Energy-efficient predictive control strategy boosts performance of electric vehicle drives

A research team at the National Institute of Technology Puducherry (NIT Puducherry) has unveiled an energy-efficient model predictive current control (MPCC) method that enhances the performance and efficiency of induction motor (IM) drives in electric vehicles (EVs).

The new approach, called Centroid-Synthesized Virtual Voltage Vector (CSVVV)-based MPCC, introduces a novel way to synthesize voltage vectors within each control cycle — significantly reducing switching losses and current ripple, which are among the key limitations of conventional MPCC techniques.

Addressing the limits of conventional predictive control

Induction motors are widely used in EVs for their robustness and cost advantages, but their control systems often face a trade-off between fast torque response and energy efficiency. Conventional finite-control-set model predictive current control (FCS-MPCC) predicts the future current trajectory for each possible switching state and selects the one minimizing a cost function. While this ensures fast dynamics, it frequently causes high inverter switching frequency, elevated losses, and current distortions under varying load conditions.

To overcome these drawbacks, the NIT Puducherry researchers developed a centroid-based vector synthesis method that integrates two adjacent active voltage vectors and a zero vector to generate a virtual reference vector located near the centroid of the feasible switching region in the α–β reference frame. This synthetic approach enables the controller to approximate the desired stator voltage vector more accurately within a single sampling period, thereby reducing current tracking error without frequent switching.

How the new control method works

In the proposed CSVVV-MPCC algorithm, the prediction model first computes the stator current behaviour for all candidate voltage vectors using the discrete-time motor equations. Instead of directly choosing one optimal vector, the controller computes the weighted time durations (t₁, t₂, t₀) for the two nearest active vectors and the zero vector, forming a virtual centroid vector that more closely aligns with the reference voltage vector demanded by the torque and flux control objectives.

This multi-vector synthesis effectively bridges the gap between traditional FCS-MPCC and continuous control methods such as space vector modulation (SVM) — delivering smoother current waveforms while maintaining the predictive nature of the control system. The optimization is performed in real time with minimal computation, allowing seamless integration with digital controllers such as TI F28379D.

Experimental validation

To validate the proposed technique, experiments were conducted on a 2.2 kW induction motor drive platform, where the control algorithm was implemented on the TMS320F28379D DSP. The performance was compared with conventional MPCC under identical test conditions.

Results revealed that the CSVVV-MPCC method achieved:

• A 25–30% reduction in inverter switching losses
• A significant decrease in current total harmonic distortion (THD)
• Improved steady-state torque smoothness
• Faster transient response under step torque and speed variations

The control scheme-maintained stability and accuracy across wide operating ranges and dynamic load profiles, demonstrating its potential for real-time electric vehicle applications.

Broader significance

The centroid-synthesized virtual vector approach offers a computationally efficient and sensorless solution adaptable to both induction and permanent magnet motor drives. By lowering energy losses and reducing thermal stress on inverter switches, the technique extends the operational life of the power converter — a vital consideration for EV manufacturers targeting higher efficiency and durability.

Beyond electric mobility, this control strategy can also benefit renewable energy systems, industrial drives, and robotic actuators, where energy optimization and precise torque control are equally critical. “Our method bridges the gap between predictive and modulated control strategies,” said one of the researchers from NIT Puducherry. “By synthesizing a centroid-based virtual voltage vector, we’ve demonstrated that it’s possible to retain the fast dynamics of MPCC while drastically improving its efficiency — a key requirement for future EV drive systems.”

The full study is accessible by DOI: 10.30941/CESTEMS.2025.00027