Jeong et al. in their work [33] introduce a BMS algorithm that can identify and calculate battery voltage, current, and temperature. A method for SOC estimate that integrates open-circuit voltage (OCV) with the Coulomb counting method (CCM) is presented to address the limitations of both OCV and CCM. This estimate approach necessitates minimal processing resources and can enhance estimation precision. The suggested algorithm employs the OCV equation, incorporating the internal resistance and efficiency of the battery. The equation may compute the battery’s charge–discharge by precisely accounting for the initial value of the CCM by the application of OCV, while taking into consideration the battery’s state. Utilizing the battery efficiency formula, a predictive model for the SOH of a battery, which is contingent upon the charging duration necessary for safe operation, is integrated into the BMS algorithm to enhance reliability. During calculating a battery’s efficiency, the charge–discharge current is measured to establish whether the ESS is in the charge–discharge condition. Based on the findings, the suggested method is capable of improving not only BMS performance but also battery safety by utilizing a problem diagnosis algorithm with precise SOH calculation. The study does not comprehensively compare with other parameter BMS algorithms nor evaluate its real-time efficacy.

Manas et al. [34] presented a centralized BMS for an evolving EV transportation system. Through simulation, the suggested BMS architecture and test findings are verified. The proposed BMS can control battery charge level, prevent overcharging and discharging, and maintain temperature protection, according to benchmarked testing results for the voltage, current, and temperature sensors. The benchmarked hardware and simulation results show that the suggested congregated BMS architecture can control temperature, avoid excessive charging or discharging, and maintain stability among the battery cells within specific battery parts. It lacks a comprehensive assessment of its abilities and adaptability in real-time applications.

Khalfi et al. [35] developed a model for a lithium-ion battery cell that takes consideration of the dynamic behavior of the battery cell in different settings and driving styles of electric vehicles on the highway, rural and urban roads. The model is composed of a second-order Thevenin model with six parameters. This work used the Trust-Region-Reflective and Levenberg–Marquardt algorithms in conjunction with a nonlinear least squares approach to predict the parameters of battery cells. As a measure of model performance, the mean square error (MSE) was used to validate the model; for the UDDS drive cycle, it is 3.8260 × 10⁴, whereas for the LA-92 drive cycle, it is 6.2674 × 10⁴. The Levenberg–Marquardt algorithm model is the best option because it takes fewer iterations than the other model method. This research is deficient in a thorough investigation of the model’s performance during extreme or prolonged usage situations, such as high-stress driving or battery degradation.

Nivedita et al. [36] examined two distinct strategies for cooling cells, namely air cooling and direct liquid cooling. The researchers proceeded to compare the outcomes of these procedures with the static cell temperature. In order to assess their efficacy, the effectiveness of these methods was evaluated using a conventional large-capacity lithium-ion pouch cell specifically designed for electric drive vehicles. This evaluation was conducted from the standpoint of coolant parasitic power consumption, maximum temperature increase, temperature differential within the cell, and the supplementary weight required for the cooling system. The findings indicated that an air-cooling system exhibited higher energy use in order to maintain an equivalent average temperature. The direct liquid cooling system has the most minimal increase in maximum temperature. However, it does not address cooling system reliability and maintainability, particularly coolant degradation and system damage in real-world electric vehicles.

Kumar et al. [37] used SIMULINK to develop a cost-effective BMS that constructs and simulates the circuit, ensuring optimal efficiency and security of the battery. The circuit is designed to monitor the battery’s essential characteristics over a period of time, hence enhancing battery longevity and facilitating more efficient battery utilization. According to the findings, it is possible to significantly reduce greenhouse gas emissions. However, it lacks a comprehensive examination of the system’s performance in real-world circumstances, specifically regarding its efficacy under diverse operating conditions.

Yan Ma et al. [38] suggested a hierarchical optimization technique for BTMS based on ant colony optimization-fuzzy sliding mode control (ACO-FSMC) to maintain the battery pack temperature at the target temperature and guarantee the cruising range of EVs. The technique for hierarchical optimization is intended to achieve control over the battery cooling rate in addition to the pump and compressor speeds. According to the simulation findings, this strategy’s maximum temperature deviation and energy usage were, respectively, 43.2% and 23.0% lower than fuzzy PID. Nevertheless, the computational difficulty and feasibility of implementing the ACO-FSMC approach in real-time applications were not considered.

Hamednia et al. [39], presented a proposal for enhancing the grid-to-meter energy utilization of a battery electric vehicle through optimal battery thermal management (BTM), charging, and eco-driving techniques. An optimization problem was formulated with the objective of finding the optimal trade-off between trip time and charging cost. The problem that has been formulated was subsequently converted into a hybrid dynamical system. In this system, the dynamics in the driving and charging modes were represented using distinct functions and involve separate state and control vectors. The suggested algorithm’s performance was evaluated on a hilly road, with two charging options considered along the driving path. Trip duration, including driving and charging times, decreased by 44% when compared to a situation without active battery heating or cooling. Nevertheless, it did not evaluate its scalability and adaptability over diverse real-world scenarios, encompassing varying road conditions, extreme weather, and numerous vehicle models.

Zhu et al. [40] introduced current integration method with open-circuit voltage method to calculate the average SOC. In this study, a BMS experimental platform was built that supports three serially coupled 3400 mAh lithium cobalt oxide batteries. It features precision in voltage and current measurement, SOC calculation, balancing control, and Liquid Crystal Display (LCD). The experimental results demonstrated that the cumulative error after 80 min charging by employing the combined technique is 1.1%, which may be further reduced to 0.6% with effective open-circuit voltage correction. Additionally, passive equilibrium control (PEC) was practical due to the simplicity and inexpensive cost. The results further demonstrated that PEC can serve successfully in strengthening the consistency of a battery pack over time. The discussion lacks the scalability of the methodology to larger battery packs and various operating and environmental circumstances, which are crucial for broader applicability in real-world scenarios.

Fault diagnostic algorithms are necessary for BMS. The primary function of these algorithms is to identify defects at an early stage and promptly implement suitable control measures for both the battery and the users. While operating a lithium-ion power battery, there is a risk of fire, explosion, smoke, high-voltage electric shock, and other risks. These hazards can occur due to numerous factors, including overcharge, overcurrent, and excessive temperature in the cell. These conditions can result in the formation of lithium dendrites or excessive heat generation. This can lead to the puncturing or decomposition of the solid electrolyte interphase (SEI) film, triggering a chain of negative reactions, ultimately resulting in a short circuit in the cell [41]. Xu et al. [42] developed a robust multi-objective nonlinear fault detection observer for lithium-ion batteries that is able to handle disturbances yet is very responsive to multiple faults in the battery. A comprehensive three-step multi-fault detection technique was devised to identify many types of battery defects, such as short-circuit faults, as well as faults in current and voltage sensors, using adaptive thresholding. The effectiveness of the suggested approach was further confirmed through multiple experimental case studies, which included various faults with different levels of severity and incorrect state of charge initialization. This research lacked real-time performance analysis at various driving cycle, which is crucial for dynamic BMS.

Si et al. [43] presented a work based on estimator algorithm the extended Kalman filter-amp-time integration-open-circuit voltage method (EKF-Ah-OCV), which improves the amp-time integration method and open-circuit voltage method primarily by utilizing the correction properties of the Kalman filter algorithm. The accuracy of the voltage method not only fixed the problem of the ampere-time integration method’s incorrect SOC initial value estimate, but it also fixed the issue of SOC estimation mistakes that have built up over time because of inaccurate long-term current measurement. The algorithm demonstrated great performance in the complex environment as assessed by SOC and was capable of fulfilling the power demands of lithium-ion batteries. The study does not address the computationally intensive or scalable aspects, which may limit its applicability for extensive, real-time applications in extensive BMS.

In addition to these models, filters can be incorporated to eliminate model uncertainties and attain satisfactory estimation performance, such as the adaptive unscented Kalman filter (AUKF) [44] and fuzzy robust two-stage unscented Kalman filter (FRTSUKF) [45]. However, both studies lack of real-time evidence to proof their efficiency.

Guo et al. [46] estimated the state of power (SOP) using a second-order partial-adaptive fractional-order model (PA-FOM). An unscented Kalman filter (UKF)-based iterative approaching algorithm (IAA) was developed to determine the peak discharge/charge current for online SOP estimation, avoiding model linearization in the prediction windows and attaining high accuracy across the entire battery operating range. The validation results indicated that the efficacy of online SOP estimation was drastically enhanced, achieving an accuracy of superior to 0.6 W in MAE and 0.7 W in RMSE across two EV driving profiles. However, it lacked comprehensive validation of the reliability of the batteries in relation to long-term aging and degradation effects throughout prolonged operational periods.

Liu et al. [47] suggested an enhanced genetic particle filter (IGPF) to estimate the SOP based on the estimated SOC. The constraints that account for the SOC, voltage, and instruction manual were utilized to assess the discharging SOP and charging SOP. The values of discharge SOP reduced gradually from a substantial value. The charging SOP values progressively increased to a stable level. These evolving trends aligned with the current circumstances, demonstrating the accuracy of the assessments. However, this study does not examine it under extreme temperature conditions, high currents, and aged batteries for the complete verification of its robustness.

Artificial Intelligence techniques have been popular among the researchers of this field. Dang et al. [48] proposed a class of differential equation-informed neural networks (DENNs) for estimating battery SOC, including differential equation-informed multilayer perception (DE-MLP), differential equation-informed recurrent neural network (DE-RNN), and differential equation-informed long short-term memory (DE-LSTM). The suggested approaches not only incorporate physical principles into data-driven methods for estimating SOC but also utilize DENNs to address an inverse problem by estimating the unknown parameters of the differential equation and network parameters simultaneously. The proposed approaches provide a viable option for precisely measuring SOC in batteries, which is critical to maintaining battery dependability and longevity in a variety of applications, including electric vehicles and renewable energy systems. However, this research does not consider other factor such as thermal condition and SOH.

Zhang et al. [49] proposed temporal convolution networks for SOH predictions. The results demonstrated the accuracy and dependability of the proposed method. The proposed method outperformed traditional methods such as ampere-hour counting, which usually suffers from accumulation of errors with extended use. The research also discusses the uncertainty expression of the SOH evaluation and the use of sampling to determine the prediction results’ confidence interval. However, the evaluation of SOH was based on a specific dataset (NASA public data set), which may restrict the generalizability of the results. It would be advantageous to validate the proposed method on a wider variety of datasets in order to evaluate its robustness and applicability in various scenarios.

Dineva [50] conducted an extensive study of the effectiveness of advanced machine learning techniques in predicting SOC with regression, specifically in the presence of dynamic loads. The results validate the significant benefit of advanced ML models in capturing crucial correlations among the variables of interest. The proposed method is effective in forecasting state of charge in battery systems. This is because it has the ability to store historical data and capture the cell dynamics, which is crucial for accurately predicting future charge levels. In another research [51], the ML approach was used to predict the remaining useful life (RUL) of lithium-ion batteries, which is critical for the proper management and maintenance of electric vehicle (EV) battery systems. Several models, including the regression model, the random forest regression model, the logistic regression model, the random forest classification model, and the support vector machine model, were compared to determine their accuracy. Despites the improvements, this paper has a limited parameter for validation and required excessive data for training.

Venugopal [52] suggested using an independently recurrent neural network (IndRNN) to estimate the SOH of lithium-ion batteries in EV. The goal of this study was to figure out exactly how long the batteries have left and when they will fail so that they can be used safely and reliably. The suggested method worked since the experimental results showed that it has a lower mean square error rate than similar architectures for recurrent neural networks. However, there is a dearth of discussion of potential difficulties or disadvantages of applying the suggested IndRNN technique in actual situations. Another drawback is that the suggested method was not compared to non-data-driven methods, like physics-based models, which would have given a more complete picture of how effectively it worked.

Meanwhile, Wang et al. [53] also utilized machine learning techniques, specifically long short-term memory (LSTM), to predict SOH. For this model training, 70% of the initial data were allocated as the training set, while the remaining 30% were designated as the prediction set. The neural network exhibited precise prediction outcomes as a consequence of parameter tuning. The model demonstrated great predictive accuracy across several vehicle types and different operational periods for individual vehicles. This demonstrated that the proposed methodology is capable of accurately evaluating the condition of health of power batteries. However, the drawback is that it necessitates significant parameter adjustments, considerable computational power, and specific knowledge for optimal functionality.

Kim et al. proposed a long short-term memory (LSTM) model to predict the SOC of lithium-ion batteries in [54]. The purpose of the research is to precisely explain the complex nonlinear behavior exhibited during the charging and discharging operations of batteries to forecast the SOC. The LSTM model was trained utilizing battery data gathered under diverse temperature and load situations. However, long-term battery aging was not accounted for, as it impacts the accuracy of SOC predictions, particularly in advanced machine learning models that using historical data. Similarly, the LSTM model was adopted in forecasting the SOC of lithium-ion batteries to improve battery management systems’ performance in [55].

Thomas et al. [56] introduced LSTM, and artificial recurrent neural network (RNN) architecture in machine learning, to estimate the expected lifespan of a battery. The various machine learning models were trained and tested using the datasets from NASA. Estimation of the battery life was conducted based on cell voltage, load voltage, temperature, and charge and discharge cycles. The protection provided by the BMS circuit includes overcharge, over-voltage, over-discharge, high or low temperatures, and cell imbalance at charge and discharge. The compact design circuit has validated its functionality as batteries have been charged successfully. The integration of machine learning for battery monitoring has proved very efficient and effective in predicting the life span of a battery. However, this method is suitable only for a battery pack containing a maximum of eight cells. If a battery has more than eight cells, a greater number of Atmega microcontrollers are required.

Harippriya et al. [57] proposed a BMS designed to forecast the remaining battery charge of the electric vehicle. The degradation of the lithium-ion battery in an electric vehicle was estimated via several machine learning and deep learning algorithms. The parameters, including voltage, current, and temperature, were collected from the sensors and provided as a dataset to the LSTM, Decision Tree (DT), K-Nearest Neighbors (KNN), Naïve Bayes (NB), and Support Vector Machine (SVM) algorithms. The experimental findings demonstrate that the Naïve Bayes algorithm yielded the most favorable outcomes for real-time data, as assessed by metrics including reliability, precision, recall, and F1-score. Naïve Bayes achieved an accuracy rate of 88% and was utilized to determine the remaining battery capacity, aiding in the prediction of lithium-ion battery aging. This research, however, does not address the SOH of batteries including their thermal management.

Sunori et al. [58] developed the machine learning approaches of linear regression and SVM, which are applied in MATLAB for SOC prediction. A secondary dataset comprising voltage, current, temperature, and SOC important in the development of the predictive models. 80% of this dataset has been allocated for training, while 20% has been designated for testing. The SVM model achieved an accuracy of 94.8%, whereas the Linear Regression model demonstrated a superior accuracy of 98%. Despite the efficacy of the suggested method, other critical characteristics must also be considered, as they significantly impact SOC estimation. The research should address additional critical characteristics that may significantly influence SOC estimation, including charge–discharge cycles and internal resistance.

The study in [59] examined the utilization of hybrid reinforcement learning (RL) models by integrating two advanced RL algorithms, namely deep Q-learning (DQL) and active-critic learning, to enhance the charging and discharging operations of lithium-ion batteries in electric vehicles. The hybrid models underwent extensive evaluation through simulation and experimental validation, proving their ability to produce effective battery management techniques. These solutions efficiently adapted for fluctuations in battery state of health (SOH) and state of charge (SOC) relative inaccuracy, reduced battery voltage degradation, and complied with complex operational limitations, including charging and discharging schedules. As a concrete step toward more dependable and sustainable electric transportation networks, the findings demonstrated how RL-based hybrid models might improve BMSs in EVs. The research did not consider the integration of the suggested hybrid reinforcement learning models with the hardware and software frameworks of current battery management systems, thus presenting an issue for practical implementation.

Li et al. [60] developed three models optimized with the particle swarm optimization (PSO) algorithm: the long short-term memory (LSTM) network, convolutional neural network (CNN), and support vector regression (SVR) for precise SOH estimation. The PSO method was utilized to optimize model parameters, resulting in the development of three predictive models: PSO-LSTM, PSO-CNN, and PSO-SVR. The findings indicated that the optimized model markedly enhanced its predictive accuracy, with the RMSE and MAE decreasing by more than 0.5%. Consequently, the minimal percentage reduction in MAPE is 38%, with R² above 0.8, thereby indicating robust fit capabilities and validating the efficacy of the PSO approach. These experiments confirm the PSO-LSTM model as a reliable benchmark for accurately estimating the state-of-health of lithium-ion batteries, highlighting its significant applicability in practical scenarios. However, this research should consider the effects of driving conditions such as acceleration, speed, and route variation on battery health that may impact the accuracy of the SOH prediction.

Khawaja et al. [61] highlight the advancement of SOC and SOH prediction using diverse machine learning methodologies: linear regression analysis, random forest, gradient boosting, light gradient boosting (LightGBM), extreme gradient boosting (XGB), and SVM regressors. According to the findings of this research, the discharge prediction produced by the random forest estimator exhibited a substantially superior performance with minimal accuracy loss. For instance, with an optimal R²-score of 0.999, the random forest regressor exhibited mean absolute error, median absolute error, and RMSE values of 0.0035, 0.0013, and 0.0097, respectively. Nonetheless, it is incomparable to other comprehensive methods, such as meta-heuristic algorithms like PSO, which may similarly require significant processing power and complexity but potentially enhance the outcomes. Additionally, several drive cycles must be conducted for the models to assess their resilience under real situations. However, machine learning algorithms require large size training data with tuning to achieve highly accurate results.

In [62], a machine learning algorithm, long short-term memory neural networks, was utilized together with an estimator algorithm the unscented Kalman filter (UKF) for SOC estimations under dynamic conditions and diverse scenarios. The results showed the good potential of the suggested methodology in accurately estimating SOC in various battery applications. However, this study only reported limited parameter for validation.

Lyu et al. [63] examined battery thermal management in advanced EV and propose an advanced hybrid BTMS architecture. The battery thermal management system integrates thermoelectric cooling, forced air cooling, and liquid cooling. The liquid coolant indirectly interacts with the battery, serving as a medium for dissipating the heat produced during operation. Forced air facilitates heat dissipation from the condenser side of the thermoelectric liquid container. An electric vehicle battery system simulation has been rigorously evaluated to determine its performance. According to experimental findings, the cooling effect is promising, and the power dissipation is acceptable. Furthermore, when 40 V is provided to the heater and 12 V to the thermoelectric cooler (TEC) module, the battery surface temperature reduces roughly 43 °C (from 55 °C to 12 °C) utilizing a TEC-based water-cooling system for a single cell with copper holder. The research lacks an examination of the long-term durability and energy efficiency of the proposed hybrid BTMS in real-world scenarios.

Harwardt et al. [64] proposed a Proximal Policy Optimization (PPO) agent that is trained using Reinforcement Learning (RL) to balance the SOC and temperature of Li-ion battery cells equipped with an active BMS. The active BMS models and battery cell models are coded in Python, together with the training environment for the agent. The efficacy of BMS based on PPO agents as determined from hyperparameter optimization resulted in a reduction in the range of balanced values by at least 28%, and in some instances, by up to 72%. However, this study does not exhaustively show the model’s effectiveness for a variety of real conditions, such as variable loading profiles, extreme temperature changes, the influence of the kinds of aging in lithium-ion systems, thus limiting the possible application.

Borah et al. [65] investigated the prediction of residual energy in a battery cell during discharge throughout a broad range of current, from low to high C-rates. This research presented a novel definition of remaining discharge energy and later conducted a systematic approach that utilized machine learning for its prediction. A machine learning method was proposed to forecast the residual discharge energy at various C-rates and predetermined voltage and temperature cut-off limits. The experimental validation demonstrated that the suggested method can forecast the remaining discharge energy with a relative error of under 3% when the current fluctuates from 0 to 8 C for an NCA cell and 0 to 15 C for a lithium-iron-phosphate (LFP) cell. However, combining machine learning with physics-based modeling introduces additional computational complexity that could be challenging to handle in practical applications.

Another category for BMS method is smart technologies. Internet of things (IoT) has also been applied for BMS. Krishnakumar et al. [66] proposed BMS using IoT to control an electric vehicle’s battery before it overheats due to continuous ON conditions. IoT-based battery management ensures online battery performance and condition monitoring. If the specified parameters fall within the acceptable range, the vehicle remains in a secure state. If the range exceeds permitted limits, the battery and the vehicle may sustain damage. The battery’s performance and voltage will be consistently monitored to prevent such situations. The battery level is displayed to the user via the IoT-enabled mobile application, allowing them to determine the remaining time till depletion. The integration of GPS allows the user to identify the nearest recharging station and the distance. This solution will assist users in avoiding hazardous situations and ensuring the safe functioning of the battery in an electric car by preserving the performance and condition of the lithium-ion battery within specified threshold values. Additionally, standardization of communication protocols and interoperability standards could further accelerate the integration of scalable and adaptive BMS designs on different EV platforms. The solution to these issues will be major enablers that help in the wide-scale adoption of next-generation BMS technologies. However, this method dependent on stable internet for connectivity.

Cao et al. [67] reported a wireless BMS (wBMS) that incorporates wireless communication technologies such as Bluetooth Low Energy (BLE), Zigbee, Near-Field Communication (NFC), Wi-Fi, and cellular networks. This method provides significantly increased flexibility in battery packing, eliminates wiring, decreases battery package weight and volume, saves service and maintenance expenses, and improves reliability by lowering the risks of connection failures in comparison to wired systems. Regardless its optimistic prospects, wBMS encounters obstacles including data security, signal interference, regulatory and standardization hurdles, as well as competition from the ongoing development of wired BMS systems, hence diminishing the apparent benefits of wBMS. Despite their potential, significant advancements must address the challenges associated with these technologies. Data security is a significant concern, and the use of robust encryption and authentication techniques is essential; blockchain appears to be a further solution for enhancing information safety against cyber threats. Signal interference constitutes an additional challenge. It can be enhanced by sophisticated algorithms for signal processing and improved shielding to guarantee reliable transmission. The absence of standardization necessitates collaboration among industry stakeholders and regulators, to develop universal standards for wireless BMS, facilitating seamless interoperability and scalability across diverse applications [67].

The highlighted research papers collectively contribute to advancing BMS for electric vehicles, offering diverse methodologies and applications. The contributions of papers in the field of BMS for electric vehicles are noteworthy. They focus, respectively, on enhancing BMS performance through accurate SOH computation, developing cost-effective BMS with circuit simulation for longevity and efficiency, proposing a centralized BMS architecture for evolving EV systems, introducing an advanced SOC estimation, suggesting a battery thermal management system (BTMS), and presenting an IoT-based BMS for real-time monitoring, contributing to enhanced safety and awareness. From the reviewed works, it can be observed that machine learning, optimization, and estimation algorithms are important in BMS. The algorithms are used to ensure that the battery is operated optimally or in prediction of the battery performance. The works reviewed above are tabulated in Table 2, highlighting the algorithms used and the main issue solved by the algorithm.

Conventional BMS methodologies have primarily concentrated on static algorithms that monitor and control battery parameters according to established models. These techniques, however, are insufficient considering the nonlinear behavior of batteries under fluctuating load circumstances and environmental factors, resulting in poor reliability and diminished battery longevity [68]. Despite technological developments, the integration of multiple computational approaches into a unified BMS framework continues to pose a significant difficulty and constraining potential advantages such as prolonged battery life, enhanced safety, and augmented energy economy.

According to [69], there are three main areas where lithium-ion battery management presents technical issues and challenges. First is the complex and nonlinear electro-thermal behaviour of lithium-ion batteries, thus complicating the modelling process. Second is the inability of sensors to directly assess the internal states of lithium-ion batteries and their high susceptibility to noise and ambient temperature. This makes it challenging to obtain a precise battery estimate. The last issue is the complicated structure when groups of lithium-ion batteries are used, which makes it hard to control precisely.

Moreover, because the work in [69] highlighted the inability to capture all the complexities in real life, inaccuracies might arise in performance predictions. Furthermore, hardware configurations and the settings of parameters can reduce their general applicability to only a few environments.

Real-time validation is one of the major challenges for BMS development. It enables effective and dependable BMS by evaluating algorithms under dynamic, real-world situations. The system’s performance is be evaluated for coverage across many scenarios, encompassing driving behaviors and environmental conditions, hence enhancing robustness and precision. A decentralized yet synchronized real-world system for intelligent battery management was developed in [70] with a universal controller with cloud computing functionality, four charge regulators, and a collection of sensor-equipped battery monitors featuring networking and Bluetooth capabilities. The results demonstrated the effective performance of the decentralized BMS system in maintaining threshold values indicated in the SOC graphs under appropriate conditions, hence preventing both overload and a decline below 80%. Regardless, substantial obstacles exist in obtaining accurate measurements, especially with state of health (SOH) assessments. These limit the proper improvement of precise predictive diagnostics for accurate remaining useful life (RUL) estimation of batteries.

Real-time analysis technologies for BMS like long range (LoRa) have made significant progress [71], but there are still problems with creating accurate real-time estimations. As a result, the majority of these solutions fail to account for all the complexities that arise in real life, which leaves room for model prediction errors. Their effectiveness are limited and their dependability for real-time application greatly impacted by differences in hardware configurations and parameter settings across various situations. Resolving the issues is essential to maximizing LoRa’s potential for enabling long-range, low-cost, and low-power remote battery monitoring.

Reliability and scalability can be improved by applying effective control strategies utilized in shipboard microgrids for battery management systems. Secondary distributed control methods, demonstrated in isolated shipboard microgrids, are crucial for addressing issues related to voltage and frequency fluctuations, accurate power distribution, and maintaining system stability in dynamic environments [72].