This paper is concerned with energy efficient operation of an integral thermal management system (ITMS) for electric vehicles using a nonlinear model predictive control (MPC). Driven by a heat pump (HP), this ITMS can handle battery thermal management (BTM) while serving the need for cabin cooling or heating need. The objectives of the ITMS MPC control strategy include minimization of power consumption and achieving temperature setpoint regulation for the battery and cabin space based on predictive information of traction power and cabin thermal load. The control design is facilitated by a gray‐box modeling framework, in which the nonlinear dynamics of HP subsystem are characterized with a data-driven Koopman subspace model, while the BTM subsystem dynamic is a bilinear physics-based model. The computational efficiency of the proposed MPC framework is improved with two aspects of convexification for the underlying receding-horizon constrained optimization problem: the Koopman-operator lifting and the McCormick envelopes implemented for handling the bilinear dynamics. The proposed control method is evaluated with simulation study, by developing a Modelica-Python cosimulation platform via the functional mockup interface (FMI), where the electric vehicle (EV)-ITMS plant is modeled in Modelica with Dymola and the MPC design is implemented in Python. By benchmarking against a recurrent-neural-networks (RNN) model based nonlinear MPC, the simulation results validate the effectiveness and improved computational efficiency of the proposed method.