Fuzzy logic based charge control strategy for lithium-ion batteries: Comparative evaluation with conventional techniques

Abstract

The global shift toward electric mobility and the growing integration of renewable energy resources have increased the demand for energy storage systems that are efficient, durable, and safe across diverse operating conditions. Lithium-ion batteries currently dominate automotive and stationary energy storage applications due to their high energy density and strong cycle performance. Despite these advantages, battery performance and service life are strongly influenced by operating factors such as temperature, current rate, and the applied charge and discharge regime. Among these factors, charging is particularly critical because it directly shapes terminal voltage response and heat generation. In this context, fast charging remains a key challenge because it can increase polarization losses and heat generation, intensify thermal stress, and accelerate degradation when it is not managed in a condition aware manner. Since batteries in practical operation are exposed to varying initial temperatures and repeated charging events, robust fast charging requires both an adequate modeling approach and a control strategy that can respond to changing operating conditions. This thesis develops a unified electrothermal aging simulation framework for lithium-ion batteries and uses it to design and evaluate an adaptive, fuzzy logic based fast charging strategy. The simulation framework is intended to assess charging performance beyond charge time by capturing temperature evolution, thermal management system activation, and long-term aging trends under repeated operation. The main motivation is that conventional charging profiles are typically fixed in structure. Although these approaches are simple and widely used, they do not naturally adapt to different thermal boundary conditions or to gradual changes in battery behavior over time. An adaptive charging strategy that responds to battery state variables can improve the tradeoff between charging speed and battery stress, especially under cold start and hot start conditions. The electrical behavior in this study is represented using an experimentally derived open circuit voltage relation as a function of state of charge, combined with a temperature dependent direct current internal resistance model. This compact representation aims to reproduce terminal voltage behavior and loss driven heat generation with low computational cost while preserving the dependencies that dominate fast charging. Thermal dynamics are captured through a lumped energy balance in which heat generation is primarily driven by resistive losses, and heat rejection is represented through heat transfer to the surroundings. The model also includes an active liquid cooling system representative of an electric vehicle application. The cooling loop is modeled to engage to regulate battery temperature toward a nominal target near room temperature. Aging is incorporated through combined calendar aging and cycle aging contributions to capacity fade and resistance growth, which enables consistent long-term comparisons among charging strategies. Model parameterization is grounded in experimental characterization. Open circuit voltage is obtained using rest period procedures after partial charge and discharge sequences, which approximates equilibrium voltage over the state of charge range. Direct current internal resistance is identified from pulse tests conducted at several temperature levels to map resistance behavior with temperature. Thermal parameters are tuned using measured transient temperature responses under controlled tests, and aging parameters are calibrated using accelerated aging data sets. Validation using data sets not used during identification indicates that the model captures the coupled electrical and thermal trends required for control evaluation, particularly the relationship among current demand, terminal voltage response, heat generation, and temperature rise. Using this unified simulation approach, five charging strategies are implemented to represent common charging philosophies and support a structured comparison. The strategies considered are constant current constant voltage, multistage constant current constant voltage, constant power constant voltage, time pulse charging, and the proposed fuzzy logic controller. Constant current constant voltage is used as the baseline. Multistage constant current constant voltage and constant power constant voltage represent more aggressive profiles that aim to reduce charge time through staged current levels or power regulation, respectively. Time pulse charging limits temperature rise by introducing rest periods between charging pulses, and it typically trades charging speed for lower sustained thermal loading. In contrast to these predefined profiles, the proposed approach continuously determines the charging current based on the battery operating condition. The fuzzy logic controller in this thesis uses a Mamdani inference structure. The inputs are battery temperature and state of charge, and the output is the charging current command. The rule base is formulated to be interpretable and suitable for tuning and potential embedded implementation. In general, the controller allows higher current at low state of charge when thermal conditions are favorable, and it reduces current as temperature increases and as state of charge moves into higher regions where charging becomes more constrained. This continuous condition based current shaping aims to reduce unnecessary stress without relying on rigid phase definitions or uniformly conservative margins. To evaluate robustness with respect to thermal boundary conditions, all strategies are tested over a common charging window from 10% to 90% state of charge under three scenarios defined by the initial battery temperature. Scenario S1 represents low temperature operation with an initial temperature of 10°C. Scenario S2 represents a reference condition with an initial temperature of 25°C. Scenario S3 represents high temperature operation with an initial temperature of 35°C. These scenarios are selected to examine controller behavior under cold start conditions, a baseline comparison case, and hot conditions where thermal constraints become more influential. Scenario S1 is important because cold start charging can be challenging even when the initial thermal state appears favorable. At low temperature, higher internal resistance and stronger polarization effects can significantly shape terminal voltage response and heat generation underload. Aggressive current demand early in the charging process may therefore be less desirable because it can increase losses and drive the system toward operational limits earlier than expected. Fixed aggressive profiles, such as staged high current approaches, may impose unnecessary stress if they do not explicitly account for temperature dependent sensitivity. Time pulse charging can reduce sustained thermal loading by introducing rest intervals, which allows partial relaxation of electrothermal effects, although it can increase the overall charging duration. The fuzzy controller addresses S1 through continuous current adaptation. By using temperature and state of charge feedback, it can reduce current when operating conditions indicate higher sensitivity and relax this moderation if conditions become more favorable as charging progresses. Scenario S2 provides a reference condition in which differences among charging strategies can be interpreted mainly as consequences of profile design rather than severe boundary constraints. Constant current constant voltage offers predictable current taper behavior but limited adaptation. Multistage constant current constant voltage and constant power constant voltage can improve charging speed depending on how aggressively they use early capability, but they remain fixed shape approaches and cannot respond to deviations in thermal response during the charging process. Time pulse charging improves thermal moderation through rest intervals, typically at the cost of increased charging time. In this baseline scenario, the fuzzy controller behavior can be interpreted as a systematic implementation of practical charging heuristics. It uses available capability at low state of charge and transitions smoothly toward more cautious current levels as temperature increases or as charging approaches regions that are inherently more constrained. Scenario S3 represents hot operation in which the initial thermal margin is reduced, and temperature control becomes critical from the start of charging. In this case, strategies that apply high current early can drive a rapid temperature rise and increase reliance on the cooling system. As a result, aggressive staged profiles may be less suitable unless they explicitly account for the thermal state. This risk is amplified when charging begins from an elevated initial temperature because thermal limits can be reached early in the event. Time pulse charging remains cautious due to its rest intervals, but this typically increases total charging time. The fuzzy controller is particularly relevant in S3 because temperature is an explicit input, allowing it to reduce current immediately as temperature increases and to support controlled thermal behavior without enforcing an overly cautious profile under all conditions. A key aspect of this study is that charging strategies are evaluated with an integrated active liquid cooling model rather than assuming a fixed temperature boundary. This enables a system-level interpretation in which profiles that generate more heat require stronger thermal management action, while condition-aware current shaping can reduce thermal burden by limiting unnecessary heat generation in sensitive regions. In addition, an aging model linking repeated operating stress to capacity fade and resistance growth enables long-horizon comparisons across strategies. In summary, this thesis presents an experimentally grounded and computationally compact electrothermal aging simulation approach and demonstrates it by comparing multiple fast charging strategies under three representative initial temperature scenarios. The results indicate that fixed profile strategies are simple but can be less robust across cold, reference, and hot start conditions, whereas an adaptive Mamdani fuzzy controller can continuously shape charging current using temperature and state-of-charge feedback. Future work could focus on broader validation at cell and module level, hardware-in-the-loop implementation, and refinement of pack-level aging and system effects.