New Hybrid Control Strategy Boosts Efficiency and Responsiveness of EV Power Systems
In the ever-evolving landscape of electric vehicle (EV) technology, power electronics continue to play a pivotal role in shaping performance, efficiency, and reliability. A recent breakthrough from researchers at Changsha University of Science and Technology introduces a novel hybrid control strategy that significantly enhances the operation of dual active bridge (DAB) DC-DC converters—key components in next-generation EV drivetrains. This innovation promises to tackle two persistent challenges in EV power systems: excessive current stress during variable-voltage operation and sluggish dynamic response under real-world driving conditions.
The study, led by Guan Weide, Li Tao, Zhong Jian, Wang Xuhong, and Xia Xiangyang, proposes a model predictive control (MPC) framework integrated with real-time current stress optimization. The resulting hybrid approach—dubbed MPC-CSO (Model Predictive Control and Current Stress Optimization)—delivers simultaneous improvements in steady-state efficiency, transient response, and robustness against parameter mismatches. These attributes are critical for modern EVs, which demand agile power delivery during frequent acceleration and regenerative braking cycles while maintaining high energy efficiency across a wide speed range.
At the heart of this advancement lies the strategic placement of a DAB converter between the main traction battery and the motor inverter. Unlike conventional EV architectures that operate with a fixed DC bus voltage—typically optimized for peak performance at high speeds—this configuration enables dynamic adjustment of the DC link voltage based on real-time motor demands. In urban driving scenarios, where vehicles spend the majority of time at low to medium speeds, lowering the bus voltage reduces switching losses, harmonic distortion, and torque ripple in the inverter and motor, thereby improving overall system efficiency.
However, this variable-voltage approach introduces a significant trade-off: when the input and output voltages of the DAB converter become mismatched—a common occurrence during low-speed operation—the circulating currents can surge, leading to elevated conduction losses, thermal stress on semiconductor devices, and potential loss of soft-switching benefits. Traditional control methods, such as single-phase-shift (SPS) modulation paired with PI controllers, struggle to mitigate these effects, especially under rapidly changing load conditions.
To address this, the research team selected dual-phase-shift (DPS) modulation as the foundation for their control strategy. DPS offers two independent control degrees of freedom—the inner phase shift (D1) and the outer phase shift (D2)—enabling infinite combinations to deliver the same power level while allowing optimization of current stress. While prior studies have explored DPS for stress reduction, they often rely on computationally intensive offline calculations or lack dynamic performance.
The innovation of MPC-CSO lies in its seamless fusion of predictive control with analytical stress minimization. Instead of treating optimization as a separate step, the team embedded the current stress minimization condition directly into the MPC cost function. Using Lagrange multiplier theory, they derived closed-form expressions that link the optimal D1 and D2 values to the desired power transfer and voltage ratio. This eliminates the need for iterative solvers or lookup tables, making the solution practical for real-time implementation on standard automotive-grade microcontrollers like the STM32F405.
Moreover, recognizing that model predictive control is inherently sensitive to inaccuracies in system parameters—such as variations in inductance due to temperature drift or manufacturing tolerances—the team introduced a lightweight error-correction mechanism. By feeding back the difference between predicted and actual output voltage into the prediction model at each control cycle, the system continuously self-corrects, maintaining high tracking accuracy even when the physical DAB deviates from its nominal model. This feedback loop transforms the otherwise open-loop nature of standard MPC into a robust closed-loop architecture without adding significant computational overhead.
Experimental validation was conducted on a scaled-down prototype platform featuring a 1 kW DAB converter interfaced with a three-phase permanent magnet synchronous motor (PMSM) drive system. Under steady-state conditions with a 48 V input and 32 V output (voltage ratio k = 1.5), the MPC-CSO strategy achieved a peak inductor current of approximately 13.1 A—comparable to other DPS-based optimized controls and markedly lower than the 16.2 A observed under conventional SPS-PI control. Infrared thermography confirmed these findings: MOSFET junction temperatures under MPC-CSO stabilized at 48.4°C, versus 57.3°C under SPS-PI, underscoring the tangible thermal benefits of reduced current stress.
But where MPC-CSO truly shines is in dynamic scenarios. During a step-load test—simulating sudden changes in motor torque demand—the system regulated the output voltage with near-zero overshoot (just 1 V) and settled within 5 milliseconds. In contrast, DPS-PI and DPS-LCFF (load current feedforward) controls exhibited 7 V and 4 V overshoots with settling times of 60 ms and 40 ms, respectively. For an EV navigating stop-and-go traffic or executing rapid lane changes, such responsiveness translates directly into smoother power delivery, enhanced drivability, and reduced stress on the battery and powertrain components.
The team also evaluated the system under realistic motor-load conditions. Using a variable DC bus strategy—16 V at 1,000 rpm, 32 V at 2,000 rpm, and 48 V at 3,000 rpm—they demonstrated a substantial efficiency gain in low- and mid-speed regimes. At 1,000 rpm, the overall system efficiency (including both DAB and inverter) rose from 68.7% under fixed-voltage operation to 81.6% with the proposed method. At 2,000 rpm, it improved from 80.6% to 84.0%. Only at the highest speed (3,000 rpm), where the fixed and variable strategies converge at 48 V, did efficiency dip slightly (88.8% vs. 87.3%)—a negligible trade-off given the dominant urban driving profile of most EVs.
Crucially, the error-correction feature proved its worth in parameter sensitivity tests. When the actual inductance was increased by 33% (from 18 µH to 24 µH) while the controller still used the nominal value, the uncorrected MPC exhibited a steady-state voltage error of 3.5 V. With error correction enabled, this deviation vanished, confirming the strategy’s robustness in real-world hardware where component tolerances and aging are unavoidable.
From an automotive engineering perspective, the implications are profound. As OEMs push toward higher power density, longer range, and faster charging, every percentage point of efficiency matters. The ability to dynamically tailor the DC bus voltage—not just for peak efficiency but also with minimal control complexity and excellent transient behavior—positions this hybrid control strategy as a compelling candidate for next-generation EV platforms. It also aligns with industry trends toward software-defined power electronics, where advanced algorithms replace bulky hardware to achieve performance gains.
Furthermore, the solution’s compatibility with standard silicon MOSFETs and modest computational requirements enhances its commercial viability. Unlike approaches relying on exotic wide-bandgap devices or high-frequency switching that demands specialized magnetics, MPC-CSO operates effectively at 20 kHz using off-the-shelf components, easing integration into existing manufacturing ecosystems.
Looking ahead, the researchers suggest that this framework could be extended to other bidirectional converter topologies or integrated with higher-level vehicle energy management systems. For instance, coupling the DAB controller with predictive navigation data could enable anticipatory voltage adjustments before approaching hills or traffic lights, further optimizing energy use.
In an era where EV innovation is increasingly driven by intelligent control rather than brute-force hardware upgrades, the work by Guan Weide and colleagues exemplifies the power of algorithmic elegance. By harmonizing predictive control with physical insight into loss mechanisms, they’ve delivered a solution that is not only technically superior but also practical, robust, and ready for the road.
Authors: Guan Weide, Li Tao, Zhong Jian, Wang Xuhong, Xia Xiangyang
Affiliation: School of Electrical and Information Engineering, Changsha University of Science and Technology, Changsha 410114, China
Journal: Transactions of China Electrotechnical Society
DOI: 10.19595/j.cnki.1000-6753.tces.230590