Cette thèse porte sur l’évaluation de plusieurs stratégies de gestion d’énergie pour un nouveau concept de moteur hybride pneumatique. Ce concept combine un moteur à combustion interne avec un système de stockage d’énergie sous forme d’air comprimée. Une soupape supplémentaire relie alors la chambre de combustion à un réservoir d’air et permet un fonctionnement en mode moteur pneumatique ou pompe pneumatique (récupératif). La première stratégie, Causale, est basée sur des principes heuristiques. La deuxième, à Coefficient de Pénalité Constant, vise la minimisation d’un critère énergétique global. Un coefficient de pondération permet de mettre en opposition, pour un travail donné, les coûts énergétiques d’un mode pneumatique d’une part et d’un mode thermique d’autre part. Le mode offrant le coût le plus faible sera choisi. La troisième stratégie, à Coefficient de Pénalité Variable, sur le même principe utilise un coefficient de pondération variable selon la quantité d’énergie pneumatique disponible. Une stratégie, à reconnaissance de situation de conduite, permet d’adapter les stratégies à la situation reconnue (par exemple, embouteillage, autoroutier). Enfin, la dernière stratégie tente de recopier la solution optimale de référence (obtenue par programmation dynamique) à l’aide d’un modèle. Toutes les stratégies ont été validées en simulation sur cycles standards. De plus une méthode, basée sur les chaînes de Markov, de constructions de cycle de conduite « artificiels » mais réalistes est proposée. Les consommations obtenues avec les différentes stratégies proposées sont comparées en référence aux consommations minimales atteignables. Les résultats montrent que 40% de gain de consommation peuvent être atteints. / This thesis presents a study of several energy management strategies for a novel hybrid pneumatic engine concept. The concept combines an internal combustion engine with a system of compressed air for energy storage. An additional charge valve connects the combustion chamber to an air pressure tank, enabling the engine to function in pneumatic motor mode or as a pneumatic pump (recuperation mode). The first strategy is called Causal and implements a rule-based control technique. The second one, called Constant Penalty Coefficient, is derived from optimal control theory and is based on an equivalent consumption minimization strategy. A penalty coefficient is introduced to evaluate, for a given torque demand, the respective energy costs of the two modes, pneumatic and conventional, enabling the mode offering the lowest cost to be chosen. The third strategy, called Variable Penalty Coefficient, is based on the same principle but uses a variable penalty coefficient depending on the amount of pneumatic energy available in the compressed air tank. Another strategy investigated, called Driving Pattern Recognition, adapts the strategies to the driving situation recognized (for example, traffic jam, or highway). The last strategy studied attempts to reproduce the optimal reference solution obtained by dynamic programming, using a neural mode. All the strategies have been validated by simulation on standard driving cycles. In addition, a method based on the Markov chain process have been develop to make ‘artificial’ yet realistic driving cycles. The consumptions obtained with the various strategies are compared with the minimal consumptions achievable. Results demonstrate that 40% of fuel saving can be achieved on certain cycles. Several of the strategies proposed give results that are close to optimal.

In early design phase for new hybrid electric vehicle (HEV) powertrains, simulation isused for the estimation of vehicle fuel consumption. For hybrid electric powertrains,fuel consumption is highly related to powertrain efficiency. While powertrainefficiency of hybrid electric powertrain is not a linear product of efficiencies ofcomponents, it has to be analysed as a sequence of energy conversions includingcomponent losses and energy interaction among components.This thesis is aimed at studying the energy losses and flows and present them in theform of Sankey diagram, later, an adaptive energy management system is developedbased on current rule-based control strategy. The first part involves developing energycalculation block in GT-SUITE corresponding to the vehicle model, calculating allthe energy losses and flows and presenting them in Sankey diagram. The secondpart involves optimizing energy management system control parameters according todifferent representative driving cycles. The third part involves developing adaptiveenergy management system by deploying optimal control parameter based on drivingpattern recognition with the help of SVM (support vector machine).In conclusion, a sturctured way to generate the Sankey diagram has been successfullygenerated and it turns out to be an effective tool to study HEV powertrain efficiencyand fuel economy. In addition, the combination of driving pattern recognition andoptimized control parameters also show a significant potential improvement in fuelconsumption. / Under den tidiga utvecklingsfasen av nya elektrifieradedrivlinor for hybridapplikationer (HEV) används simulering för uppskattning avfordonets bränsleförbrukning. För dess drivlinor är bränsleförbrukningen i hög gradkopplad till drivlinans verkningsgrad. Även om drivlinans verkningsgrad inte ären linjär prokukt av komponenternas verkningsgrad behöve rden analyseras somen sekvens av energiomvandlingar, inklusive förluster och energipåverkan mellankomponenter.Detta examensarbete syftar till att undersöka energiförluster och flöden samtpresentera dessa i form av sankey diagram. Senare utvecklas ett anpassningsbartenergihanteringssystem baserat på nuvarande regelbaserad kontrollstrategi. Deninledande delen involverar utvecklandet av energianalys i GT-SUITE som motsvararfordonsmodellen, beräkningar av totala energiförluster och flöden samt presentationav dessa i ett sankey diagram. Den andra delen innefattar optimering avenergihanteringssystems kontrollparametrar enligt olika representativa körcykler.Den tredje delen involverar utveckling av anpassningsbara energihanteringssystemgenom användning av optimala kontrollparameterar baserad på detektering avkörbeteende med hjälp av SVM ( stödvektormaskin).Slutligen, ett strukturerat sätt att generera sankey diagrammet har med framgånggenererats och visat sig vara ett effektivt verktyg för studier av HEV drivlinorseffektivitet och bränsleekonomi. Dessutom visar kombinationen av detektering avkörbeteende och optimerade kontrollparametrar på en markant potentiell förbättringi bränsleförbrukning.