New Control Strategy Boosts Performance of High-Speed Permanent Magnet Motors in EVs
In the rapidly evolving world of electric vehicles (EVs), the quest for higher efficiency, greater power density, and extended driving range continues to push the boundaries of motor control technology. A recent breakthrough from researchers at Hefei University of Technology promises to significantly enhance the performance of interior permanent magnet synchronous motors (IPMSMs)—a key component in modern EV drivetrains—especially under high-speed, low-carrier-ratio operating conditions that have long posed formidable challenges for digital control systems.
Led by Yupu Zhu, Shuying Yang, and Qishuai Wang from the School of Electrical Engineering and Automation, the team has developed a novel current control strategy based on a delay-corrected extended state observer (ESO). Their work, published in the March 2024 issue of Transactions of China Electrotechnical Society, addresses a critical yet often overlooked issue in high-speed motor drives: digital control delay.
As EV manufacturers strive to increase motor speeds to achieve higher power density without proportionally increasing system size or weight, they encounter a fundamental limitation: the switching frequency of power inverters cannot scale indefinitely due to thermal constraints, semiconductor losses, and electromagnetic interference. This results in a shrinking ratio between the inverter’s switching frequency and the motor’s fundamental electrical frequency—a condition known as “low carrier ratio.” Under such conditions, conventional vector control strategies, typically designed in the continuous domain and later discretized for digital implementation, suffer from degraded performance due to increased discretization errors, exacerbated cross-coupling between d- and q-axis currents, and—most critically—control delays inherent in digital systems.
These delays, stemming from the time required for current sampling, computation, and pulse-width modulation (PWM) signal generation, cause a mismatch between the control action and the actual motor response. In practical terms, the voltage command applied at a given control cycle influences the motor current only in the subsequent cycle. For high-speed operations where electrical periods are extremely short, even a single sampling delay can introduce significant phase lag, destabilizing the current loop and reducing torque accuracy.
Traditional proportional-integral (PI) controllers, while simple and widely adopted, lack the robustness needed to handle these dynamic distortions. Advanced methods such as feedforward decoupling or complex vector control have been proposed, but they often assume idealized models and fail to account for real-world uncertainties like parameter variations due to temperature drift or magnetic saturation, as well as external disturbances from load fluctuations or grid harmonics.
Enter the extended state observer—a concept pioneered by Jingqing Han in the 1990s as part of active disturbance rejection control (ADRC). The ESO treats all unmodeled dynamics, parameter uncertainties, and external disturbances as a single “total disturbance” and estimates it in real time alongside the system states. This estimated disturbance can then be fed forward into the control law for compensation, effectively turning an uncertain system into a well-behaved, nominal one.
However, applying ESO in digital motor drives is not straightforward. The standard ESO assumes that its two inputs—the control voltage and the measured current—are synchronized in time. But under digital control with inherent one-step delay, the measured current corresponds to the previous voltage command, not the current one. This temporal misalignment corrupts the observer’s estimation, particularly in low-carrier-ratio scenarios where the delay constitutes a large fraction of the electrical cycle.
Recognizing this, the Hefei University team conducted a rigorous analysis in the discrete-time domain and proposed several delay-aware ESO architectures. They evaluated three distinct approaches: a model-based delay ESO (M-DESO) that approximates the delay as a first-order inertial element using Padé approximation; a Smith predictor-based ESO (Smith-DESO) that forecasts the current one step ahead to align it with the present voltage command; and a voltage-delay ESO (Ud-DESO) that intentionally delays the voltage input by one sample to match the delayed current measurement.
Through extensive simulations and experimental validation on a real EV drive platform, the researchers demonstrated that both Smith-DESO and Ud-DESO dramatically outperform conventional ESO and M-DESO in terms of estimation accuracy, current tracking speed, and disturbance rejection. Notably, the Smith-DESO achieved the fastest response, reducing disturbance recovery time from over 50 milliseconds under traditional PI control to just 2.5 milliseconds. This translates to smoother torque delivery, reduced current ripple, and improved efficiency—critical advantages for high-performance EVs.
Yet, the study also revealed a crucial trade-off. While Smith-DESO delivers superior dynamic performance when motor parameters are accurate, it exhibits heightened sensitivity to parameter mismatches—such as errors in inductance values due to magnetic saturation or temperature changes. In contrast, Ud-DESO, though slightly slower in response, demonstrates remarkable robustness even when the inductance used in the observer deviates by up to 70% from its true value. This resilience makes Ud-DESO particularly attractive for real-world applications where precise, real-time parameter identification remains challenging.
The experimental setup, built around a Texas Instruments TMS320F28379 digital signal controller and a 18-kW IPMSM rated for 7,500 rpm, confirmed these findings. At 3,000 rpm with a 4-kHz switching frequency (carrier ratio of ~13), all four strategies performed adequately. But as speed increased to 6,000 rpm (carrier ratio dropping to ~10), the conventional ESO began to show instability, while Smith-DESO maintained stable operation even at 800 Hz electrical frequency—equivalent to a carrier ratio of just 5.
This achievement is not merely academic. Low-carrier-ratio operation is increasingly common in next-generation EVs, especially in high-speed traction applications like performance sedans or commercial vehicles requiring wide constant-power speed ranges. By enabling stable, high-fidelity current control under these demanding conditions, the Hefei team’s strategy paves the way for smaller, lighter, and more efficient motor drives without requiring costly increases in switching frequency or computational hardware.
Moreover, the fully discrete-domain design philosophy eliminates the approximation errors introduced by continuous-to-discrete transformations, ensuring that theoretical performance aligns closely with real-world implementation. This approach also simplifies controller tuning, as the observer bandwidth becomes the primary design parameter, directly influencing both response speed and noise sensitivity.
From an industry perspective, the implications are profound. Enhanced current control directly translates to better torque precision, reduced acoustic noise, lower copper losses, and extended battery range. For automakers racing to meet stringent emissions targets and consumer expectations for driving dynamics, such incremental gains can be decisive.
The research also underscores a broader trend in power electronics: the shift from model-based control to model-agnostic, disturbance-rejection paradigms. As systems grow more complex and operating conditions more variable, reliance on precise mathematical models becomes impractical. Observers like ESO offer a pragmatic alternative—leveraging real-time estimation to compensate for uncertainty rather than attempting to eliminate it through ever-more-complex modeling.
Looking ahead, the team suggests several avenues for further development. Integrating the delay-corrected ESO with predictive control or machine learning-based parameter adaptation could yield even greater robustness. Additionally, extending the framework to sensorless operation—where rotor position is also estimated—could reduce system cost and improve reliability by eliminating position encoders.
In an era where every watt-hour counts and every millisecond of response time matters, innovations like this exemplify the quiet but critical engineering advances powering the EV revolution. While headlines often focus on battery breakthroughs or autonomous driving, it is often in the domain of motor control—where physics, computation, and real-time decision-making converge—that the true performance envelope of an electric vehicle is defined.
The work by Zhu, Yang, and Wang not only solves a pressing technical challenge but also sets a new benchmark for robustness and efficiency in high-speed motor drives. As the automotive industry accelerates toward electrification, such foundational research will be instrumental in turning the promise of sustainable mobility into a high-performance reality.
Authors: Yupu Zhu, Shuying Yang, Qishuai Wang — School of Electrical Engineering and Automation, Hefei University of Technology, Hefei 230009, China
Published in: Transactions of China Electrotechnical Society, Vol. 39, No. 6, March 2024
DOI: 10.19595/j.cnki.1000-6753.tces.222286