New Control Strategy Boosts EV DC-DC Converter Performance Under Dynamic Loads
In the rapidly evolving landscape of electric mobility, the quest for more efficient, responsive, and robust power electronics has never been more urgent. A recent breakthrough from a team of researchers at Beijing Information Science and Technology University, in collaboration with the People’s Public Security University of China and Beijing Nucleus Tongchuang Technology Co., Ltd., promises to significantly enhance the performance of bidirectional DC-DC converters—a critical component in electric vehicle (EV) low-voltage power systems.
Published in the Journal of Chongqing University of Technology (Natural Science), the study introduces a novel control methodology that merges model predictive control (MPC) with fuzzy logic to dynamically adjust weighting parameters in real time. This approach, termed Fuzzy-Varying Weight Model Predictive Control (F-VMPC), addresses long-standing challenges in maintaining stable and precise output voltage during abrupt load changes—common scenarios in real-world EV operation, such as during rapid acceleration, regenerative braking, or the simultaneous activation of multiple onboard electronics.
Traditional control strategies, including the widely used Proportional-Integral (PI) controllers and even conventional MPC implementations, often struggle to balance speed, stability, and overshoot when faced with dynamic load fluctuations. While PI controllers are simple and reliable, they tend to exhibit sluggish response times and significant voltage overshoot under sudden load shifts. Conventional MPC improves upon this by leveraging predictive models to anticipate system behavior, but its performance is heavily dependent on fixed weighting parameters in the cost function—parameters that are typically tuned offline and cannot adapt to real-time operating conditions.
The innovation presented by Ma Bin, Shi Yongle, Chai Haonan, Jiang Wenlong, and Chen Yong lies in recognizing that the optimal weighting between voltage tracking error and control effort is not static—it must evolve with the system’s state. By integrating a fuzzy inference system that continuously evaluates input voltage and the magnitude of output voltage deviation, the F-VMPC controller dynamically recalibrates the weight assigned to the control input (duty cycle adjustment) within the MPC optimization framework.
This real-time adaptation allows the converter to respond with unprecedented agility. In simulation studies conducted using MATLAB/Simulink, the F-VMPC controller demonstrated a remarkable 26.1% average reduction in voltage overshoot compared to standard MPC when subjected to step changes in load resistance—from 70 Ω down to 10 Ω and 40 Ω. More impressively, in one test case involving a drop from 70 Ω to 40 Ω, the F-VMPC system exhibited virtually no overshoot (just 0.01 V), whereas the PI controller produced a 0.4 V spike and the conventional MPC still showed a 0.16 V deviation.
Beyond overshoot suppression, the new method also shortens settling time. In multiple dynamic scenarios—including both load increases (loading) and decreases (unloading)—the F-VMPC consistently achieved target voltage stabilization in approximately 0.2 milliseconds, outperforming PI controllers that required up to 2.7 ms in extreme cases. Even compared to conventional MPC, which already offers faster response than PI, the F-VMPC shaved off critical fractions of a millisecond, translating to smoother power delivery and enhanced system reliability.
The researchers validated their approach not only through extensive simulations but also via a physical experimental platform built around an HM-cSPACE real-time control system. The hardware setup included a two-switch bidirectional DC-DC converter prototype controlled by an STM32F030K6T6 microcontroller, operating at a 200 kHz switching frequency. Real-world tests—such as stepping the load from 40 Ω to 30 Ω or 20 Ω—confirmed the simulation results: the F-VMPC controller reached the desired output voltage faster and with lower steady-state error than the PI benchmark.
This advancement carries significant implications for the broader EV ecosystem. Modern electric vehicles rely on a complex array of low-voltage systems—ranging from infotainment and lighting to advanced driver-assistance systems (ADAS) and electronic control units (ECUs)—all powered through a 12 V or 24 V auxiliary network. The DC-DC converter acts as the bridge between the high-voltage traction battery (typically 400 V or 800 V) and this low-voltage domain. Any instability in this conversion process can lead to voltage sags, brownouts, or transient spikes that may disrupt sensitive electronics or even trigger safety-critical shutdowns.
By ensuring tighter voltage regulation under dynamic conditions, the F-VMPC method enhances not only performance but also functional safety and user experience. For instance, during regenerative braking, when energy flows back from the motor to the battery through the bidirectional converter in boost mode, the system must manage sudden surges in current without destabilizing the low-voltage bus. Similarly, when multiple high-power accessories (like heated seats, air conditioning compressors, or radar sensors) activate simultaneously, the converter must respond instantly to prevent voltage droop.
Moreover, the proposed control strategy aligns with industry trends toward software-defined power electronics. As automotive manufacturers increasingly adopt over-the-air (OTA) updates and adaptive vehicle control systems, the ability to implement intelligent, self-tuning algorithms like F-VMPC becomes a strategic advantage. Unlike methods that require computationally intensive online optimization (such as particle swarm optimization, which the authors note can hinder real-time execution), the fuzzy-based weight adjustment in F-VMPC is lightweight and suitable for deployment on standard automotive-grade microcontrollers.
The research team’s choice of a Takagi-Sugeno fuzzy model further underscores their engineering pragmatism. This first-order fuzzy inference system provides a good balance between approximation accuracy and computational simplicity, making it well-suited for embedded control applications where processing resources and memory are constrained.
From a design perspective, the F-VMPC framework is also highly modular. The core MPC engine remains unchanged; only the weight parameter is modulated by the fuzzy layer. This means existing MPC-based converter designs could potentially be upgraded with minimal architectural overhaul—simply by integrating the fuzzy regulator as a supervisory module.
Looking ahead, the methodology could be extended beyond bidirectional DC-DC converters. The principle of adaptive weighting based on real-time system states is broadly applicable to other power electronic systems, including inverters for motor drives, onboard chargers, and even grid-tied renewable energy interfaces. In fact, the authors hint at this potential in their conclusion, noting that their approach provides a theoretical foundation for optimizing control in a variety of dynamic power conversion scenarios.
Industry experts anticipate that innovations like F-VMPC will play a pivotal role in next-generation EV architectures, particularly as vehicles become more electrified and software-centric. With global automakers racing to improve energy efficiency, extend driving range, and enhance reliability, even marginal gains in power electronics performance can yield substantial system-level benefits.
The publication of this work in the Journal of Chongqing University of Technology (Natural Science) underscores the growing contribution of Chinese research institutions to the global advancement of automotive electrification. The team’s multidisciplinary collaboration—spanning mechanical-electrical engineering, new energy vehicle laboratories, and public safety technology—reflects the integrative nature of modern transportation research.
As electric vehicles continue to dominate automotive roadmaps worldwide, the demand for smarter, more resilient power management solutions will only intensify. The F-VMPC control strategy represents a meaningful step forward in that journey—offering a practical, effective, and implementable solution to a persistent engineering challenge. For automotive engineers, power electronics designers, and EV developers, this research provides not just a new algorithm, but a new paradigm for thinking about adaptive control in dynamic energy systems.
Authors: Ma Bin¹,²,³, Shi Yongle¹, Chai Haonan⁴, Jiang Wenlong⁵, Chen Yong¹,²,³
Affiliations:
¹ School of Mechanical and Electrical Engineering, Beijing Information Science and Technology University, Beijing 100192, China
² Beijing Laboratory for New Energy Vehicle, Beijing 100192, China
³ Beijing Collaborative Innovation Center of Electrics, Beijing 100192, China
⁴ Beijing Nucleus Tongchuang Technology Co., Ltd., Beijing 100192, China
⁵ School of Public Order and Traffic Management, People’s Public Security University of China, Beijing 100038, China
Published in: Journal of Chongqing University of Technology (Natural Science), 2024, Vol. 38, No. 7, pp. 12–20
DOI: 10.3969/j.issn.1674-8425(z).2024.07.002