A lithium battery power monitoring and low power warning system

Abstract

The present invention discloses a lithium battery power monitoring and low power warning system, and relates to the technical field of battery management. The present invention estimates SOC by combining a power depth mapping algorithm with battery chemical characteristics and a working environment, thereby improving the accuracy of SOC estimation and reducing the risk of sudden equipment shutdown due to inaccurate power display; learning user behavior patterns through an adaptive user behavior recognition and optimization algorithm, and adjusting parameters and warning thresholds of an SOC estimation model according to learning results, thereby improving the intelligence level of the warning system; monitoring the battery health state through an SOC estimation and status monitoring module, and adjusting the warning threshold according to the actual state of the battery, thereby improving the accuracy of the warning; in addition, the user behavior learning module dynamically adjusts the alarm threshold according to the user's usage habits, thereby improving the practicability and effectiveness of the warning; the present invention greatly improves the monitoring accuracy of the existing lithium battery power monitoring and low power warning system.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and more specifically to a lithium battery power monitoring and low power warning system. Background Art

[0002] With the development of science and technology and the enhancement of environmental awareness, lithium batteries have been widely used in electric vehicles, power tools, consumer electronics and other fields due to their high energy density, lightness and durability. Especially in the electric vehicle industry, lithium batteries have gradually replaced traditional lead-acid batteries as one of the mainstream power sources. However, due to the significant differences in the working principles and characteristics of lithium batteries and lead-acid batteries, the power display system on traditional vehicles often cannot accurately reflect the actual remaining power of the lithium battery, which brings many inconveniences to users, especially when the power is close to exhaustion, it may cause the device to suddenly stop working. In order to solve this problem, various low-battery alarm reminder devices have appeared on the market, but these devices still have certain limitations.

[0003] Low-battery alarms currently available on the market mainly use the vehicle's own alarm system to remind users that the battery is about to run out. Although this approach solves the problem of low battery to a certain extent, many older vehicles are not equipped with such a function, and even for new models, their power display systems are usually designed based on lead-acid batteries and cannot accurately adapt to the characteristics and working conditions of lithium batteries. Most existing low-battery alarms rely on simple voltage measurements to determine the remaining power, which is not always reliable in the application scenarios of lithium batteries. The voltage of lithium batteries changes relatively slowly during discharge, especially when the battery is close to full discharge, the voltage drops faster, which leads to inaccurate voltage-based SOC (state of charge) estimates. As a result, users may face the situation that when the device suddenly stops working, they find that the battery is actually exhausted. In addition, most existing low-battery alarms do not have intelligent functions, and they cannot dynamically adjust according to the user's usage habits or the actual state of the battery. For example, these systems cannot learn when the user usually starts charging, or cannot adjust the alarm threshold according to the health of the battery. This means that even if the battery performance deteriorates, the alarm threshold will not change accordingly, which may lead to early or delayed alarms.

[0004] Therefore, in order to better meet the needs of users, the present invention discloses a lithium battery power monitoring and low power warning system, so as to give users timely and effective reminders when the power is close to being exhausted, avoiding unnecessary troubles and losses. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention discloses a lithium battery power monitoring and low power warning system, which greatly improves the monitoring accuracy of the existing lithium battery power monitoring and low power warning system.

[0006] In order to achieve the above technical effects, the present invention adopts the following technical solutions:

[0007] A lithium battery power monitoring and low power warning system, comprising: a data acquisition module, used to collect battery operation data and environmental data in real time, and output the data to a SOC estimation and status monitoring module; the data acquisition module collects data through a sensor, and improves signal quality through a signal conditioning circuit; the battery operation data and environmental data include but are not limited to battery voltage, current, temperature and internal resistance data;

[0008] An SOC estimation and status monitoring module is used to estimate the remaining power of the battery, monitor the battery health status, and output the SOC estimation result and the battery health status assessment report based on the collected data of the data acquisition module; the battery health status includes but is not limited to capacity decay and internal resistance change; the SOC estimation and status monitoring module estimates the SOC through a power depth mapping algorithm combined with battery chemical characteristics and working environment, and monitors the battery aging trend through data analysis;

[0009] A user behavior learning module is used to receive the SOC estimation result and battery health status assessment result of the SOC estimation and status monitoring module to analyze user usage habits, optimize power estimation and early warning strategies, and output optimized parameters and thresholds to the SOC estimation and status monitoring module and early warning notification module; the user behavior learning module learns user behavior patterns through an adaptive user behavior recognition and optimization algorithm, and adjusts the parameters and early warning thresholds of the SOC estimation model according to the learning results; the user usage habits include but are not limited to charging time and discharging mode;

[0010] The early warning notification module is used to perform early warning logic judgment based on the SOC value and battery health status evaluation result of the SOC estimation and status monitoring module, and the alarm threshold parameter of the user behavior learning module, and send an early warning notification to the user through a speaker.

[0011] As a further technical solution of the present invention, the SOC estimation method of the SOC estimation and status monitoring module is:

[0012] S1, data input, obtaining the operation data and environmental data of the data acquisition module, and removing input data noise and abnormal values ​​through Kalman filtering;

[0013] S2, initial SOC setting, after system startup or battery replacement, the initial SOC value is set based on the stored data of the battery management system and the estimation result of the open circuit voltage method;

[0014] S3, multi-parameter fusion estimation of SOC, based on the battery chemical characteristics and working environment, the battery power-SOC mapping model is constructed through the power deep mapping algorithm; the battery chemical characteristics include but are not limited to battery materials and electrolyte properties; the working environment includes but is not limited to temperature, humidity and charge and discharge rate; the power-SOC mapping model fuses the input data information in S1 through a multi-parameter fusion method to perform real-time SOC estimation;

[0015] S4, ampere-hour integration method assists in estimation and error correction, the ampere-hour integration method assists in estimating the change in SOC, and performs error correction in combination with the output result of S3; the ampere-hour integration method calculates the change in battery power by integrating the charge and discharge current of the battery, and updates the SOC value. The working steps of the ampere-hour integration method are:

[0016] S401, measuring the charge and discharge current of the battery in real time through a Hall effect sensor, and converting the charge and discharge current of the battery into a digital signal through an analog-to-digital converter;

[0017] S402, obtaining a change in battery power by integrating the current over time using the Simpson integration method;

[0018] S403, converting the amount of power change into the amount of SOC change, and updating the current SOC value;

[0019] S404, combining the result of the power depth mapping algorithm, regularly correcting the estimation result of the ampere-hour integration method to eliminate the accumulated error;

[0020] S5, open circuit voltage method calibration, after the battery is left to stand for a long time, measure the open circuit voltage of the battery, and find the corresponding SOC value through the OCV-SOC curve, and calibrate the power-SOC mapping model and the estimation result of the ampere-hour integration method; the OCV-SOC curve realizes the construction of the corresponding relationship of the open circuit voltage of the battery cell under different battery remaining power values ​​through data fitting;

[0021] S6. Output the calibrated SOC value.

[0022] As a further technical solution of the present invention, the multi-parameter fusion method integrates information from different sensors and data sources through a data fusion algorithm, and the data fusion algorithm estimates the state of the linear dynamic system through Kalman filtering to remove noise and outliers in the input data; based on the complexity of battery chemical characteristics and working environment, the multi-parameter fusion method also constructs a power-SOC mapping model through a multi-parameter mapping mechanism to achieve accurate estimation of SOC; the multi-parameter mapping mechanism captures the complex nonlinear relationship between battery power and SOC through a polynomial regression method, and estimates SOC based on multi-parameter fusion processing; during the operation of the power-SOC mapping model, the multi-parameter mapping mechanism also dynamically adjusts model parameters according to changes in the current battery state and working conditions through a real-time parameter adjustment method.

[0023] As a further technical solution of the present invention, the working steps of the power depth mapping algorithm include:

[0024] Step 1: System modeling and prediction;

[0025] Based on the physical characteristics and working principle of the battery, a dynamic model of the model battery system is constructed to predict the future value of the battery state. The dynamic model is used to describe the linear and nonlinear relationship between the battery state and the change of external input over time; the external input includes but is not limited to the current and voltage values; during the prediction process, the dynamic model quantifies the uncertainty of the model through the covariance matrix; the formula expression of the dynamic model is:

[0026] { (1)

[0027] In formula (1), represents a state transfer matrix, which is used to reflect the change of battery state over time; the state transfer matrix Depends on the battery's temperature coefficient and aging factor ; is a predicted state vector, which is used to represent the predicted value of the current state of the battery, including but not limited to SOC, voltage and current integral parameters; is the input matrix, which is used to reflect the current efficiency The actual effect that affects the input current; Represents a nonlinear perturbation function, which is used to measure the complexity of the chemical reaction inside the battery; represents process noise, which is used to measure model uncertainty; is an input vector, including but not limited to current and voltage;

[0028] Step 2: Linearize the nonlinear function;

[0029] In the prediction step of the dynamic model, the nonlinear system model is linearized at the prediction point by the Jacobian matrix to allow the update rule of the power depth mapping algorithm to be applied; the formula expression for linearization using the Jacobian matrix is:

[0030] (2)

[0031] In formula (2), Represents the Jacobian matrix, which is used to linearize nonlinear models; Represents the capacity decay coefficient, which is used to measure the rate of change of battery capacity over time; Indicates the internal resistance variation coefficient, which is used to measure the rate of change of the battery's internal resistance over time; Represents the temperature influence coefficient, which is used to measure the impact of temperature changes on battery performance;

[0032] Step 3: Setting the observation equation;

[0033] At the same time, the real-time observation data of the battery is obtained through the sensor, and the relationship between the observation data and the battery state is established through the observation equation to reduce the influence of noise and interference; the real-time observation data of the battery includes but is not limited to the battery voltage and current; the formula expression of the observation equation is:

[0034] (3)

[0035] In formula (3), represents an observation vector for correcting the predicted state, the observation vector including but not limited to battery voltage and current; represents an observation parameter, which is used to adjust the observation equation, and the observation parameter includes but is not limited to battery temperature and battery age; represents the observation function, which is used to measure the change factors of the open circuit voltage and internal resistance of the battery; represents observation noise, which is used to measure the uncertainty in the observations; An additional parameter representing the observation noise, used to measure the distribution of the noise;

[0036] Step 4: Update the prediction error covariance;

[0037] Next, the difference between the observed value and the model predicted value is calculated by the prediction error covariance update function; the formula expression of the prediction error covariance update function is:

[0038] (4)

[0039] In formula (4), represents the forecast error covariance matrix, which is used to quantify the uncertainty of the forecast state; Represents the updated error covariance matrix of the previous moment; Represents the process noise covariance matrix, quantifying the statistical characteristics of the process noise; represents the process noise intensity coefficient; represents the process noise correlation coefficient;

[0040] Step 5: Kalman gain calculation and state update;

[0041] The weights of the predicted value and the observed value when updating the battery state estimation are balanced by the Kalman gain; the Kalman gain quantifies the uncertainty of the prediction and observation respectively through the prediction error covariance matrix and the observation noise covariance matrix, and minimizes the covariance of the estimation error; wherein the Kalman gain calculation and state update include the following steps:

[0042] Step 501: Calculate the contribution of the observation value in updating the state estimate through the Kalman gain, and the formula expression is:

[0043] (5)

[0044] In formula (5), is the Kalman gain matrix, which is used to determine the influence of the observation value on the state estimation; Represents the observation function Status The Jacobian matrix of is used to map the state vector to the observation space; represents the observation noise covariance matrix, which is used to measure the noise level in the observations; represents the observation noise intensity coefficient; represents the observation noise correlation coefficient;

[0045] Step 502: After obtaining a new observation value, update the state estimate through a state update function to minimize the prediction error; the formula expression of the state update function is:

[0046] (6)

[0047] In formula (6), is the updated state estimate; represents the observation vector value; represents the observation noise intensity coefficient; represents the forecast state adjustment amount; represents the Kalman gain matrix; represents the prediction error covariance matrix;

[0048] Step 503: After the state is updated, the error covariance matrix is ​​updated to reflect the uncertainty of the new estimate. The formula is:

[0049] (7)

[0050] In formula (7), represents the updated error covariance matrix, which is used to reflect the uncertainty of the updated state estimation; is the identity matrix;

[0051] Step 6: Deep learning correction;

[0052] A deep learning model is trained based on at least historical voltage, current and temperature data to build a power-SOC mapping model, predict the SOC correction amount, and output a probability distribution including a mean and a variance;

[0053] Step 7: Fusion and output;

[0054] The SOC correction value predicted by the power-SOC mapping model is fused with the estimated value of the extended Kalman filter by a weighted fusion method to generate a final SOC estimated value; the fusion formula of the weighted fusion method is expressed as:

[0055] (8)

[0056] In formula (8), is the estimated SOC value after deep learning correction; It is the SOC estimation value after the extended Kalman filter is updated; The SOC correction value predicted by the power-SOC mapping model; is the normal distribution function, which is used to measure the uncertainty of deep learning predictions; It is the mean value of the correction value predicted by the power-SOC mapping model, which is used to adjust the SOC estimation value; is the weighting coefficient used to balance the contribution of the deep learning model and the extended Kalman filter; Represents the uncertainty variance of deep learning predictions, which is used to measure the reliability of the predictions.

[0057] As a further technical solution of the present invention, the power-SOC mapping model includes an input layer, a feature extraction layer, a hidden layer, an uncertainty estimation layer and an output layer; the working principle steps of the power-SOC mapping model are as follows:

[0058] W1, receiving pre-processed battery operation data through the input layer;

[0059] W2, extracting features from input data through the feature extraction layer; the feature extraction layer extracts spatial features of time series data through a convolutional neural network, and captures temporal features of time series data through a recurrent neural network; the spatial features of the time series data are the fluctuation features of the battery voltage; the temporal features of the time series data are the changing trends of the battery current;

[0060] W3, performing feature conversion and nonlinear mapping through the hidden layer; the hidden layer performs feature conversion through multiple layers of fully connected layers to increase the expressive power of the model, and introduces nonlinear mapping through a nonlinear activation function to capture the complex relationship between the battery SOC and the input data;

[0061] W4, estimating the uncertainty of the model prediction through the uncertainty estimation layer; the uncertainty estimation layer prevents overfitting and improves the generalization ability of the model through a regularization method; and estimating the uncertainty of the model prediction through a Bayesian neural network, the output of the uncertainty estimation layer includes a probability distribution of the mean and variance;

[0062] W5. Output the final SOC estimation value and the uncertainty of the SOC estimation value through the output layer; the output layer represents the SOC estimation value and the uncertainty of the SOC estimation value through Gaussian distribution.

[0063] As a further technical solution of the present invention, the state monitoring working method of the SOC estimation and state monitoring module is:

[0064] R1. Periodically collect the battery capacity, internal resistance, charge and discharge performance, temperature and cycle number parameter data through the battery management system;

[0065] R2. Monitoring the capacity decay of the battery by a model-based capacity estimation method; the model-based capacity estimation method predicts the battery capacity decay trend by exponential smoothing method;

[0066] R3. Based on AC impedance and DC internal resistance data, analyze the change trend of internal resistance through moving average and standard deviation;

[0067] R4. The charging and discharging process of the battery is controlled by the BMS controller, and the charging and discharging efficiency is automatically calculated and continuously updated by the efficiency formula; the BMS controller controls the output of the charger and the discharger by the PWM signal; the formula expression of the efficiency formula is:

[0068] (9)

[0069] In formula (9), represents the discharge energy, Indicates charging energy;

[0070] R5. Based on the above historical cycle times, temperature, charge and discharge efficiency characteristics and corresponding battery capacity data, the battery health model is trained through support vector machine to predict battery life;

[0071] R6. Combine the monitoring data from R2 to R5 and generate a battery health report through a template engine. The battery health report includes but is not limited to the comprehensive health status, performance level and remaining life prediction information of the battery.

[0072] As a further technical solution of the present invention, the adaptive user behavior recognition and optimization algorithm includes an input layer, a data layer, a feature engineering layer, a user behavior pattern recognition layer, a behavior pattern modeling layer, an adaptive adjustment layer, an output layer and a feedback optimization layer; the working method steps of the adaptive user behavior recognition and optimization algorithm are as follows:

[0073] U1, obtain SOC estimation data and status monitoring data through the input layer, and the input layer ensures data integrity through data verification protocol CRC check;

[0074] U2, cleaning, denoising and normalizing the received raw data through the data layer; the data layer processes the data through missing value filling, outlier detection and standardization methods;

[0075] U3, extracting features from the preprocessed data through the feature engineering layer; the feature engineering layer constructs a feature set through packaging method, embedding method, polynomial features and PCA dimensionality reduction;

[0076] U4. Based on the extracted features, the user's charging and discharging usage behavior patterns are identified through the user behavior pattern recognition layer; the user behavior pattern recognition layer classifies and clusters the user behaviors through conditional random fields and k-clustering algorithms;

[0077] U5. Modeling the identified user behavior pattern through the behavior pattern modeling layer, wherein the behavior pattern modeling layer models the user behavior through time series analysis and gradient boosting tree;

[0078] U6. According to the change of user behavior pattern, the parameters and warning threshold of the SOC estimation model are dynamically adjusted through the adaptive adjustment layer; the adaptive adjustment layer automatically optimizes the model parameters and thresholds according to real-time feedback through the policy gradient method;

[0079] U7, based on the SOC estimation result, battery health status and user behavior pattern information, generates power estimation and early warning strategy through the output layer; the output layer outputs the optimal decision through the Bayesian network and rule engine;

[0080] U8. Collect user feedback and system operation data through the feedback optimization layer, evaluate the effectiveness of the current strategy, and guide the next round of optimization iterations; the feedback optimization layer collects feedback data through user satisfaction surveys and system log analysis, and evaluates the optimization effect through regression analysis.

[0081] As a further technical solution of the present invention, the early warning notification module includes an early warning logic judgment submodule, an alarm display submodule and a power management submodule; the early warning logic judgment submodule is used to receive the SOC estimation value, the battery health status assessment result and the alarm threshold parameters output by the user behavior learning module, and perform real-time data analysis and logic judgment through an embedded processor; when the alarm is triggered, the alarm display submodule broadcasts the remaining power through a speaker; the power management submodule includes a power control unit and an energy consumption optimization unit; the power control unit controls the switching and distribution of the power supply through a power management integrated circuit to ensure stable power supply to the system; the energy consumption optimization unit reduces the energy consumption of the system in an inactive state through a sleep mode and a wake-up mechanism.

[0082] As a further technical solution of the present invention, the system also includes a remote monitoring and data management module, which is used to receive data from each module, perform big data analysis to identify potential problems, and generate maintenance suggestions; the remote monitoring and data management module includes a data access submodule, a data management submodule, a big data analysis submodule and a maintenance suggestion generation submodule; the data access submodule realizes data exchange with front-end sensors and equipment through a data communication protocol, and eliminates noise and ensures the validity of data through outlier detection and missing value filling methods; the data management submodule is used to store preprocessed data and perform data management and query; the data management submodule includes a distribution The distributed database unit and the data index unit; the distributed database unit stores historical data through the NoSQL database; the data index unit improves the data retrieval speed through the inverted index; the big data analysis submodule is used to perform in-depth analysis on the stored data to identify the trend and potential failure of battery performance; the big data analysis submodule predicts the change trend of battery performance through time series analysis and moving average method, and identifies the failure mode through support vector machine; based on the analysis results of the big data analysis submodule, the maintenance suggestion generation submodule formulates a maintenance plan through an expert system and a decision tree algorithm, and automatically generates a maintenance suggestion report through natural language processing and a chart library.

[0083] Positive beneficial effects:

[0084] The present invention estimates SOC by combining the battery chemical characteristics and working environment with the power depth mapping algorithm, solves the problem that the traditional voltage-based SOC estimation method is inaccurate in lithium battery applications, improves the accuracy of SOC estimation, and reduces the risk of sudden equipment shutdown due to inaccurate power display. The user behavior pattern is learned by the adaptive user behavior recognition and optimization algorithm, and the parameters and warning threshold of the SOC estimation model are adjusted according to the learning results, which solves the problem that the existing warning system cannot be dynamically adjusted according to the user's usage habits, improves the intelligence level of the warning system, and reduces the situation of false alarms or missed alarms. The battery health status is monitored by the SOC estimation and status monitoring module, including capacity attenuation and internal resistance changes, and the battery aging trend is monitored by data analysis, which solves the problem that the existing system cannot accurately monitor the battery health status, can timely discover the situation of battery performance degradation, and adjust the warning threshold according to the actual state of the battery, thereby improving the accuracy of the warning. In addition, the user behavior learning module dynamically adjusts the alarm threshold according to the user's usage habits, solves the problem that the existing warning system cannot adapt to the usage habits of different users, provides a personalized warning strategy, and improves the user experience. Secondly, the early warning notification module performs early warning logic judgment based on the SOC estimation results, battery health status assessment results and the alarm threshold parameters of the user behavior learning module, which solves the problems of simple judgment logic and lack of intelligent judgment in the existing early warning system. It can more accurately judge when to issue a early warning notification, reduce unnecessary alarms, and improve the practicality and effectiveness of the early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work, among which:

[0086] Figure 1 A framework diagram of a lithium battery power monitoring and low power warning system of the present invention;

[0087] Figure 2 A working method step diagram of the SOC estimation method of the SOC estimation and state monitoring module of the present invention;

[0088] Figure 3 It is a step diagram of the algorithm working principle of the power depth mapping algorithm of the present invention;

[0089] Figure 4 It is a principle framework diagram of the remote monitoring and data management module of the present invention;

[0090] Figure 5It is a framework diagram of the early warning notification module of the present invention. DETAILED DESCRIPTION

[0091] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0092] like Figure 1-Figure 5 As shown, a lithium battery power monitoring and low power warning system includes:

[0093] A data acquisition module is used to collect the operating data and environmental data of the battery in real time, and output the data to the SOC estimation and status monitoring module; the data acquisition module collects data through sensors and improves the signal quality through signal conditioning circuits; the operating data and environmental data of the battery include but are not limited to battery voltage, current, temperature and internal resistance data;

[0094] An SOC estimation and status monitoring module is used to estimate the remaining power of the battery, monitor the battery health status, and output the SOC estimation result and the battery health status assessment report based on the collected data of the data acquisition module; the battery health status includes but is not limited to capacity decay and internal resistance change; the SOC estimation and status monitoring module estimates the SOC through a power depth mapping algorithm combined with battery chemical characteristics and working environment, and monitors the battery aging trend through data analysis;

[0095] A user behavior learning module is used to receive the SOC estimation result and battery health status assessment result of the SOC estimation and status monitoring module to analyze user usage habits, optimize power estimation and early warning strategies, and output optimized parameters and thresholds to the SOC estimation and status monitoring module and early warning notification module; the user behavior learning module learns user behavior patterns through an adaptive user behavior recognition and optimization algorithm, and adjusts the parameters and early warning thresholds of the SOC estimation model according to the learning results; the user usage habits include but are not limited to charging time and discharging mode;

[0096] The early warning notification module is used to perform early warning logic judgment based on the SOC value and battery health status evaluation result of the SOC estimation and status monitoring module, and the alarm threshold parameter of the user behavior learning module, and send an early warning notification to the user through a speaker.

[0097] In a specific embodiment, the lithium battery power monitoring and low power warning system monitors the key parameters of the battery such as voltage, current, temperature and internal resistance in real time through the data acquisition module. These parameters are the basis for evaluating the battery status. The SOC estimation and status monitoring module adopts a power depth mapping algorithm, which can convert the collected data into an estimated value of the remaining power SOC of the battery according to the chemical characteristics and working environment of the battery. At the same time, by monitoring the change of the internal resistance and capacity decay of the battery, the health status and aging trend of the battery are evaluated. The user behavior learning module uses machine learning technology to analyze the user's usage habits, such as charging and discharging patterns, so as to optimize the SOC estimation model and warning strategy. The warning notification module performs logical judgment and issues a warning based on the SOC estimation results and the battery health status, combined with the parameters provided by the user behavior learning module. Its specific implementation process is: first, it includes the installation of sensors to collect the operation and environmental data of the battery. Subsequently, the signal conditioning circuit is designed and implemented to ensure the accuracy and reliability of the data. Next, the algorithm of the SOC estimation and status monitoring module is developed, which usually involves complex mathematical models and data analysis techniques. The development of the user behavior learning module requires the integration of machine learning algorithms to identify and adapt to the user's behavior patterns. Finally, the implementation of the early warning notification module requires the design of a logical judgment mechanism and its integration with the user interface or notification system to ensure that users can receive early warning information in a timely manner.

[0098] The sensors of the data acquisition module must be able to accurately measure the key parameters of the battery, and the signal conditioning circuit needs to properly amplify, filter and convert these signals to meet the needs of subsequent processing. The algorithm of the SOC estimation and status monitoring module needs to consider the dynamic characteristics and long-term performance changes of the battery to achieve accurate SOC estimation and health status assessment. The user behavior learning module needs to continuously learn and optimize the model from actual usage data to adapt to the behavioral habits of different users. The logical judgment mechanism of the early warning notification module needs to comprehensively consider the SOC value, battery health status and user behavior to determine when to issue an early warning.

[0099] This lithium battery power monitoring and low power warning system can predict the battery aging trend and detect potential problems in advance through real-time monitoring and intelligent analysis, thereby extending the battery life. In addition, the system can also optimize power estimation and warning strategies based on user habits to improve user experience. In electric vehicles, mobile devices, and other applications that require precise battery management, this system can effectively reduce energy waste and improve energy efficiency, which is of great significance to promoting sustainable development and environmental protection.

[0100] In the above embodiment, the SOC estimation method of the SOC estimation and status monitoring module is:

[0101] S1, data input, obtaining the operation data and environmental data of the data acquisition module, and removing input data noise and abnormal values ​​through Kalman filtering;

[0102] S2, initial SOC setting, after system startup or battery replacement, the initial SOC value is set based on the stored data of the battery management system and the estimation result of the open circuit voltage method;

[0103] S3, multi-parameter fusion estimation of SOC, based on the battery chemical characteristics and working environment, the battery power-SOC mapping model is constructed through the power deep mapping algorithm; the battery chemical characteristics include but are not limited to battery materials and electrolyte properties; the working environment includes but is not limited to temperature, humidity and charge and discharge rate; the power-SOC mapping model fuses the input data information in S1 through a multi-parameter fusion method to perform real-time SOC estimation;

[0104] S4, ampere-hour integration method assists in estimation and error correction, the ampere-hour integration method assists in estimating the change in SOC, and performs error correction in combination with the output result of S3; the ampere-hour integration method calculates the change in battery power by integrating the charge and discharge current of the battery, and updates the SOC value. The working steps of the ampere-hour integration method are:

[0105] S401, measuring the charge and discharge current of the battery in real time through a Hall effect sensor, and converting the charge and discharge current of the battery into a digital signal through an analog-to-digital converter;

[0106] S402, obtaining a change in battery power by integrating the current over time using the Simpson integration method;

[0107] S403, converting the amount of power change into the amount of SOC change, and updating the current SOC value;

[0108] S404, combining the result of the power depth mapping algorithm, regularly correcting the estimation result of the ampere-hour integration method to eliminate the accumulated error;

[0109] S5, open circuit voltage method calibration, after the battery is left to stand for a long time, measure the open circuit voltage of the battery, and find the corresponding SOC value through the OCV-SOC curve, and calibrate the power-SOC mapping model and the estimation result of the ampere-hour integration method; the OCV-SOC curve realizes the construction of the corresponding relationship of the open circuit voltage of the battery cell under different battery remaining power values ​​through data fitting;

[0110] S6. Output the calibrated SOC value.

[0111] In the above embodiment, the multi-parameter fusion method integrates information from different sensors and data sources through a data fusion algorithm, and the data fusion algorithm estimates the state of the linear dynamic system through Kalman filtering to remove noise and outliers in the input data; based on the complexity of the battery chemical characteristics and the working environment, the multi-parameter fusion method also constructs a power-SOC mapping model through a multi-parameter mapping mechanism to achieve accurate estimation of SOC; the multi-parameter mapping mechanism captures the complex nonlinear relationship between battery power and SOC through a polynomial regression method, and estimates SOC based on multi-parameter fusion processing; during the operation of the power-SOC mapping model, the multi-parameter mapping mechanism also dynamically adjusts model parameters according to changes in the current battery state and working conditions through a real-time parameter adjustment method.

[0112] In the above embodiment, the working steps of the power depth mapping algorithm include:

[0113] Step 1: System modeling and prediction;

[0114] Based on the physical characteristics and working principle of the battery, a dynamic model of the model battery system is constructed to predict the future value of the battery state. The dynamic model is used to describe the linear and nonlinear relationship between the battery state and the change of external input over time; the external input includes but is not limited to the current and voltage values; during the prediction process, the dynamic model quantifies the uncertainty of the model through the covariance matrix; the formula expression of the dynamic model is:

[0115] { (1)

[0116] In formula (1), represents a state transfer matrix, which is used to reflect the change of battery state over time; the state transfer matrix Depends on the battery's temperature coefficient and aging factor ; is a predicted state vector, which is used to represent the predicted value of the current state of the battery, including but not limited to SOC, voltage and current integral parameters; is the input matrix, which is used to reflect the current efficiency The actual effect that affects the input current; Represents a nonlinear perturbation function, which is used to measure the complexity of the chemical reaction inside the battery; represents process noise, which is used to measure model uncertainty; is an input vector, including but not limited to current and voltage;

[0117] Step 2: Linearize the nonlinear function;

[0118] In the prediction step of the dynamic model, the nonlinear system model is linearized at the prediction point by the Jacobian matrix to allow the update rule of the power depth mapping algorithm to be applied; the formula expression for linearization using the Jacobian matrix is:

[0119] (2)

[0120] In formula (2), Represents the Jacobian matrix, which is used to linearize nonlinear models; Represents the capacity decay coefficient, which is used to measure the rate of change of battery capacity over time; Indicates the internal resistance variation coefficient, which is used to measure the rate of change of the battery's internal resistance over time; Represents the temperature influence coefficient, which is used to measure the impact of temperature changes on battery performance;

[0121] Step 3: Setting the observation equation;

[0122] At the same time, the real-time observation data of the battery is obtained through the sensor, and the relationship between the observation data and the battery state is established through the observation equation to reduce the influence of noise and interference; the real-time observation data of the battery includes but is not limited to the battery voltage and current; the formula expression of the observation equation is:

[0123] (3)

[0124] In formula (3), represents an observation vector for correcting the predicted state, the observation vector including but not limited to battery voltage and current; represents an observation parameter, which is used to adjust the observation equation, and the observation parameter includes but is not limited to battery temperature and battery age; represents the observation function, which is used to measure the change factors of the open circuit voltage and internal resistance of the battery; represents observation noise, which is used to measure the uncertainty in the observations; An additional parameter representing the observation noise, used to measure the distribution of the noise;

[0125] Step 4: Update the prediction error covariance;

[0126] Next, the difference between the observed value and the model predicted value is calculated by the prediction error covariance update function; the formula expression of the prediction error covariance update function is:

[0127] (4)

[0128] In formula (4), represents the forecast error covariance matrix, which is used to quantify the uncertainty of the forecast state; Represents the updated error covariance matrix of the previous moment; Represents the process noise covariance matrix, quantifying the statistical characteristics of the process noise; represents the process noise intensity coefficient; represents the process noise correlation coefficient;

[0129] Step 5: Kalman gain calculation and state update;

[0130] The weights of the predicted value and the observed value when updating the battery state estimation are balanced by the Kalman gain; the Kalman gain quantifies the uncertainty of the prediction and observation respectively through the prediction error covariance matrix and the observation noise covariance matrix, and minimizes the covariance of the estimation error; wherein the Kalman gain calculation and state update include the following steps:

[0131] Step 501: Calculate the contribution of the observation value in updating the state estimate through the Kalman gain, and the formula expression is:

[0132] (5)

[0133] In formula (5), is the Kalman gain matrix, which is used to determine the influence of the observation value on the state estimation; Represents the observation function Status The Jacobian matrix of is used to map the state vector to the observation space; represents the observation noise covariance matrix, which is used to measure the noise level in the observations; represents the observation noise intensity coefficient; represents the observation noise correlation coefficient;

[0134] Step 502: After obtaining a new observation value, update the state estimate through a state update function to minimize the prediction error; the formula expression of the state update function is:

[0135] (6)

[0136] In formula (6), is the updated state estimate; represents the observation vector value; represents the observation noise intensity coefficient; represents the forecast state adjustment amount; represents the Kalman gain matrix; represents the prediction error covariance matrix;

[0137] Step 503: After the state is updated, the error covariance matrix is ​​updated to reflect the uncertainty of the new estimate. The formula is:

[0138] (7)

[0139] In formula (7), represents the updated error covariance matrix, which is used to reflect the uncertainty of the updated state estimation; is the identity matrix;

[0140] Step 6: Deep learning correction;

[0141] A deep learning model is trained based on at least historical voltage, current and temperature data to build a power-SOC mapping model, predict the SOC correction amount, and output a probability distribution including a mean and a variance;

[0142] Step 7: Fusion and output;

[0143] The SOC correction value predicted by the power-SOC mapping model is fused with the estimated value of the extended Kalman filter by a weighted fusion method to generate a final SOC estimated value; the fusion formula of the weighted fusion method is expressed as:

[0144] (8)

[0145] In formula (8), is the estimated SOC value after deep learning correction; It is the SOC estimation value after the extended Kalman filter is updated; The SOC correction value predicted by the power-SOC mapping model; is the normal distribution function, which is used to measure the uncertainty of deep learning predictions; It is the mean value of the correction value predicted by the power-SOC mapping model, which is used to adjust the SOC estimation value; is the weighting coefficient used to balance the contribution of the deep learning model and the extended Kalman filter; Represents the uncertainty variance of deep learning predictions, which is used to measure the reliability of the predictions.

[0146] In the above embodiment, the power-SOC mapping model includes an input layer, a feature extraction layer, a hidden layer, an uncertainty estimation layer and an output layer; the working principle steps of the power-SOC mapping model are as follows:

[0147] W1, receiving pre-processed battery operation data through the input layer;

[0148] W2, extracting features from input data through the feature extraction layer; the feature extraction layer extracts spatial features of time series data through a convolutional neural network, and captures temporal features of time series data through a recurrent neural network; the spatial features of the time series data are the fluctuation features of the battery voltage; the temporal features of the time series data are the changing trends of the battery current;

[0149] W3, performing feature conversion and nonlinear mapping through the hidden layer; the hidden layer performs feature conversion through multiple layers of fully connected layers to increase the expressive power of the model, and introduces nonlinear mapping through a nonlinear activation function to capture the complex relationship between the battery SOC and the input data;

[0150] W4, estimating the uncertainty of the model prediction through the uncertainty estimation layer; the uncertainty estimation layer prevents overfitting and improves the generalization ability of the model through a regularization method; and estimating the uncertainty of the model prediction through a Bayesian neural network, the output of the uncertainty estimation layer includes a probability distribution of the mean and variance;

[0151] W5. Output the final SOC estimation value and the uncertainty of the SOC estimation value through the output layer; the output layer represents the SOC estimation value and the uncertainty of the SOC estimation value through Gaussian distribution.

[0152] In a specific embodiment, the core of SOC estimation is to use the voltage, current, temperature and other parameters collected in real time by the data acquisition module to calculate through the power depth mapping algorithm. The power depth mapping algorithm first builds a relationship model between the power and voltage and current of the battery under different working conditions based on the chemical characteristics of the battery, such as the rated capacity and discharge curve of the battery. This model takes into account multiple factors such as battery internal resistance, self-discharge, temperature influence and charge and discharge efficiency to ensure the accuracy and reliability of the estimation.

[0153] In actual operation, the SOC estimation and status monitoring module first obtains the battery's real-time voltage and current data through high-precision sensors, and considers the battery's temperature information. These data are transmitted to the processing unit in real time. The processing unit uses the power depth mapping algorithm and combines the chemical characteristics of the battery to conduct a comprehensive analysis of the collected data to calculate the current battery SOC value. This process involves accurately calculating the accumulated power during the battery's charging and discharging process, while considering the impact of factors such as charging and discharging efficiency, internal resistance loss, and self-discharge on SOC estimation.

[0154] In order to improve the accuracy of SOC estimation, the module also adopts a variety of optimization strategies. For example, the extended Kalman filter algorithm is used to dynamically correct the SOC estimation results to reduce the cumulative error caused by factors such as current measurement error, battery capacity change and temperature fluctuation. In addition, the current capacity of the battery is calibrated by regularly performing full charge or full discharge tests to update the key parameters in the power depth mapping algorithm to ensure the long-term accuracy of SOC estimation.

[0155] In terms of battery health status monitoring, the SOC estimation and status monitoring module evaluates the battery aging trend by analyzing key indicators such as battery capacity decay and internal resistance change. This process relies on the accumulation and analysis of historical data, and identifies the trend and cause of battery performance degradation by comparing the initial performance of the battery with the current performance. At the same time, the module also uses data analysis technologies, such as machine learning algorithms, to predict and warn the battery health status, providing strong support for system maintenance.

[0156] Among them, the power depth mapping algorithm is an algorithm that builds a relationship model between power and parameters such as voltage and current based on the battery's chemical characteristics and working environment. First, the power depth mapping algorithm builds a relationship model between the battery's power and voltage and current under different conditions based on the battery's chemical characteristics (such as the battery's rated capacity, discharge curve, etc.) and working environment (such as temperature, charge and discharge rate, etc.). This model needs to take into account multiple factors such as battery internal resistance, self-discharge, temperature effects, and charge and discharge efficiency. Next, the battery's voltage, current, and temperature parameters are collected in real time through high-precision sensors. These data are the basic inputs of the power depth mapping algorithm. Then, the remaining power (SOC) of the battery is estimated using the constructed relationship model combined with the real-time collected data. This process needs to take into account the accumulation of power during the battery's charge and discharge process, as well as the effects of factors such as charge and discharge efficiency, internal resistance loss, and self-discharge on SOC estimation. In order to improve the accuracy of power estimation, the power depth mapping algorithm also needs to optimize the model. This includes regular calibration of the battery's current capacity, updating key parameters in the relationship model, and using more advanced algorithms and technologies to reduce errors.

[0157] In the battery management system, the extended Kalman filter algorithm usually describes the change of SOC over time through the state equation, but due to the complexity of the battery charging and discharging process, this change is often nonlinear. Therefore, in the extended Kalman filter, the state equation is expressed in the form of a nonlinear function. The core of the extended Kalman filter is to convert nonlinear problems into linear problems through linearization techniques (usually Taylor series expansion and ignoring high-order terms). In each iteration, the algorithm linearizes the nonlinear function at the current estimated state to obtain an approximate Jacobian matrix (that is, the matrix of partial derivatives of the function with respect to the state). The specific prediction steps are similar to those of the standard Kalman filter. The extended Kalman filter first predicts the current state and its covariance matrix based on the linearized state equation and the state estimate at the previous moment.

[0158] When new observations arrive, the extended Kalman filter uses the linearized observation equation and Kalman gain matrix to combine the predicted and observed values ​​to update the state estimate and its covariance matrix. The calculation of the Kalman gain matrix takes into account the uncertainty of the prediction (covariance matrix) and the covariance of the observation noise. In addition, the extended Kalman filter algorithm is a recursive process that dynamically adjusts the state estimate based on the new observation data by repeatedly predicting and updating, thereby improving the accuracy and robustness of SOC estimation.

[0159] In the lithium battery power monitoring and low power warning system, the extended Kalman filter algorithm can achieve accurate estimation of battery SOC by effectively processing the nonlinear characteristics of the battery system and combining the initial information and real-time observation data provided by the power depth mapping algorithm. At the same time, it can also adapt to changes in battery status and deal with noise and interference in the system, providing strong support for the optimization of the battery management system and battery maintenance.

[0160] In specific implementation, the power depth mapping algorithm achieves accurate estimation of SOC by constructing a complex mapping relationship between battery power (SOC) and measurable parameters (such as voltage, current, temperature, etc.). Its technical principle is based on the electrochemical characteristics of the battery and the influence of the actual working environment. Through mathematical modeling and data analysis, it can accurately predict the remaining power of the battery. Specifically, the estimation of battery SOC first depends on a deep understanding of the electrochemical process inside the battery. During the charging and discharging process of the battery, its parameters such as voltage, current and temperature will change with the change of SOC, and these change relationships are usually nonlinear. Therefore, the power depth mapping algorithm needs to be based on the electrochemical model of the battery, considering factors such as battery internal resistance, polarization effect, and self-discharge. Based on the understanding of the electrochemical characteristics of the battery, the power depth mapping algorithm collects a large amount of experimental data and uses statistical and machine learning methods to establish a mapping model between battery SOC and measurable parameters. This model needs to be able to accurately reflect the behavioral characteristics of the battery under different working conditions, including SOC changes at different charging and discharging rates and different temperature conditions. Since the performance of the battery will change with the use time and environmental conditions (such as capacity decay, internal resistance increase, etc.), the power depth mapping algorithm needs to have the ability to dynamically adjust and optimize. This is usually achieved through regular calibration, model update and parameter adjustment to ensure the accuracy and reliability of SOC estimation. In specific implementation, the power depth mapping algorithm significantly improves the accuracy and reliability of SOC estimation by constructing a precise mapping relationship between battery power and measurable parameters. In addition, accurate SOC estimation helps to achieve precise charge and discharge control of the battery, avoid harmful operations such as overcharging and over-discharging, and thus extend the service life of the battery. Secondly, the implementation of the low power warning function depends on accurate SOC estimation. The power depth mapping algorithm provides reliable data support for the low power warning system, which helps to improve the safety and stability of the entire system. At the same time, accurate SOC estimation helps to achieve optimal configuration and management of energy, improve energy utilization efficiency, and reduce energy consumption costs.

[0161] Specifically, Experiment 1 is designed to verify the accuracy and reliability of the power depth mapping algorithm in the SOC estimation of lithium batteries. The experimental equipment includes: Lithium battery test system: used to control the charging and discharging process of the battery and record related data. High-precision voltage and current sensors: used to monitor the voltage and current of the battery in real time. Temperature controller: used to control the experimental environment temperature to ensure the consistency of experimental conditions. Power depth mapping algorithm software: used to process experimental data and realize SOC estimation. The experimental steps are as follows:

[0162] Preparation stage: Select 10 lithium batteries with different serial numbers as experimental objects, and ensure that each battery is fully charged before testing.

[0163] Set experimental parameters: According to the experimental design, set the initial SOC, ambient temperature, discharge current and other parameters of each battery.

[0164] Start discharge test: perform discharge test on the battery according to preset parameters and record data such as voltage, current and time during the discharge process.

[0165] Data recording: When the battery voltage drops to the preset termination voltage, the discharge is stopped and the voltage value and discharge time at this time are recorded.

[0166] Actual SOC measurement: The actual SOC value of each battery at the end of discharge is measured by laboratory standard methods (such as coulomb counter method) or high-precision battery tester.

[0167] Error calculation: Compare the SOC value estimated by the algorithm with the actual measured SOC value and calculate the error percentage. The error calculation formula is: Error (%) = [(estimated SOC - actual SOC) / actual SOC] * 100%.

[0168] The experimental data records are shown in Table 1:

[0169] Table 1 Algorithm experimental test table

[0170] Battery serial number Initial SOC(%) Ambient temperature (℃) Discharge current(A) Discharge time (hours) End voltage(V) Estimated SOC (%) Actual SOC(%) error(%) B001 100 25 1.0 3.0 3.00 69.8 70.2 -0.57 B002 85 20 1.5 2.5 2.90 60.1 59.9 0.33 B003 70 25 2.0 2.0 2.80 45.2 45.3 -0.20 B004 55 30 0.8 4.0 2.65 28.3 28.1 0.71 B005 40 15 1.2 3.5 2.55 15.6 15.8 -1.27 B006 25 25 0.5 6.0 2.40 7.8 7.7 1.30 B007 10 35 1.8 1.0 3.10 3.2 3.1 3.23 B008 80 10 3.0 1.5 2.70 50.0 50.1 -0.20 B009 65 20 0.7 5.0 2.50 30.0 29.8 0.67 B010 35 30 2.5 1.2 2.30 10.5 10.6 -0.94

[0171] As can be seen from Table 1, the power depth mapping algorithm can accurately estimate the SOC of lithium batteries in most cases, with an error within an acceptable range. The estimation accuracy of the algorithm may be affected to a certain extent under certain conditions, such as extreme temperature, high current discharge, etc.

[0172] In addition, Experiment 2 is designed to evaluate the performance of the power depth mapping algorithm and the traditional SOC estimation and battery health status monitoring algorithms under different conditions; the traditional SOC estimation and battery health status monitoring algorithms are tested using the open circuit voltage method, ampere-hour integration method, and discharge experiment method; the experimental data table is shown in Table 2:

[0173] Table 2 Experimental test data comparison table

[0174] Experimental conditions Test items Power Depth Mapping Algorithm Traditional Algorithms Improved results Initial SOC: 100% Estimated error when discharging to 50% 1.2% 3.5% Error reduction of 65.7% Room temperature 25°C Discharge to 20% estimated error 0.8% 4.2% Error reduction of 80.9% Discharge current: 1C Discharge to 10% estimated error 0.6% 5.1% Error reduction of 88.2% High temperature 50°C Estimated error when discharging to 50% 1.5% 5.0% Error reduction 70.0% Low temperature -10°C Discharge to 20% estimated error 1.8% 6.8% Error reduction of 73.5% Aging battery (300 cycles) Estimated error when discharging to 50% 2.0% 7.5% Error reduction of 73.3% Capacity decay monitoring accuracy 98.5% (actual attenuation) 95.0% (actual attenuation) Accuracy improved by 3.5% Internal resistance change monitoring accuracy 97.0% (actual change) 92.0% (actual change) Accuracy improved by 5.0%

[0175] By comparison, it can be seen that the power depth mapping algorithm is significantly better than the traditional algorithm in terms of SOC estimation accuracy, especially under extreme conditions (such as high temperature, low temperature) and battery aging, its estimation error is smaller. At the same time, in terms of battery health status monitoring, the algorithm also shows higher accuracy and can more accurately monitor the battery's capacity decay and internal resistance changes; experimental data show that the power depth mapping algorithm has significant positive and beneficial effects in lithium battery power monitoring and low power warning systems, which not only improves the accuracy of SOC estimation, but also enhances the reliability of battery health status monitoring. This is of great significance for extending battery life and improving system safety and stability.

[0176] In the above embodiment, the state monitoring working method of the SOC estimation and state monitoring module is:

[0177] R1. Periodically collect the battery capacity, internal resistance, charge and discharge performance, temperature and cycle number parameter data through the battery management system;

[0178] R2. Monitoring the capacity decay of the battery by a model-based capacity estimation method; the model-based capacity estimation method predicts the battery capacity decay trend by exponential smoothing method;

[0179] R3. Based on AC impedance and DC internal resistance data, analyze the change trend of internal resistance through moving average and standard deviation;

[0180] R4. The charging and discharging process of the battery is controlled by the BMS controller, and the charging and discharging efficiency is automatically calculated and continuously updated by the efficiency formula; the BMS controller controls the output of the charger and the discharger by the PWM signal; the formula expression of the efficiency formula is:

[0181] (9)

[0182] In formula (9), represents the discharge energy, Indicates charging energy;

[0183] R5. Based on the above historical cycle times, temperature, charge and discharge efficiency characteristics and corresponding battery capacity data, the battery health model is trained through support vector machine to predict battery life;

[0184] R6. Combine the monitoring data from R2 to R5 and generate a battery health report through a template engine. The battery health report includes but is not limited to the comprehensive health status, performance level and remaining life prediction information of the battery.

[0185] In a specific embodiment, periodic data acquisition uses high-precision sensors and data acquisition systems (DAQ) to monitor key battery parameters such as capacity, internal resistance, charge and discharge performance, temperature and number of cycles in real time. At the same time, advanced filtering algorithms such as Kalman filtering or particle filtering are used to improve data quality and reduce the impact of environmental noise and measurement errors. Model-based capacity estimation uses electrochemical models, such as equivalent circuit models (ECM), to model battery capacity. Exponential smoothing methods, such as the Holt-Winters method, are applied for trend prediction, which captures trends and seasonal changes in data through exponentially weighted moving averages. The internal resistance change trend analysis obtains the AC impedance characteristics of the battery through electrochemical impedance spectroscopy (EIS) technology. The moving average algorithm is used to filter out high-frequency noise, and the standard deviation analysis is used to evaluate the statistical significance of internal resistance changes, thereby identifying signs of battery aging. Automatic calculation of charge and discharge efficiency uses the PWM signal of the BMS controller to accurately control the charge and discharge process and achieve accurate measurement of energy input and output. Through the efficiency formula, the discharge energy is compared with the charging energy to quantify the energy conversion efficiency of the battery; the battery health model training collects historical data, including the number of cycles, temperature, charge and discharge efficiency, etc., as a training data set. Then the support vector machine (SVM) algorithm is applied to process the high-dimensional feature space through the kernel technique to train the classification or regression model of the battery health status and predict the battery life. The battery health report generation uses the template engine and natural language generation (NLG) technology to automatically generate an easy-to-understand battery health report based on the monitoring data and model prediction results. The report content includes the comprehensive health status, performance level and remaining life prediction of the battery, providing decision support for battery maintenance and replacement.

[0186] In the specific implementation, when implementing the condition monitoring working method, it is first necessary to establish a high-precision data acquisition system to ensure that the operating parameters of the battery can be accurately collected. Then, a model-based capacity estimation algorithm is developed to implement the prediction model of the exponential smoothing method. Next, an analysis tool for the internal resistance change trend is implemented, including moving average and standard deviation calculations. In addition, the PWM signal control logic of the BMS controller and the automatic calculation module of the charge and discharge efficiency are developed. Subsequently, the SVM model is trained using the collected data to establish a battery health prediction model. Finally, a template engine is developed to generate a detailed battery health report based on the monitoring data.

[0187] In addition, the implementation of the condition monitoring working method can not only monitor the health status of the battery in real time, detect performance degradation and potential failures in a timely manner, but also predict the aging trend of the battery, providing a scientific basis for battery maintenance and replacement. In addition, through condition monitoring, the battery usage strategy can be optimized, the battery life can be extended, energy waste can be reduced, and the economy and environmental friendliness of the battery system can be improved. In the long run, the development of condition monitoring technology will promote the development of battery management systems to a higher level of intelligence and automation, and provide strong technical support for the development of the new energy field.

[0188] In the above embodiment, the adaptive user behavior recognition and optimization algorithm includes an input layer, a data layer, a feature engineering layer, a user behavior pattern recognition layer, a behavior pattern modeling layer, an adaptive adjustment layer, an output layer and a feedback optimization layer; the working method steps of the adaptive user behavior recognition and optimization algorithm are:

[0189] U1, obtain SOC estimation data and status monitoring data through the input layer, and the input layer ensures data integrity through data verification protocol CRC check;

[0190] U2, cleaning, denoising and normalizing the received raw data through the data layer; the data layer processes the data through missing value filling, outlier detection and standardization methods;

[0191] U3, extracting features from the preprocessed data through the feature engineering layer; the feature engineering layer constructs a feature set through packaging method, embedding method, polynomial features and PCA dimensionality reduction;

[0192] U4. Based on the extracted features, the user's charging and discharging usage behavior patterns are identified through the user behavior pattern recognition layer; the user behavior pattern recognition layer classifies and clusters the user behaviors through conditional random fields and k-clustering algorithms;

[0193] U5. Modeling the identified user behavior pattern through the behavior pattern modeling layer, wherein the behavior pattern modeling layer models the user behavior through time series analysis and gradient boosting tree;

[0194] U6. According to the change of user behavior pattern, the parameters and warning threshold of the SOC estimation model are dynamically adjusted through the adaptive adjustment layer; the adaptive adjustment layer automatically optimizes the model parameters and thresholds according to real-time feedback through the policy gradient method;

[0195] U7, based on the SOC estimation result, battery health status and user behavior pattern information, generates power estimation and early warning strategy through the output layer; the output layer outputs the optimal decision through the Bayesian network and rule engine;

[0196] U8. Collect user feedback and system operation data through the feedback optimization layer, evaluate the effectiveness of the current strategy, and guide the next round of optimization iterations; the feedback optimization layer collects feedback data through user satisfaction surveys and system log analysis, and evaluates the optimization effect through regression analysis.

[0197] In a specific embodiment, the SOC estimation and status monitoring module accurately monitors the health status of the battery by combining the power depth mapping algorithm, battery chemical characteristics and working environment parameters. The module predicts the battery aging trend by analyzing the characteristics of capacity decay and internal resistance change that occur during the use of the battery. Specifically, the power depth mapping algorithm realizes real-time estimation of the remaining battery power by establishing a nonlinear mapping relationship between battery voltage, current, temperature and other parameters and SOC. The algorithm not only takes into account the basic electrochemical characteristics of the battery, but also integrates the complex influence of the working environment, such as temperature fluctuations, changes in charge and discharge rates, etc. In addition, the performance of lithium batteries is significantly affected by their internal chemical reaction processes. This module uses the electrochemical behavior characteristics of the battery during the charge and discharge process, such as the efficiency of lithium ion insertion and deinsertion, the activity changes of electrode materials, etc., to evaluate the health status of the battery. Secondly, by monitoring the environmental parameters such as temperature, humidity, vibration, etc. of the battery operation, the influence of these factors on the battery performance is analyzed, so as to more accurately evaluate the degree of battery aging and remaining life.

[0198] Among them, the specific implementation process of the condition monitoring module includes the following steps:

[0199] Data acquisition: First, use the data acquisition module to obtain the battery's voltage, current, temperature and other parameters in real time. These data are the basis for subsequent analysis and processing.

[0200] Preprocessing: Preprocess the collected data, including filtering, denoising, correction, etc., to ensure the accuracy and reliability of the data.

[0201] Power Deep Mapping: Using the power deep mapping algorithm, the real-time SOC value of the battery is calculated based on the pre-processed data. At the same time, the dynamic performance of the battery is evaluated by comparing the SOC changes under different working conditions.

[0202] Health status monitoring: Based on the SOC estimation, the battery capacity attenuation and internal resistance change are further analyzed. This is usually achieved by comparing the difference between the current performance parameters and the initial or historical performance parameters. In addition, the electrochemical model is combined to simulate the internal reaction process of the battery and predict the aging trend of the battery.

[0203] Report output: Output the SOC estimation results and battery health status assessment report to the user. The report content includes information such as the remaining battery power, health status level, and estimated remaining life.

[0204] In specific implementation, the module integrates multi-dimensional data such as battery voltage, current, temperature, etc., comprehensively considers the performance of the battery in a complex working environment, and improves the accuracy and comprehensiveness of monitoring. The module also introduces the electrochemical model of the battery, combines the macroscopic performance of the battery with the internal microscopic reaction process, and makes the monitoring results more consistent with the actual working conditions of the battery. During the monitoring process, the algorithm parameters and model parameters are dynamically adjusted according to the actual performance of the battery and changes in the working environment to ensure the real-time and accuracy of the monitoring results.

[0205] In the specific implementation, this module estimates the remaining power and health status of the battery in real time, so that users can arrange the battery usage plan more reasonably, avoid improper operations such as over-discharge and over-charging, and thus improve the battery usage efficiency. In addition, by predicting and monitoring the battery aging trend, users can promptly discover and deal with potential battery problems, avoid premature failure of the battery due to long-term poor working conditions, and thus extend the battery life. In key areas such as electric vehicles and energy storage systems, the stable operation of batteries is crucial to the safety and reliability of the entire system. This state monitoring method can promptly detect battery health problems and provide strong guarantees for the stable operation of the system. At the same time, the successful application of this method in the field of battery monitoring will provide strong support for the further development of related technologies. At the same time, with the continuous advancement of technologies such as big data and artificial intelligence, the performance of this method will be further improved and perfected.

[0206] In the above embodiment, the early warning notification module includes an early warning logic judgment submodule, an alarm display submodule and a power management submodule; the early warning logic judgment submodule is used to receive the SOC estimation value, the battery health status assessment result and the alarm threshold parameters output by the user behavior learning module, and perform real-time data analysis and logic judgment through an embedded processor; when the alarm is triggered, the alarm display submodule broadcasts the remaining power through a speaker; the power management submodule includes a power control unit and an energy consumption optimization unit; the power control unit controls the switching and distribution of the power supply through a power management integrated circuit to ensure stable power supply to the system; the energy consumption optimization unit reduces the energy consumption of the system in an inactive state through a sleep mode and a wake-up mechanism.

[0207] In a specific embodiment, the early warning notification module realizes dynamic monitoring and timely feedback of the battery status based on real-time data analysis and logical judgment. This module consists of three submodules: an early warning logic judgment submodule, an alarm display submodule, and a power management submodule. The early warning logic judgment submodule uses an embedded processor to process the SOC estimation value, the battery health status assessment results, and the alarm threshold parameters provided by the user behavior learning module in real time, and determines whether the battery is close to a low power state through a preset algorithm. Once it is detected that the power is lower than the safety threshold, the alarm display submodule will be activated and a sound broadcast will be made through the speaker to remind the user to pay attention to the power status. The power management submodule is responsible for the energy supply and optimization of the system, and ensures the energy efficiency and stability of the system in different states through the power control unit and the energy consumption optimization unit.

[0208] Specifically, the early warning logic judgment submodule uses data fusion technology to combine the real-time SOC estimation value, health status and output of the user behavior learning algorithm of the battery to conduct a comprehensive analysis of multiple parameters. This algorithm not only takes into account the immediate state of the battery, but also predicts the short-term behavior trend of the battery, thereby improving the accuracy and foresight of the early warning. The alarm display submodule converts the battery status into natural language through speech synthesis technology and conveys it to the user in the form of a broadcast. This interactive method is not only intuitive, but also can provide effective notifications when the user is distracted or unable to view the screen. The power management submodule adopts advanced power control strategies and energy consumption optimization technologies. The power control unit dynamically adjusts the power supply by monitoring the system load and power status in real time to ensure the stability and reliability of the system under different working conditions. The energy consumption optimization unit uses advanced sleep mode and wake-up mechanism to reduce the energy consumption of the system in standby or low load state and extend the battery life. The three submodules of the early warning notification module work together through efficient data exchange and event triggering mechanism to ensure the system's rapid response and processing of battery status changes. This design not only improves the system's response speed, but also enhances the system's robustness.

[0209] In the specific implementation process, we first need to design and develop the algorithm of the early warning logic judgment submodule to ensure that it can accurately interpret the SOC estimation value and battery health status, and combine it with the output of the user behavior learning module to perform real-time logic judgment. Subsequently, we develop the alarm display submodule, select suitable speaker hardware, and program the sound broadcast function to ensure that the information can be clearly and accurately conveyed to the user when the alarm is triggered. At the same time, we design the power management submodule, select the appropriate power management integrated circuit, and implement the functions of the power control unit and energy consumption optimization unit to optimize the overall energy consumption of the system and ensure a stable supply of power.

[0210] Specifically, the early warning logic judgment submodule needs to analyze the relationship between the SOC estimation value and the battery health status, as well as the impact of user behavior on these parameters, so as to accurately set the alarm threshold. The alarm display submodule needs to analyze the clarity of the sound broadcast and the user's reaction time to the alarm to determine the best alarm strategy. The analysis of the power management submodule focuses on the stability of the power supply and energy consumption optimization, reducing the energy consumption of the system in an inactive state through sleep mode and wake-up mechanism, while ensuring rapid response when needed.

[0211] The design and implementation of the early warning notification module is of great significance for the lithium battery power monitoring and low power warning system. It not only improves the practicality and user-friendliness of the system, but also increases the safety of battery use through real-time monitoring and timely feedback. In addition, through the energy consumption optimization of the power management submodule, the system can extend the battery life and reduce energy waste while ensuring performance. This modular and intelligent design enables the early warning system to adapt to different usage environments and user needs, with high flexibility and scalability, and has a positive role in promoting the development and application of battery management systems.

[0212] In the above embodiment, the system also includes a remote monitoring and data management module, which is used to receive data from each module, perform big data analysis to identify potential problems, and generate maintenance suggestions; the remote monitoring and data management module includes a data access submodule, a data management submodule, a big data analysis submodule, and a maintenance suggestion generation submodule; the data access submodule realizes data exchange with front-end sensors and devices through a data communication protocol, and eliminates noise and ensures the validity of data through outlier detection and missing value filling methods; the data management submodule is used to store pre-processed data and perform data management and query; the data management submodule includes a distributed database unit and a data index unit; the distributed database unit stores historical data through a NoSQL database; the data index unit improves data retrieval speed through an inverted index; the big data analysis submodule is used to perform in-depth analysis on the stored data to identify trends and potential faults in battery performance; the big data analysis submodule predicts the changing trend of battery performance through time series analysis and moving average method, and identifies fault modes through support vector machines; based on the analysis results of the big data analysis submodule, the maintenance suggestion generation submodule formulates a maintenance plan through an expert system and a decision tree algorithm, and automatically generates a maintenance suggestion report through natural language processing and a chart library.

[0213] In a specific embodiment, the remote monitoring and data management module is based on efficient processing, storage, analysis and visualization of real-time data streams to achieve in-depth understanding of battery status and predictive maintenance. The data access submodule adopts advanced data communication protocols to seamlessly exchange data with front-end sensors and devices, and ensures the integrity and accuracy of data through outlier detection and missing value filling technology. The data management submodule uses distributed database and data indexing technology to achieve large-scale data storage, management and rapid retrieval. The big data analysis submodule uses algorithms such as time series analysis, moving average method and support vector machine to conduct in-depth analysis of battery performance and identify trends and potential faults. The maintenance suggestion generation submodule combines expert system and decision tree algorithm to automatically formulate maintenance plans, and uses natural language processing technology to generate easy-to-understand maintenance suggestion reports.

[0214] In the specific implementation process, the data communication protocol of the data access submodule is first used to implement the outlier detection and missing value processing algorithms. Then, the distributed database and data index unit are configured through the data management submodule to support the storage and retrieval of large-scale data. Subsequently, the big data analysis submodule integrates analysis tools such as time series analysis, moving average method and support vector machine to conduct in-depth analysis of battery performance. Finally, the maintenance recommendation generation submodule is developed to integrate expert system and decision tree algorithm, as well as natural language processing and chart library to automatically generate maintenance recommendation reports.

[0215] The implementation of the remote monitoring and data management module not only improves the system's ability to monitor the battery status, but also provides the possibility of predictive maintenance through big data analysis. This modular and intelligent design enables the system to adapt to different usage environments and user needs, with high flexibility and scalability. Through the automatically generated maintenance recommendation report, users can more easily understand the battery status and take maintenance measures in time, thereby extending the battery life and reducing the risk of unexpected failures. In addition, the implementation of this module also helps to reduce maintenance costs and improve energy efficiency, which has a positive role in promoting the development and application of battery management systems.

[0216] Although the specific embodiments of the present invention are described above, it should be understood by those skilled in the art that these specific embodiments are only illustrative, and those skilled in the art may omit, replace, and change the details of the above methods and systems in various ways without departing from the principles and essence of the present invention. For example, merging the above method steps so as to perform substantially the same functions in substantially the same manner to achieve substantially the same results is within the scope of the present invention. Therefore, the scope of the present invention is limited only by the appended claims.

Claims

1. A lithium battery power monitoring and low power warning system, characterized in that: The system comprises: A data acquisition module is used to collect the operating data and environmental data of the battery in real time, and output the data to the SOC estimation and status monitoring module; the data acquisition module collects data through sensors and improves the signal quality through signal conditioning circuits; the operating data and environmental data of the battery include battery voltage, current, temperature and internal resistance data; The SOC estimation and status monitoring module is used to estimate the remaining power of the battery, monitor the battery health status, and output the SOC estimation result and the battery health status assessment report based on the collected data of the data acquisition module; the battery health status includes capacity decay and internal resistance change; the SOC estimation and status monitoring module estimates the SOC through a power depth mapping algorithm combined with the battery chemical characteristics and working environment, and monitors the battery aging trend through data analysis; The user behavior learning module is used to receive the SOC estimation result and battery health status assessment result of the SOC estimation and status monitoring module to analyze the user's usage habits, optimize the power estimation and warning strategy, and output the optimized parameters and thresholds to the SOC estimation and status monitoring module and the warning notification module; the user behavior learning module learns the user behavior pattern through the adaptive user behavior recognition and optimization algorithm, and adjusts the parameters and warning thresholds of the SOC estimation model according to the learning results; the user's usage habits include charging time and discharging mode; the adaptive user behavior recognition and optimization algorithm includes an input layer, a data layer, a feature engineering layer, a user behavior pattern recognition layer, a behavior pattern modeling layer, an adaptive adjustment layer, an output layer and a feedback optimization layer; the working method steps of the adaptive user behavior recognition and optimization algorithm are as follows: U1, obtain SOC estimation data and status monitoring data through the input layer, and the input layer ensures data integrity through data verification protocol CRC check; U2, cleaning, denoising and normalizing the received raw data through the data layer; the data layer processes the data through missing value filling, outlier detection and standardization methods; U3, extracting features from the preprocessed data through the feature engineering layer; the feature engineering layer constructs a feature set through packaging method, embedding method, polynomial features and PCA dimensionality reduction; U4. Based on the extracted features, the user's charging and discharging usage behavior patterns are identified through the user behavior pattern recognition layer; the user behavior pattern recognition layer classifies and clusters the user behaviors through conditional random fields and k-clustering algorithms; U5. Modeling the identified user behavior pattern through the behavior pattern modeling layer, wherein the behavior pattern modeling layer models the user behavior through time series analysis and gradient boosting tree; U6. According to the changes in user behavior patterns, the parameters and warning thresholds of the SOC estimation model are dynamically adjusted through the adaptive adjustment layer; the adaptive adjustment layer automatically optimizes the model parameters and thresholds according to real-time feedback through a policy gradient method; U7, based on the SOC estimation result, battery health status and user behavior pattern information, generates power estimation and early warning strategy through the output layer; the output layer outputs the optimal decision through the Bayesian network and rule engine; U8. Collect user feedback and system operation data through the feedback optimization layer, evaluate the effect of the current strategy, and guide the next round of optimization iteration; the feedback optimization layer collects feedback data through user satisfaction surveys and system log analysis, and evaluates the optimization effect through regression analysis; The early warning notification module is used to perform early warning logic judgment based on the SOC value and battery health status evaluation result of the SOC estimation and status monitoring module, and the alarm threshold parameter of the user behavior learning module, and send an early warning notification to the user through a speaker.

2. A lithium battery power monitoring and low power warning system according to claim 1, characterized in that: The SOC estimation method of the SOC estimation and status monitoring module is: S1, data input, obtaining the operation data and environmental data of the data acquisition module, and removing input data noise and abnormal values ​​through Kalman filtering; S2, initial SOC setting, after system startup or battery replacement, the initial SOC value is set based on the stored data of the battery management system and the estimation result of the open circuit voltage method; S3, multi-parameter fusion estimation of SOC, based on the battery chemical characteristics and working environment, the battery power-SOC mapping model is constructed through the power deep mapping algorithm; the battery chemical characteristics include battery materials and electrolyte properties; the working environment includes temperature, humidity and charge and discharge rate; the power-SOC mapping model fuses the input data information in S1 through a multi-parameter fusion method to perform real-time SOC estimation; S4, ampere-hour integration method assists in estimation and error correction, the ampere-hour integration method assists in estimating the change in SOC, and performs error correction in combination with the output result of S3; the ampere-hour integration method calculates the change in battery power by integrating the charge and discharge current of the battery, and updates the SOC value. The working steps of the ampere-hour integration method are: S401, measuring the charge and discharge current of the battery in real time through a Hall effect sensor, and converting the charge and discharge current of the battery into a digital signal through an analog-to-digital converter; S402, obtaining a change in battery power by integrating the current over time using the Simpson integration method; S403, converting the amount of power change into the amount of SOC change, and updating the current SOC value; S404, combining the result of the power depth mapping algorithm, regularly correcting the estimation result of the ampere-hour integration method to eliminate the accumulated error; S5, open circuit voltage method calibration, after the battery is left to stand for a long time, measure the open circuit voltage of the battery, and find the corresponding SOC value through the OCV-SOC curve, and calibrate the power-SOC mapping model and the estimation result of the ampere-hour integration method; the OCV-SOC curve realizes the construction of the corresponding relationship of the open circuit voltage of the battery cell under different battery remaining power values ​​through data fitting; S6. Output the calibrated SOC value.

3. A lithium battery power monitoring and low power warning system according to claim 2, characterized in that: The multi-parameter fusion method integrates information from different sensors and data sources through a data fusion algorithm, which estimates the state of a linear dynamic system through a Kalman filter to remove noise and outliers in the input data; Based on the complexity of battery chemical characteristics and working environment, the multi-parameter fusion method also constructs a power-SOC mapping model through a multi-parameter mapping mechanism to achieve accurate estimation of SOC; the multi-parameter mapping mechanism captures the complex nonlinear relationship between battery power and SOC through a polynomial regression method, and performs SOC estimation based on multi-parameter fusion processing; During the operation of the power-SOC mapping model, the multi-parameter mapping mechanism also dynamically adjusts the model parameters according to the changes in the current battery state and working conditions through a real-time parameter adjustment method.

4. A lithium battery power monitoring and low power warning system according to claim 1, characterized in that: The working steps of the power depth mapping algorithm include: Step 1: System modeling and prediction; Based on the physical characteristics and working principle of the battery, a dynamic model of the model battery system is constructed to predict the future value of the battery state. The dynamic model is used to describe the linear and nonlinear relationship between the battery state and the change of external input over time; the external input includes current and voltage values; during the prediction process, the dynamic model quantifies the uncertainty of the model through the covariance matrix; the formula expression of the dynamic model is: { （1） In formula (1), represents a state transfer matrix, which is used to reflect the change of battery state over time; the state transfer matrix Depends on the battery's temperature coefficient and aging factor ; is the predicted state vector, which is used to represent the predicted value of the current state of the battery, including SOC, voltage and current integration parameters; is the input matrix, which is used to reflect the current efficiency The actual effect that affects the input current; Represents a nonlinear perturbation function, which is used to measure the complexity of the chemical reaction inside the battery; represents process noise, which is used to measure model uncertainty; is the input vector, including current and voltage; Step 2: Linearize the nonlinear function; In the prediction step of the dynamic model, the nonlinear system model is linearized at the prediction point by the Jacobian matrix to allow the update rule of the power depth mapping algorithm to be applied; the formula expression for linearization using the Jacobian matrix is: （2） In formula (2), Represents the Jacobian matrix, which is used to linearize nonlinear models; Represents the capacity decay coefficient, which is used to measure the rate of change of battery capacity over time; Indicates the internal resistance variation coefficient, which is used to measure the rate of change of the battery's internal resistance over time; Represents the temperature influence coefficient, which is used to measure the impact of temperature changes on battery performance; Step 3: Setting the observation equation; At the same time, the real-time observation data of the battery is obtained through the sensor, and the relationship between the observation data and the battery state is established through the observation equation to reduce the influence of noise and interference; the real-time observation data of the battery includes the battery voltage and current; the formula expression of the observation equation is: （3） In formula (3), represents an observation vector, used to correct the predicted state, wherein the observation vector includes battery voltage and current; represents observation parameters, which are used to adjust the observation equation, and the observation parameters include battery temperature and battery age; represents the observation function, which is used to measure the change factors of the open circuit voltage and internal resistance of the battery; represents observation noise, which is used to measure the uncertainty in the observations; An additional parameter representing the observation noise, used to measure the distribution of the noise; Step 4: Update the prediction error covariance; Next, the difference between the observed value and the model predicted value is calculated by the prediction error covariance update function; the formula expression of the prediction error covariance update function is: （4） In formula (4), represents the forecast error covariance matrix, which is used to quantify the uncertainty of the forecast state; Represents the updated error covariance matrix of the previous moment; Represents the process noise covariance matrix, quantifying the statistical characteristics of the process noise; represents the process noise intensity coefficient; represents the process noise correlation coefficient; Step 5: Kalman gain calculation and state update; The weights of the predicted value and the observed value when updating the battery state estimation are balanced by the Kalman gain; the Kalman gain quantifies the uncertainty of the prediction and observation respectively through the prediction error covariance matrix and the observation noise covariance matrix, and minimizes the covariance of the estimation error; wherein the Kalman gain calculation and state update include the following steps: Step 501: Calculate the contribution of the observation value in updating the state estimate through the Kalman gain, and the formula expression is: （5） In formula (5), is the Kalman gain matrix, which is used to determine the influence of the observation value on the state estimation; Represents the observation function Status The Jacobian matrix of is used to map the state vector to the observation space; represents the observation noise covariance matrix, which is used to measure the noise level in the observations; represents the observation noise intensity coefficient; represents the observation noise correlation coefficient; Step 502: After obtaining a new observation value, update the state estimate through a state update function to minimize the prediction error; the formula expression of the state update function is: （6） In formula (6), is the updated state estimate; represents the observation vector value; represents the observation noise intensity coefficient; represents the forecast state adjustment; represents the Kalman gain matrix; represents the prediction error covariance matrix; Step 503: After the state is updated, the error covariance matrix is ​​updated to reflect the uncertainty of the new estimate. The formula is: （7） In formula (7), represents the updated error covariance matrix, which is used to reflect the uncertainty of the updated state estimation; is the identity matrix; Step 6: Deep learning correction; A deep learning model is trained based on at least historical voltage, current and temperature data to build a power-SOC mapping model, predict the SOC correction amount, and output a probability distribution including a mean and a variance; Step 7: Fusion and output; The SOC correction value predicted by the power-SOC mapping model is fused with the estimated value of the extended Kalman filter by a weighted fusion method to generate a final SOC estimated value; the fusion formula of the weighted fusion method is expressed as: （8） In formula (8), is the estimated SOC value after deep learning correction; It is the SOC estimation value after the extended Kalman filter is updated; The SOC correction value predicted by the power-SOC mapping model; is the normal distribution function, which is used to measure the uncertainty of deep learning predictions; It is the mean value of the correction value predicted by the power-SOC mapping model, which is used to adjust the SOC estimation value; is the weighting coefficient used to balance the contribution of the deep learning model and the extended Kalman filter; Represents the uncertainty variance of deep learning predictions, which is used to measure the reliability of the predictions.

5. A lithium battery power monitoring and low power warning system according to claim 4, characterized in that: The power-SOC mapping model includes an input layer, a feature extraction layer, a hidden layer, an uncertainty estimation layer and an output layer; the working principle steps of the power-SOC mapping model are as follows: W1, receiving pre-processed battery operation data through the input layer; W2, extracting features from input data through the feature extraction layer; The feature extraction layer extracts the spatial features of the time series data through a convolutional neural network, and captures the temporal features of the time series data through a recurrent neural network; the spatial features of the time series data are the fluctuation features of the battery voltage; The time series feature of the time series data is the changing trend of the battery current; W3, performing feature conversion and nonlinear mapping through the hidden layer; The hidden layer performs feature conversion through multiple fully connected layers to increase the expressiveness of the model, and introduces nonlinear mapping through a nonlinear activation function to capture the complex relationship between the battery SOC and the input data; W4, estimating the uncertainty of the model prediction through the uncertainty estimation layer; The uncertainty estimation layer prevents overfitting and improves the generalization ability of the model through a regularization method; and estimates the uncertainty of the model prediction through a Bayesian neural network, and the output of the uncertainty estimation layer includes a probability distribution of a mean and a variance; W5. Output the final SOC estimation value and the uncertainty of the SOC estimation value through the output layer; the output layer represents the SOC estimation value and the uncertainty of the SOC estimation value through Gaussian distribution.

6. A lithium battery power monitoring and low power warning system according to claim 1, characterized in that: The state monitoring working method of the SOC estimation and state monitoring module is: R1. Periodically collect the battery capacity, internal resistance, charge and discharge performance, temperature and cycle number parameter data through the battery management system; R2. Monitoring the capacity decay of the battery by a model-based capacity estimation method; the model-based capacity estimation method predicts the battery capacity decay trend by exponential smoothing method; R3. Based on AC impedance and DC internal resistance data, the changing trend of internal resistance is analyzed by moving average and standard deviation; R4. The charging and discharging process of the battery is controlled by the BMS controller, and the charging and discharging efficiency is automatically calculated and continuously updated by the efficiency formula; the BMS controller controls the output of the charger and the discharger by the PWM signal; the formula expression of the efficiency formula is: （9） In formula (9), represents the discharge energy, Indicates charging energy; R5. Based on the above historical cycle times, temperature, charge and discharge efficiency characteristics and corresponding battery capacity data, the battery health model is trained through support vector machine to predict battery life; R6. Combine the monitoring data from R2 to R5 and generate a battery health report through a template engine. The battery health report includes the comprehensive health status, performance level and remaining life prediction information of the battery.

7. A lithium battery power monitoring and low power warning system according to claim 1, characterized in that: The warning notification module includes a warning logic judgment submodule, an alarm display submodule and a power management submodule; The early warning logic judgment submodule is used to receive the SOC estimation value, the battery health status assessment result and the alarm threshold parameters output by the user behavior learning module, and perform real-time data analysis and logic judgment through the embedded processor; when the alarm is triggered, the alarm display submodule broadcasts the remaining power through the speaker; the power management submodule includes a power control unit and an energy consumption optimization unit; the power control unit controls the switching and distribution of the power supply through the power management integrated circuit to ensure stable power supply of the system; the energy consumption optimization unit reduces the energy consumption of the system in an inactive state through the sleep mode and wake-up mechanism.

8. A lithium battery power monitoring and low power warning system according to claim 1, characterized in that: The system also includes a remote monitoring and data management module, which is used to receive data from each module, perform big data analysis to identify potential problems, and generate maintenance suggestions; the remote monitoring and data management module includes a data access submodule, a data management submodule, a big data analysis submodule, and a maintenance suggestion generation submodule; the data access submodule realizes data exchange with front-end sensors and devices through a data communication protocol, and eliminates noise and ensures data validity through outlier detection and missing value filling methods; the data management submodule is used to store pre-processed data and perform data management and query; the data management submodule includes a distributed database unit and a data index unit; the distributed database unit stores historical data through a NoSQL database; the data index unit improves data retrieval speed through an inverted index; the big data analysis submodule is used to perform in-depth analysis on the stored data to identify trends and potential faults in battery performance; the big data analysis submodule predicts the changing trend of battery performance through time series analysis and moving average method, and identifies fault modes through a support vector machine; Based on the analysis results of the big data analysis submodule, the maintenance suggestion generation submodule formulates a maintenance plan through an expert system and a decision tree algorithm, and automatically generates a maintenance suggestion report through natural language processing and a chart library.

Images