 400 includes: Step 410, System Application: Officially apply the debugged and optimized system to the monitoring and management scenarios of the battery pack, such as communication base stations, energy storage power stations, etc.; view the operating status, SOC, SOH and other information of the battery pack in real time through the mobile APP and the remote monitoring platform, and receive abnormal early warning notifications; use the digital twin mapping system to build a virtual mirror of the battery pack inside the gateway, synchronize the parameters of the single battery in real time, simulate the fault scenario through finite element simulation, predict potential faults 72 hours in advance and generate personalized maintenance work orders; Step 420, Daily Operation and Maintenance: Regularly check and maintain the hardware devices of the system, including sensor calibration, cleaning and performance detection of the intelligent battery gateway 12, etc.; update and upgrade the system software, and repair vulnerabilities and optimize functions in a timely manner; analyze the data accumulated during the operation of the system, summarize the operation rules and fault modes of the battery pack, and provide a basis for subsequent operation and maintenance decisions; when the system issues an abnormal early warning, conduct fault troubleshooting and handling in a timely manner according to the early warning information and repair suggestions.

[0056] In this embodiment, through multi-dimensional fusion perception technology and dual-drive intelligent algorithms, the aging signs of the internal microstructure of the battery are accurately captured, and the remaining power and health status of the battery are estimated with high precision. In terms of early warning performance, a multi-dimensional linkage detection model and an intelligent early warning mechanism are established to effectively improve the accuracy and timeliness of fault warnings and significantly advance the warning time. At the management and operation level, the dynamic grouping intelligent balancing strategy and fault inference engine not only improve the efficiency of battery pack management, but also realize the root cause diagnosis of faults and provide maintenance suggestions. At the same time, the edge computing architecture optimizes data processing and protects user privacy.

[0057] In addition, it should be noted that the combination of the various technical features in this case is not limited to the combination described in the claims of this case or the combination described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way unless there is a contradiction between them.

[0058] It should be noted that the above examples are only specific embodiments of the present invention, and the present invention is obviously not limited to the above examples, and there are many similar variations. All variations directly derived or associated from the contents disclosed by the technicians in this field should fall within the protection scope of the present invention.

[0059] The above are only preferred embodiments of the present invention and are not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A storage battery abnormal warning system based on an intelligent battery gateway, characterized in that Including: A sensor module for real-time acquisition of parameters such as the voltage, internal resistance, temperature, thermal runaway state, and floating charge voltage of a single battery; An intelligent battery gateway, including a data processing module and an early warning module. The data processing module performs fusion analysis on sensor data through intelligent algorithms, calculates the state of charge (SOC) of the battery remaining and the state of health (SOH), and executes an intelligent equalization strategy; The early warning module generates multi-level early warning signals according to parameter thresholds and an anomaly detection model; A communication module for local data interaction and remote platform data upload, supporting Bluetooth, serial port, and network communication; A human-machine interaction module for visual early warning of abnormal states and operation and maintenance management.

2. The battery anomaly warning system based on the intelligent battery gateway according to claim 1, wherein The sensor module includes: A composite parameter measurement unit is used to synchronously collect the voltage response curve and the AC internal resistance of a single cell by means of pulse current injection method, and obtain the electrochemical impedance spectroscopy (EIS) data including the charge transfer resistance , the double-layer capacitance ; A thermal runaway detection unit for identifying precursors of thermal runaway through a three-dimensional linkage criterion of temperature gradient, internal resistance mutation rate, and second derivative of voltage. The criterion formula is: ； In the formula, is the temperature change rate per unit time, is the change rate of the internal resistance relative to the initial value, is the curvature of the voltage curve.

3. The abnormal warning system for storage battery based on intelligent battery gateway according to claim 2, wherein The sensor module further includes a self-calibration unit, and the self-calibration unit includes: A MEMS temperature and pressure compensation module for real-time monitoring of ambient air pressure and humidity, and dynamically correcting the thermal resistance coefficient of the temperature sensor through a look-up table method. The corrected temperature measurement accuracy is ±0.5°C; A standard voltage source module for automatically calibrating the voltage acquisition channel every 24 hours. After calibration, the voltage drift error < 0.05%.

4. The abnormal warning system for storage batteries based on an intelligent battery gateway according to claim 1, wherein The SOC calculation algorithm of the data processing module is an improved volumetric Kalman filtering algorithm. The calculation of the state of charge (SOC) of the battery remaining includes: Set up a third-order RC equivalent circuit model to identify the ohmic internal resistance in real time , charge transfer resistance , double-layer capacitance and other model parameters; Construct a multi-input measurement equation to fuse voltage , internal resistance, and temperature T Construct a state-space equation: ； Among them, is the open-circuit voltage, is the ohmic voltage drop, is the current flowing through the battery, is the polarization voltage, is the rate of change of the polarization voltage with time, is the change in the polarization voltage caused by the current charging the electric double-layer capacitance, is the polarization voltage relaxation time constant; Setting an anti-outlier adaptive mechanism, judging the measurement residual through the Mahalanobis distance, and switching to the square root volumetric Kalman filter when the limit is exceeded. The calculation accuracy of SOC is ±3%.

5. The battery abnormal warning system based on the intelligent battery gateway according to claim 1, characterized in that, The SOH prediction algorithm of the data processing module is a gated recurrent unit algorithm based on an attention mechanism, including: Constructing a multi-feature input layer, extracting 12-dimensional aging features such as the internal resistance growth rate and capacity attenuation rate, and constructing a time window sliding matrix; Constructing an attention mechanism layer, calculating the contribution degree of historical data to the current SOH through weighted calculation, and focusing on the features in the accelerated aging stage. The weight formula is: ； Among them, is t the moment attention weight, is the hidden layer state, is the hidden layer state at the previous moment, is a scoring function used to measure the hidden layer state at the previous moment and the hidden layer state at the current moment for the degree of association between them; Constructing a physical constraint correction layer to correct the prediction result through a capacity-internal resistance coupling equation: ； Among them, represents the change ratio of the measured internal resistance relative to the initial internal resistance and the failure internal resistance. The coefficient 0.8 indicates that the internal resistance growth dominates the weight in the SOH assessment. is the capacity attenuation ratio. The coefficient 0.2 shows that the influence of capacity attenuation is relatively small, and the corrected SOH prediction accuracy is ±5%.

6. The abnormal warning system for storage battery based on intelligent battery gateway according to claim 1, characterized in that, The intelligent equalization strategy is dynamic grouping equalization based on impedance spectrum clustering, including: Constructing a health state stratification, clustering the principal components of the impedance spectrum through the K-means algorithm, and dividing the batteries into a healthy group, a declining group, and a fault warning group; Setting a time-varying threshold trigger mechanism, and setting different equalization thresholds according to the charge and discharge conditions: ； Monitor the internal resistance change rate during the equalization process and automatically terminate when is reached.

7. The abnormal warning system for storage batteries based on an intelligent battery gateway according to claim 1, wherein The early warning module includes: A five-dimensional feature anomaly detection unit, constructing a feature space including voltage, internal resistance, temperature, SOC, and SOH, and using the isolation forest algorithm to detect the isolation degree of data points; A Bayesian network inference engine, establishing a causal relationship knowledge base including 56 fault modes, deriving the root cause of the fault through posterior probability calculation, and outputting maintenance suggestions.

8. The battery anomaly warning system based on an intelligent battery gateway according to claim 1, characterized in that, The intelligent battery gateway adopts an edge computing architecture. The intelligent battery gateway further includes: A data hierarchical processing module for uploading abnormal data segments and 15-dimensional feature vectors, and processing regular data locally; A federated learning update module, which has a lightweight AM-GRU model built-in, for synchronizing model parameter gradients through federated learning.

9. The battery abnormal warning system based on the intelligent battery gateway according to claim 1, wherein The system further includes: Digital twin mapping module. The digital twin mapping module builds a virtual mirror of the battery pack inside the intelligent battery gateway, synchronizes the parameters of individual batteries in real time, simulates the evolution law of fault scenario parameters through finite element simulation, predicts potential faults and generates personalized maintenance work orders.

Images