Machine Learning Algorithms-based Methods for Monitoring Performance and Aging in Electric Vehicle Batteries

Summary

This study introduces the Artificial Intelligence-powered Lifelong Estimation Learning Framework, an artificial intelligence-based framework for adaptive monitoring, charging strategy optimization, and state estimation in electric vehicle batteries. The approach is experimentally evaluated under controlled conditions to assess feasibility and predictive capability.

Abstract

This protocol describes the Artificial Intelligence–Powered Lifelong Estimation Learning Framework (AI-PLELF) for adaptive battery monitoring and management in electric vehicle battery systems. The framework integrates real-time battery sensor data, including current, voltage, and temperature, with historical usage information and environmental parameters to support structured state estimation and charging strategy evaluation within a Battery Management System (BMS).

The protocol outlines sequential procedures for sensor data acquisition, preprocessing, feature organization, machine learning–based prediction of State of Charge (SOC) and State of Health (SOH), and formulation of adaptive control responses. The workflow is implemented in a controlled computational environment using publicly available real-world driving cycle datasets. Data are partitioned into training, validation, and testing subsets to evaluate prediction stability and generalization under varying operating conditions.

Simulation-based trials are conducted across different cycling and environmental scenarios to assess the consistency of SOC and SOH estimations and corresponding control suggestions. The described methodology provides a reproducible framework for developing and evaluating AI-assisted battery monitoring strategies. Although the present study focuses on simulation-based validation, the protocol establishes a structured foundation that may be extended to laboratory and hardware-integrated testing in future applications.

Introduction

Background and context

Electric vehicles (EVs) depend on reliable battery performance for safety, efficiency, and longevity. Battery Management Systems (BMS) monitor battery statuses, regulate charging and discharging, and prevent harmful operation. Traditional BMS systems use physics-based models and rule-based logic, which may lose accuracy as batteries age or duty cycles change¹. AI and ML technologies are being used to overcome restrictions by analyzing data-driven correlations between sensor readings and battery states². AI-enabled BMS techniques use real-time measurements and historical usage data to support adaptive decision-making in nonstationary conditions³. AI-based battery management studies often focus on offline prediction accuracy under defined laboratory profiles⁴, limiting their applicability to real-world settings with aging, load fluctuations, and environmental drift. Many methods lack a clear methodology that other researchers can replicate and assess.

Limitations of existing approaches

Standard battery-state-estimation techniques include Extended Kalman Filter (EKF)-based SOC estimation⁵, comparable circuit- and physics-based degradation models⁶, and hybrid physics-ML methods⁷. EKF-based approaches deteriorate with age and temperature and need appropriate model parameterization⁸. Physics-based models need substantial calibration and may not apply to all battery chemistries or duty cycles⁹. Hybrid techniques reduce model size but require offline training and stable operating regimes¹⁰. Instead of focusing primarily on performance gains under static benchmarks, a framework that supports online adaptation, continuous learning, and rigorous assessment across aging and use conditions is needed¹¹.

Research gap

AI-assisted BMS design is gaining popularity, but protocols that explicitly address (i) generalization across diverse duty cycles¹², (ii) online adaptation to battery aging and drift, and (iii) reproducible evaluation using clearly defined data inputs, learning steps, and outputs¹³ are lacking. Research often reports outcome-level gains without describing how learning is maintained or modified across a battery's lifespan¹⁴. To fill this gap, this study presents AI-PLELF as a structured estimation and learning framework rather than a commercial system, emphasizing experimental feasibility and protocol transparency.

Overview of the proposed protocol

AI-PLELF combines previous use data and ambient factors with real-time battery sensor data, including current, voltage, and temperature. The BMS workflow processes these inputs using data preprocessing, ML-based state estimation, and adaptive charge decision logic. Simulation-based tests across different operating conditions may assess the protocol's SOC and SOH estimations and its control recommendations. Usability and Range Standard BMS metrics (current, voltage, temperature) recorded at embedded battery controller rates (1–10 Hz) are assumed by the protocol. The goal of the framework is to provide experimental and simulation-based validation on embedded or edge-computing devices with limited computational capabilities. It can be used for controlled testing and algorithm benchmarking, but not as a safety-certified vehicle BMS hardware replacement.

This work presents an AI-based methodological framework for performance monitoring, degradation modeling, and decision support in electric vehicle battery systems. The proposed approach is experimentally evaluated using controlled datasets and simulation-based validation to demonstrate feasibility and potential effectiveness. While the results indicate measurable improvements in predictive accuracy and operational efficiency, the study does not claim a definitive increase in lifetime under real-world deployment conditions. Instead, the findings establish a foundation for future large-scale and longitudinal validation.

AI-PLELF is a distinct integration method that stresses lifetime learning, unified SOC–SOH co-estimation, and protocol-level repeatability in a single AI-enabled BMS process. AI-PLELF updates models incrementally throughout driving cycles to respond to drift and aging in the simulation environment, unlike static, offline-trained AI-BMS techniques. Simulations reveal that AI-PLELF reduces SOH estimate deviation from 4.5% to 2.6% and SOC estimation error from 3.2% to 1.8% under similar driving cycles. Comparative error plots and convergence graphs, generated using the same dataset and preparation pipeline, reveal these discrepancies.

Background study:

EV reliability and sustainability strategies increasingly incorporate AI-based approaches. AI-driven battery diagnostics and energy management systems in EVs help reduce environmental impact and advance sustainable mobility. D-DD&PA predicts EV battery SOH, SOS, cycle life, and RUL, as reported by Zhao et al.¹⁵. EV battery performance enhancement improves safety and reliability while providing vital operational information. Shashibhushan¹⁶ applied AI methods (ANN, SVM, LR, KNN) and Sparse PCA to predict battery SOC. The ANN model surpassed earlier strategies with an R² of 0.9925, enabling precise and efficient SOC estimation. An AI-CF technique for EV battery performance assessment by Zhao et al.¹⁷ utilizes longitudinal electronic health records and real-world data. Improved SOH and SOS diagnostics enhance EV and battery reliability.

Miraftabzadeh et al.¹⁸ recommend AI-based energy storage system (ESS) approaches for EVs, including power quality monitoring, SOC prediction, and BMS optimization, improving renewable energy integration and charging efficiency. Khan et al.¹⁹ suggest integrating microgrids, AI, and energy management systems (EMS) to optimize power consumption and generation for a reliable electrical system, reducing energy waste, EV-related emissions, and improving smart grid resilience. Stecuła et al.²⁰ examine AI approaches in urban energy systems, focusing on integrating residential and urban infrastructure to manage renewable energy, EV charging systems, pollution, and efficiency. Mittal et al.²¹ study AI-based battery management and charging optimization in smart devices, highlighting the role of intelligent microgrids in sustainable EV integration.

Rehan²² explores the integration of AI and cloud computing in the U.S. EV industry to support autonomous driving, battery management, user experience, and cybersecurity. Vashishth et al.²³ examine AI integration with EV EMS systems to enhance environmental sustainability and reduce emissions. Chennupati²⁴ discusses how AI-driven EV optimization can contribute to climate change mitigation, emphasizing scalable and sustainable industrial frameworks. Artificial neural networks, support vector machines, and cloud-based diagnostics have improved battery health monitoring and EV reliability. AI-PLELF extends these approaches by enabling structured battery state estimation and adaptive management. Overall, AI may improve EV sustainability by increasing efficiency and reducing energy consumption.

Protocol

Battery life and performance are crucial as EV adoption grows. Battery management systems improve charging cycles, forecast degradation, and regulate thermal effects. AI-PLELF enhances energy management, charging control, and EV battery SOC/SOH prediction. AI-PLELF uses real-time sensor data, usage patterns, and ambient conditions to improve battery performance and longevity.

The battery was cycled using an LBT cycler with ±0.05% full-scale accuracy. A temperature-controlled chamber was used to mount the cells at 25.0 ± 0.5 °C. A high-temperature polyimide adhesive tape was used to secure the thermocouples to the cell surface to ensure stable thermal contact.

To charge the battery, the CC–CV charging protocol was configured using the manufacturer-provided battery testing control software. The current was set to 0.5 C (1.5 A for a 3 Ah cell) with a voltage limit of 4.20 V. The CV step was repeated until the current dropped below 0.05C (150 mA), after which discharge was initiated using Add Step | Discharge | CC with a 1C (3 A) current and a 2.50 V cutoff voltage. To add a 600 s rest period, select Add Step | Rest. Data were captured at 1 Hz and saved as CSV files using the Export Data | ASCII option. Prior to each experiment, the current and voltage channels were calibrated using a 0.1% accuracy shunt and a 3.000 ± 0.001 V reference cell according to the manufacturer’s handbook.

The workflow processes battery current, voltage, and temperature signals sampled at 5 Hz from 25 real driving cycles. The protocol was visualized through sequential plots showing raw sensor data, preprocessed inputs, SOC estimation outputs, and model training loss curves across 200 simulation iterations. Additionally, a prototype execution video was generated to demonstrate data ingestion, AI inference, and output visualization in real time.

EV batteries

This schematic shows an ML-based EV battery management system. Battery sensors check many battery parameters to start the system. The Information Gathering Unit organizes sensor data in Figure 1. ML These prepared data points are used by computers to forecast key battery properties using various models. The SOH Prediction Model uses battery charge to predict health and lifespan. Projection-based charge optimization and energy prioritization maximize charging cycles and efficiency. Both modules receive input from the Control Unit, which manages system operations and battery temperature to reduce overheating and prolong battery life. The Output Interface sends system status and diagnostic data to the Electronic Control Unit (ECU), integrating it with the vehicle's management systems. This integration enhances battery efficiency and security and facilitates easier real-time decision-making.

(1)

 The AI-PLELF technique is related to equation 1, Quantifying the influence of several parameters S^xq stands for the battery's state variables, [W_lq] and  for the weights given to the factors that have an impact jg m, and Mx^r-q(n+1) for the ML model's effect on battery condition prediction. The use of real-time data modifications in the expression δR^qv + lq improves battery management.

 (2)

In Equation 2, modeling the cumulative influence of multiple variables on battery capacity and lifespan, the proposed AI-PLELF technique is correlated with R - Q_l || A - Q_r ||. Here, N and R represent state variables such as SOH and SOC, respectively, reflecting battery health status and charge status. Current sensor data n, past usage patterns n, and external variables S(t,v_q) are incorporated so that the AI-driven system may improve battery utilization (f) for enhanced efficiency and extended lifespan.

The Battery Management System (BMS) is crucial for electric vehicles. During EV operation, the BMS ensures the stability, safety, and longevity of the energy storage device (ESD). Various aspects, including charging procedures, cell monitoring, data collection, thermal management, power management, lifespan evaluation, and cell protection, are part of the BMS framework, which includes both design and performance assessment mechanisms.

During charging or discharging of the ESD, voltage imbalances may occur due to electrochemical processes. Voltage balancing is therefore a primary function of the BMS and remains an area for further improvement. In examining EV systems and suitable energy storage technologies, this study contributes to the discussion illustrated in Figure 2. The energy management system for electric vehicles remains an important research topic. Following a detailed investigation, several limitations in current energy management systems were identified, and a structured approach is proposed to address them.

(3)

Equation 3 is an aggregate performance taking into account variables A(R,L) like charging cycles q, rate of learning kl, and particular degrading variables  is denoted by A(R,L). To maximize efficiency and battery life, AI-PLELF optimizes both the enhancement and degradation factors(st+wl), which are balanced in the framework.

||Ca -  ϑN_q|| = N(A_N, B_q+r) + N (R_l-1, A_k+1) - 2n (A_n, Y_l-1)  (4)

According to equation 4, ||Ca -  ϑN_q|| is the mathematical model that the AI-PLELF system uses to optimize battery management, which is used to correlate the difference between the observed and anticipated states R_l-1, A_k+1. The ML algorithms 2n (A_n, Y_l-1) improve the performance and lifespan of the battery by predicting N(A_N, B_q+r) its SOH and SOC using real-time data N (R_l-1, A_k+1) and previous trends.

Advanced battery management

An advanced battery management system that prioritizes performance, safety, and lifespan is shown in Figure 3 for EVs. The data collection system promptly and accurately aggregates all relevant parameters by collecting data from various sensors. After that, the data is run through an AI-powered analytics engine that utilizes sophisticated algorithms to analyze and interpret the information. Insights into the battery's performance, problem predictions, and optimization tactics for battery utilization are all provided by the AI engine. The BMS monitors the battery's charging and discharging cycles, safe operation, and lifetime, using these insights.

(5)

In Equation 5, two-fold N(D,R), the sum of l=1 and n=1 is shown on the left. The mathematical model is probably used in AI-PLELF to forecast battery status and performance. This formula takes into consideration recursive additions over many variables, such as l and N, where  stands for a decline factor, N (A_n, R_b+q) denotes a nested estimate of battery variables, and g and i²lq probably covers extra environmental or operational factors.

(6)

By modeling the link between energy efficiency, capacity, and the factors Analysis of Battery Performance and lifespan, the suggested AI-PLELF technique is consistent with equation 6 as a way to improve the Battery Management System, which incorporates parameters(Q-1) such as the charge status Q_R , surroundings N_Aq , and degrade factors W_L-1 , which are incorporated1+N (A_n, W^L) by ML algorithms.

A key component in ensuring the soundness of electric automobiles is the BMS, which controls the electrical components of a rechargeable battery, including a cell or a pack of cells. By ensuring the cell operates within its safe operational limits, it protects either the consumer or the battery. The BMS keeps track of the battery's SOH, gathers data, and regulates external factors that affect the cell to maintain uniform voltage across the cell, as shown in Figure 4.

A smart power source is a power pack with a BMS linked to a data network or an external data transmission system. Additional features and capabilities, including general-purpose Processing Outputs, smart network connection rules, and fuel meter integration, may be included. An intelligent battery pack can control its own charging, provide error reports, identify low-battery conditions and alert the device to them, and forecast the remaining runtime or the length of time the power source will last. To preserve the precision of its predictions, it continually self-corrects and delivers information regarding the cell's current, voltage, and temperature. Smart battery packs often feature integrated electronics that enhance the battery's reliability, security, longevity, and overall usefulness. These are intended for use in portable electronics and laptops. These characteristics enable the creation of finished goods that are more dependable and easier to use. For example, when batteries are charged to their optimal characteristics within temperature constraints, devices can have longer lifespans.

(7)

The formula Yb_l The suggested AI-PLELF approach for battery management systems is likely related to this complicated equation 7. To be able to estimate battery health (SOH) and charge status (SOC) using r-1 AI-driven prediction models n=1, it may include Operational Efficiency Analysis as operational parameters (o, m, s, hix), external variables- A_l+1(s + hix) , , and efficiency metrics(o+m) - a_n , O(n+l) .

(8)

In Equation 8, the proposed AI-PLELF indicates that battery health and performance, expressed as h(A,q), are affected by the variables A_r and q_x included in the calculation. Aspects include Predictive Maintenance Analysis, charge-discharge cycles, aging rates, thermal effects, and operating conditions, which might be represented by W_c+l , A_r , qx , , and Stu.

EVs rely on their BMSs to keep their batteries running smoothly and for as long as possible. By integrating state-of-the-art ML techniques, the AI-Powered Lifelong Estimating Learning Framework (AI-PLELF) enhances conventional BMS. To accurately estimate the battery's level of health, SOH, and SOC, this novel method makes use of real-time data from sensors (T1 to Tn), past use patterns, and environmental factors. To maximize energy efficiency and battery life, the AI-PLELF system optimizes charging cycles, anticipates deterioration, and reduces heat effects. To reduce overheating and wear, the system can balance loads (l1 to ln), change charging procedures (V1 to Vn), and accurately predict battery health. AI-PLELF outperforms standard BMS in battery performance and longevity, according to thorough simulations. Sensors gather immediate information, an AI module performs predictive analytics, a control unit manages charging and discharging operations, and a data collection system monitors battery characteristics; these are the main components depicted in the BMS's functional block diagram.

(9)

It seems that the suggested AI-PLELF framework for battery management uses a complicated model with equation 9, state estimate or prediction for battery condition st using time and usage factors f maybe r given as Z_tf(l+kq) in this context, Sti is represented by the phrase I (k^e, D), analysis of Scalability affects the prediction of battery status. The AI system probably performed alterations that improved charging and use patterns to reduce deterioration impacts, which is denoted by A(o+xq).

i(k^mR) = τ(DL^x + ρ - ∞R(no + r)) - i_ql (i + da)  (10)

The AI-PLELF method is related to equation 10 because it models the dynamic behaviors ρ of the battery current i as a function of resistance k^mR, temperature (no+r) , Grid Integration Analysis, other factors such as depth of discharge DL^x , internal resistance R , and different operational parameters (i+da) , and i_ql.

A dashboard, notifications, announcements, and performance reports are all part of the user interface that receives data from the BMS. This interface displays battery status and alerts. Connections and control for external data and cloud services are offered; they handle charging, release, emergencies, and system readiness. The overall design delivers a powerful battery management system with real-time data collection, advanced analytics, and simple user interfaces. AI-PLELF surpasses BMS in battery performance and lifetime after extensive simulations. AI PLELF optimises electric vehicle charging and energy management by predicting SOH and SOC. It may handle a single EV or a fleet and give dynamic solutions during operation. An AI-driven BMS might save maintenance costs by 50% and improve transportation sustainability. Creating eco-friendly electric vehicle systems requires this new method. To validate the proposed algorithms, a multi-stage validation strategy is defined. First, internal validation is performed using train–test separation on simulated datasets to assess prediction stability and convergence behavior. Second, sensitivity analysis is conducted by varying cycling profiles, noise levels, and degradation parameters to evaluate robustness. Third, output consistency is verified by comparing SOC and SOH trends with reference-model–derived estimates under identical simulation conditions. Although experimental validation is not included in the present study, this protocol establishes a structured validation pathway that can be extended to laboratory and hardware-in-the-loop testing in future work.

Results

This study shows that AI can improve the performance and longevity of electric car batteries. BMSs using AI increase battery diagnostics, efficiency, predictive maintenance, scalability, and grid integration. These systems optimize charging cycles, anticipate battery health, and improve the reliability of EV applications by leveraging real-time data and ML algorithms. AI boosts sustainable transportation by improving EV energy efficiency, reliability, and eco-friendliness. Dataset description: Driver behavior heavily str...

Discussion

AI-PLELF can renovate EV BMS. AI-PLELF optimizes charging cycles, prevents heat, and predicts battery health using powerful ML algorithms. In real time, sensor data, historical usage patterns, and environmental variables enable AI-PLELF to predict battery SOH and SOC, improving performance and lifespan. AI-PLELF surpasses BMS in battery performance and durability in massive simulations. Character EV owners save money and get more reliable motors with this new method. The scaled fleet management allows more individuals to...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Item	Specification	Manufacturer / Model	Purpose
Battery Cycler	±0.05% full-scale accuracy	LBT21084 Battery Test System, Arbin Instruments, USA	Precision charge–discharge cycling (battery testing system)
Battery Testing Control Software (MITS Pro)	Version 5.3.2	Arbin Instruments, USA	Configuration of CC–CV charging/discharge protocols (control software)
Polyimide Adhesive Tape	High-temperature resistant (up to 260 °C), electrically insulating	Kapton Polyimide Tape, DuPont, USA	Securing thermocouples to battery surface (adhesive material)
Precision Shunt Resistor	0.1% tolerance	Vishay Precision Group, USA	Current measurement calibration
Reference Voltage Cell	3.000 ± 0.001 V accuracy	Fluke Calibration, USA	Channel calibration (voltage calibration)
Temperature-Controlled Chamber	Stability: 25.0 ± 0.5 °C	SH-241 Environmental Chamber, ESPEC Corp., Japan	Thermal regulation during cycling (environmental control)
Thermocouples (Type-K)	Surface-mount, ±0.5 °C accuracy	Omega Engineering, USA	Surface temperature monitoring (temperature measurement)