Philosophiae Doctor - PhD / Proton exchange membrane fuel cells (PEMFC) converts chemical energy from the electrochemical reaction of oxygen and hydrogen into electrical while emitting heat, oxygen depleted air (ODA) and water as by-products. The by-products have useful functions in aircrafts, such as heat that can be used for ice prevention, deoxygenated air for fire retardation and drinkable water for use on board. Consequently, the PEMFC is also studied to optimize recovery of the useful products. Despite the progress made, durability and reliability remain key challenges to the fuel cell technology. One of the reasons for this is the limited understanding of PEMFC behaviour in the aeronautic environment. The aim of this thesis was to define a comprehensive non-intrusive diagnostic technique that provides real time diagnostics on the PEMFC State of Health (SoH). The framework of the study involved determining factors that have direct influence on fuel cell life in aeronautic environment through a literature survey, examining the effects of the factors by subjecting the PEMFC to simulated conditions, establishing measurable parameters reflective of the factors and defining the diagnostic tool based on literature review and this thesis finding.

Operating batteries in parallel is attempted to overcome the problems with conventionally used battery bank, in which batteries are connected in series. The problems and the management with the operation of serial connected batteries are first addressed. The related topics to the parallel configuration are reviewed. Then, the parallel configuration with battery power modules is proposed. The battery power module can be realized with different dc-to-dc converters for different applications. When batteries are charged in parallel, the problem of over-charge can be avoided. With parallel operation, the discharging currents of the batteries are independently controlled but are coordinated to execute a full amount load current. This allows for scheduling the discharging profiles under different operating conditions. As a result, a sophisticated discharging profile can be realized to utilize the available stored energy in batteries. On the other hand, some of the batteries may take rest or be isolated from the system for the detections at a time. This facilitates the estimations of the state of charge (SOC) and the state of health (SOH). Moreover, the completely exhausted or damaged batteries can be isolated from the battery power supply bank without interrupting the system operation. Experiments are carried out on battery power modules with lead-acid batteries incorporating with associated buck-boost converters. The experimental results demonstrate that a more efficient utilization of battery energy can be achieved. On the other hand, a more reasonable management can be done with simple estimation methods of the SOC and the SOH.

This thesis proposes an enhanced ampere-hour counting method based on the depth-of-discharge (DOD) to estimate the state-of-charge (SOC) and state-of-health (SOH) for lead-acid batteries. Not only the losses at different discharging currents, but also the releasable capacity at the exhausted state caused by the larger discharging current are considered and compensated. Furthermore, the SOH is revaluated at the exhausted state by the maximum releasable capacity, consequently leading to more accurate SOC estimation. Through the experiments that emulate practical operations, the experimental results reveal that the maximum error is less than 6 %.

A novel battery power system configured by the battery power modules (BPMs) is proposed. Each BPM consists of a single battery pack or a battery bank equipped with an associated DC/DC converter. The output ports of BPMs can be connected in series for the high voltage applications, or in parallel to cope with a higher power or energy. For a large scale battery power system, a number of BPMs can be arrayed with combination of series and parallel connections to meet the load requirements. These all configurations allow the BPMs be operated individually. Consequently, the discharging currents of the batteries can be independently controlled, but coordinated to provide a full amount of the load current. The performances of BPMs connected in both parallel and series at outputs are analyzed theoretically and discussed from the experimental results. Batteries operating independently do not suffer from charge imbalance, and thus can avoid being over-charged or over-discharged, so that the life cycle can be prolonged. Furthermore, sophisticated discharging profiles such as intermittent currents can be realized to equalize the charges and thus to efficiently utilize the available stored energy in batteries. During the operation period, some of the batteries may take rest or be isolated from the system for the open-circuit measurement, facilitating the estimation of the state-of-charge (SOC) and the evaluation of the state-of-health (SOH). With the benefit of independent operation, the BPMs can be discharged with a scheduled current profile, such as intermittent discharging. The investigation results show that the average current plays the most important role in current discharging. By detecting the battery voltage at the break time, an SOC estimation method based on the dynamically changed open-circuit voltage exhibits an acceptable accuracy in a shorter time with considerations of the previous charging/discharging currents and the depth-of- discharge (DOD). In addition, the coulomb counting method can be enhanced by evaluating the SOH at the exhausted and fully charged states, which can be intended on the independently operated BPMs. Through the experiments that emulate practical operations, the SOC estimation methods are verified on lead-acid batteries and lithium-ion batteries to demonstrate the effectiveness and accuracy.

Dans cette thèse, les mécanismes de vieillissement des cellules Li-ion sont analysés sur un niveau théorique,assisté par une AMDEC (Analyse des modes de défaillance, de leurs effets et de leur criticité). L'accent est mis surla famille des cellules lithium fer phosphate (LFP) utilisées comme batteries de traction dans les applicationsvéhicules électriques.L'objectif de la partie xpérimentale de cette thèse est le développement d'un concept d'un outil de diagnostic pourla détermination rapide d'état de la batterie. Une expérience de vieillissement accélérée est réalisée avec un groupede cellules LFP de haute capacité (70Ah). Les cellules sont analysées en utilisant des méthodes de mesured'impédance dans les domaines temporel et fréquentiel. La pectroscopie d'Impédance Électrochimique (SIE, ouEIS en anglais) s'est trouvée être un bon outil pour révéler des informations intéressantes sur l'état de santé (Stateof-Health, SOH) de la batterie.Des modèles de batterie sont utilisés pour l'interprétation des résultats de mesure. En comparant différents modèlesdu circuit équivalent (ECMs), un modèle est choisi. Ce modèle est utilisé pour la détermination du SOC et étendupour la détermination du SOH. Un concept pour la détermination du SOH est développé, permettant uneapproximation de la capacité de la batterie dans une période de temps de moins de 30s, si les onditions de labatterie et d'environnement, comme la température et l'état de charge de la batterie, sont connus. / In this thesis, the aging mechanisms withing Li-ion cells are analyzed on a theoretical level, supported by an FMEA(Failure ode and Effects Analysis). The focus lies on the group of lithium iron phosphate (LFP) cells used fortraction batteries in electric vehicles. Scope of the experimental part of the thesis is the development of a diagnosticconcept for the quick battery state determination. A group of high capacity LFP cells (70Ah) designed for tractionpurposes in electric vehicles is aged artificially and investigated afterwards by impedance measurements in the timeand frequency domain. Electrochemical impedance spectroscopy (EIS) is found to reveal interesting information onthe battery's State-of-Health (SOH).For the interpretation of the measurement results, battery models are employed. Different equivalent circuit models(ECMs) are compared and an appropriate model is chosen, which is used for the SOC (State-of-Charge)determination and extended for the SOH (State-of-Health) determination. An SOH determination concept isdeveloped, which allows the approximation of the cell capacity in less than 30s, if the battery and environmentalconditions, such as the temperature and the cell's SOC, are known.

Pour assurer une énergie électrique de qualité et de façon permanente, il existe des systèmes électroniques d’alimentation spécifiques. Il s’agit des Alimentations Sans Interruptions (ASI). Une ASI comme tout autre système peut tomber en panne ce qui peut entrainer une perte de redondance. Cette perte induit une maintenance corrective donc une forme d’indisponibilité ce qui représente un coût. Nous proposons dans cette thèse de travailler sur deux composants parmi les plus sensibles dans les ASI à savoir les condensateurs électrolytiques et les batteries au plomb. Dans une première phase, nous présentons, les systèmes de surveillance existants pour ces deux composants en soulignant leurs principaux inconvénients. Ceci nous permet de proposer le cahier des charges à mettre en œuvre. Pour les condensateurs électrolytiques, nous détaillons les différentes étapes de caractérisation et de vieillissement ainsi que la procédure expérimentale de vieillissement standard accéléré et les résultats associés. D’autre part, nous présentons les résultats de simulation du système de surveillance et de prédiction de pannes retenu. Nous abordons la validation expérimentale en décrivant le système développé. Nous détaillons les cartes électroniques conçues, les algorithmes mis en œuvre et leurs contraintes d’implémentation respectifs pour une réalisation temps réel. Enfin, pour les batteries au plomb étanches, nous présentons les résultats de simulation du système de surveillance retenu permettant d’obtenir le SOC et le SOH. Nous détaillons la procédure expérimentale de vieillissement en cycles de charge et décharge de la batterie nécessaire pour avoir un modèle électrique simple et précis. Nous expliquons les résultats expérimentaux de vieillissement pour finir avec des propositions d’amélioration de notre système afin d’obtenir un SOH plus précis. / To ensure power quality and permanently, some electronic system supplies exist. These supplies are the Uninterrupted Power Supplies (UPS). An UPS like any other system may have some failures. This can be a cause of redundancy loss. This load loss causes a maintenance downtime which may represent a high cost. We propose in this thesis to work on two of the most sensitive components in the UPS namely electrolytic capacitors and lead acid batteries. In a first phase, we present the existing surveillance systems for these two components, highlighting their main drawbacks. This allows us to propose the specifications which have to be implemented for this system. For electrolytic capacitors, we detail different stages of characterization ; the aging accelerated standard experimental procedure and their associated results. On the other hand, we present the simulation results of monitoring and failure prediction system retained. We discuss the experimental validation, describing the developed system. We detail the electronic boards designed, implemented algorithms and their respective constraints for a real time implementation. Finally, for lead acid batteries, we present the simulation results of the monitoring system adopted to obtain the SOC and SOH. We describe the aging experimental procedure of charging and discharging cycles of the batteries needed to find a simple and accurate electric models. We explain the aging experimental results and in the end we give suggestions for improving our system to get a more accurate SOH.

La mesure d'impédance électrique en embarqué sur véhicule est un sujet clé pour améliorer les fonctions de diagnostic d'un pack batterie. On cherche en particulier à fournir ainsi des mesures supplémentaires à celles du courant pack et des tensions cellules, afin d'enrichir les indicateurs de vieillissement dans un premier temps, et d'état de santé et de charge dans un second temps. Une méthode classique de laboratoire pour obtenir des mesures d'impédance d'une batterie est la spectroscopie d'impédance électrochimique (ou EIS). Elle consiste à envoyer un signal sinusoïdal en courant (ou tension) de fréquence variable balayant une gamme de fréquences d'intérêt et mesurer ensuite la réponse en tension (ou courant) pour chaque fréquence. Une technique d'identification active basée sur l'utilisation des signaux large bande à motifs carrés est proposée. En particulier, des simulations ont permis de comparer les performances d'identification de différents signaux d'excitation fréquemment utilisés dans le domaine de l'identification et de vérifier les conditions correspondant à un comportement linéaire et invariant dans le temps de l'élément électrochimique. L'évaluation de la qualité d'estimation est effectuée en utilisant une grandeur spécifique : la cohérence. Cette grandeur statistique permet de déterminer un intervalle de confiance sur le module et la phase de l'impédance estimée. Elle permet de sélectionner la gamme de fréquence où la batterie respecte les hypothèses imposées par la méthode d'identification large bande. Afin de valider les résultats, une électronique de test a été conçue. Les résultats expérimentaux permettent de mettre en valeur l'intérêt de cette approche par motifs carrés. Un circuit de référence est utilisé afin d'évaluer les performances en métrologie des méthodes. L'étude expérimentale est ensuite poursuivie sur une batterie Li-ion soumise à un courant de polarisation et à différents états de charge. Des essais comparatifs avec l'EIS sont réalisés. Le cahier de charge établi à l'aide d'un simulateur de batterie Li-ion a permis d'évaluer les performances de la technique large bande proposée et de structurer son utilité pour l'estimation des états de vieillissement et de charge. / Embedded electrical impedance measurement is a key issue to enhance battery monitoring and diagnostic in a vehicle. It provides additional measures to those of the pack's current and cell's voltage to enrich the aging's indicators in a first time, and the battery states in a second time. A classical method for battery impedance measurements is the electrochemical impedance spectroscopy (EIS). At each frequency, a sinusoidal signal current (or voltage) of a variable frequency sweeping a range of frequencies of interest is at the input of the battery and the output is the measured voltage response (or current). An active identification technique based on the use of wideband signals composed of square patterns is proposed. Particularly, simulations were used to compare the performance of different excitation signals commonly used for system identification in several domains and to verify the linear and time invariant behavior for the electrochemical element. The evaluation of the estimation performance is performed using a specific quantity: the spectral coherence. This statistical value is used to give a confidence interval for the module and the phase of the estimated impedance. It allows the selection of the frequency range where the battery respects the assumptions imposed by the non-parametric identification method. To experimentally validate the previous results, an electronic test bench was designed. Experimental results are used to evaluate the wideband frequency impedance identification. A reference circuit is first used to evaluate the performance of the used methodology. Experimentations are then done on a Li–ion battery. Comparative tests with EIS are realized. The specifications are established using a simulator of Li-ion battery. They are used to evaluate the performance of the proposed wide band identification method and fix its usefulness for the battery states estimation: the state of charge and the state of health.

Electric vehicles have received substantial attention in the past few years since they provide a more sustainable, efficient, and greener transportation alternative in comparison to conventional fossil-fuel powered vehicles. Lithium-Ion batteries represent the most important component in the electric vehicle powertrain and thus require accurate monitoring and control. Many challenges are still facing the mass market production of electric vehicles; these challenges include battery cost, range anxiety, safety, and reliability. These challenges can be significantly mitigated by incorporating an efficient battery management system. The battery management system is responsible for estimating, in real-time, the battery state of charge, state of health, and remaining useful life in addition to communicating with other vehicle components and subsystems. In order for the battery management system to effectively perform these tasks, a high-fidelity battery model along with an accurate, robust estimation strategy must work collaboratively at various power demands, temperatures, and states of life. Lithium ion batteries are considered in this research. For these batteries, electrochemical models represent an attractive approach since they are capable of modeling lithium diffusion processes and track changes in lithium concentrations and potentials inside the electrodes and the electrolyte. Therefore, electrochemical models provide a connection to the physical reactions that occur in the battery thus favoured in state of charge and state of health estimation in comparison to other modeling techniques. The research presented in this thesis focuses on advancing the development and implementation of battery models, state of charge, and state of health estimation strategies. Most electrochemical battery models have been verified using simulation data and have rarely been experimentally applied. This is because most electrochemical battery model parameters are considered proprietary information to their manufacturers. In addition, most battery models have not accounted for battery aging and degradation over the lifetime of the vehicle using real-world driving cycles. Therefore, the first major contribution of this research is the formulation of a new battery state of charge parameterization strategy. Using this strategy, a full-set of parameters for a reduced-order electrochemical model can be estimated using real-world driving cycles while accurately calculating the state of charge. The developed electrochemical model-based state of charge parameterization strategy depends on a number of spherical shells (model states) in conjunction with the final value theorem. The final value theorem is applied in order to calculate the initial values of lithium concentrations at various shells of the electrode. Then, this value is used in setting up constraints for the optimizer in order to achieve accurate state of charge estimation. Developed battery models at various battery states of life can be utilized in a real-time battery management system. Based on the developed models, estimation of the battery critical surface charge using a relatively new estimation strategy known as the Smooth Variable Structure Filter has been effectively applied. The technique has been extended to estimate the state of charge for aged batteries in addition to healthy ones. In addition, the thesis introduces a new battery aging model based on electrochemistry. The model is capable of capturing battery degradation by varying the effective electrode volume, open circuit potential-state of charge relationship, diffusion coefficients, and solid-electrolyte interface resistance. Extensive experiments for a range of aging scenarios have been carried out over a period of 12 months to emulate the entire life of the battery. The applications of the proposed parameterization method combined with experimental aging results significantly improve the reduced-order electrochemical model to adapt to various battery states of life. Furthermore, online and offline battery model parameters identification and state of charge estimation at various states of life has been implemented. A technique for tracking changes in the battery OCV-R-RC model parameters as battery ages in addition to estimation of the battery SOC using the relatively new Smooth Variable Structure Filter is presented. The strategy has been validated at both healthy and aged battery states of life using driving scenarios of an average North-American driver. Furthermore, online estimation of the battery model parameters using square-root recursive least square (SR-RLS) with forgetting factor methodology is conducted. Based on the estimated model parameters, estimation of the battery state of charge using regressed-voltage-based estimation strategy at various states of life is applied. The developed models provide a mechanism for combining the standalone estimation strategy that provide terminal voltage, state of charge, and state of health estimates based on one model to incorporate these different aspects at various battery states of life. Accordingly, a new model-based estimation strategy known as the interacting multiple model (IMM) method has been applied by utilizing multiple models at various states of life. The method is able to improve the state of charge estimation accuracy and stability, when compared with the most commonly used strategy. This research results in a number of novel contributions, and significantly advances the development of robust strategies that can be effectively applied in real-time on-board of a battery management system. / Thesis / Doctor of Philosophy (PhD)

Due to the large share of production costs, the lifespan of an electric vehicle’s (EV) lithium-ion traction battery should be as long as possible. The optimisation of the EV’s operating strategy with regard to battery life requires a regular evaluation of the battery’s state-of-health (SOH). Yet the SOH, the remaining battery capacity, cannot be measured directly through sensors but requires the elaborate conduction of special characterisation tests. Considering the limited number of test facilities as well as the rapidly growing number of EVs, time-efficient and scalable SOH estimation methods are urgently needed and are the object of investigation in this work. The developed virtual SOH experiment originates from the incremental capacity measurement and solely relies on the commonly logged battery management system (BMS) signals to train the digital battery twins. The first examined dataset with identical load profiles for new and aged battery state serves as proof of concept. The successful SOH estimation based on the second dataset that consists of varying load profiles with increased complexity constitutes a step towards the application on real driving cycles. Assuming that the load cycles contain pauses and start from the fully charged battery state, the SOH estimation succeeds either through a steady shift of the load sequences (variant one) with an average deviation of 0.36% or by random alignment of the dataset’s subsequences (variant two) with 1.04%. In contrast to continuous capacity tests, the presented framework does not impose restrictions to small currents. It is entirely independent of the prevailing and unknown ageing condition due to the application of battery models based on the novel encoder–decoder architecture and thus provides the cornerstone for a scalable and robust estimation of battery capacity on a pure data basis.

The continuous monitoring of the state-of-health (SOH) of electric vehicles (EV) represents a problem with great research relevance due to the time-consuming battery cycling and capacity measurements that are usually required to create a SOH estimation model. Instead of the widely used approach of modelling the battery’s degradation behaviour with as little cycling effort as possible, the applied SOH monitoring approach is the first of its kind that is solely based on commonly logged battery management system (BMS) signals and does not rely on tedious capacity measurements. These are used to train the digital battery twins, which are subsequently subjected to virtual capacity tests to estimate the SOH. In this work, transfer-learning is applied to increase the data and computational efficiency of the digital battery twins training process to facilitate a real-world application as it enables SOH estimation for unknown ageing states due to the selective parameter initialisation at less than a tenth of the common training time. However, the successful SOH estimation with a mean SOH deviation of 0.05% using transfer-learning still requires the presence of pauses in the dataset. Meta-learning extends the idea of transfer-learning as the baseline model simultaneously takes several ageing states into account. Learning the basic battery-electric behaviour it is forced to preserve a certain level of uncertainty at the same time, which seems crucial for the successful fine-tuning of the model parameters based on three pause-free load profiles resulting in a mean SOH deviation of 0.85%. This optimised virtual SOH experiment framework provides the cornerstone for a scalable and robust estimation of the remaining battery capacity on a pure data basis.