Abstract-An optimized control strategy for induction machines is presented and compared to classic strategies. The described method allows to optimize voltage and frequency for a steady state equivalent circuit model of three-phase induction machine. This method is applied to an electric vehicle by simulating driving cycles and calculating energy consumption. The potential gain for the optimized control strategy is discussed.

This paper proposes a method to the model constraints from different models to run an optimization over models with different granularities. Through machine learning, the proposed method has proven to be able to accurately map the constraints and minimize the number of call to the model. It handles both continuous and discrete variables and mixes design rules to statistic approach to create a surrogate of the model. Index Terms-Constraint modeling, finite-element (FE) model, machine learning, optimal design, random forest.

Purpose The purpose of this paper is to compare two design optimization architectures for the optimal design of a complex device that integrates simultaneously the sizing of system components and the control strategy for increasing the energetic performances. The considered benchmark is a battery electric passenger car. Design/methodology/approach The optimal design of an electric vehicle powertrain is addressed within this paper, with regards to performances and range. The objectives and constraints require simulating several vehicle operating points, each of them has one degree of freedom for the electric machine control. This control is usually determined separately for each point with a sampling or an optimization loop resulting in an architecture called bi-level. In some conditions, the control variables can be transferred to the design optimization loop by suppressing the inner loop to get a mono-level formulation. The paper describes in which conditions this transformation can be done and compares the results for both architectures. Findings Results show a calculation time divided by more than 30 for the mono-level architecture compared to the natural bi-level on the study case. Even with the same models and optimization algorithms, the structure of the problem should be studied to improve the results, especially if computational cost is high. Originality/value The compared architectures bring new guidelines in the field optimal design for electric powertrains. The way to formulate a design optimization with some inner degrees of freedom can have a significant impact on computing time and on the problem understanding

In this paper, a global optimization methodology is described to pre-design an electric vehicle powertrain in order to find the best compromises between components. The modeled system includes a transmission, an electric machine, an inverter and a battery pack. The challenge is to find the dedicated formulations, with the vehicle performance requirements, electric range, and cost calculation that include the whole system without exploding computational time. Bi-objective, range/costs, optimizations with performance constraints are performed to find the potential gain with the system model. Keywords-Electric vehicle powertrain, optimization methodology, electric system design, methodology Notations:Sp: Vehicle Speed (km/h) t: driving cycle time (s) Sl: Road slope dp: number of operating points that represent a portion of a driving cycle pp: number of operation points that represent a performance requirements ng: number of gears of the transmission Tw: required torque at the wheel (Nm) Ωw: wheel speed (rpm) Tm: required machine torque (Nm) Ωm: machine speed (rpm) Tm max : maximum machine achievable torque (Nm) Tm req : required machine torque for performance (Nm) f: stator frequency (Hz) Um: machine voltage (V) Im: machine current (A) cos(φ): machine power factor J: stator current density (A/mm²) Ib: battery current (A) Ub: battery voltage (V) x i : design variable ns: number of cells in series for the battery pack np: number of cells in parallel for the battery pack SOC: battery state of charge