Porsche Engineering tests AI-based soft switching for EV inverters

Porsche Engineering reports on AI-controlled soft switching that cuts inverter losses and improves electric vehicle efficiency. Learn how the system works.

Porsche Engineering is turning to artificial intelligence to improve the efficiency of electric vehicles, starting with one of the most critical yet often underestimated components of the powertrain: the inverter. A significant share of energy losses in electric drivetrains occurs during the switching of power transistors, directly affecting range, thermal load, and component size.

The new approach is based on the principle of soft switching. Unlike conventional hard switching, where current and voltage overlap during transitions, soft switching aims to minimize their product by carefully controlling the moments when transistors are turned on and off. In practice, this means switching occurs when voltage or current is close to zero.

Porsche Engineering has opted for Zero Voltage Switching (ZVS), a method particularly well suited to inductive loads such as electric motors. ZVS also aligns well with modern silicon-carbide and gallium-nitride power transistors, which are increasingly used in electric vehicles for their ability to operate efficiently at higher switching frequencies.

At the heart of the concept lies the Auxiliary Resonant Commutated Pole (ARCP) inverter topology. While ARCP has been known in power electronics for decades, its use in traction inverters has traditionally been limited by the complexity of controlling it under rapidly changing operating conditions. This is where artificial intelligence comes into play.

A pre-trained AI algorithm processes dozens of live vehicle parameters in real time, including load, torque, and temperature. Based on this data, it calculates optimal switching instants for the power transistors within fractions of a second. Porsche Engineering is currently evaluating both recurrent neural networks, known for high predictive accuracy, and reinforcement learning methods, which offer advantages for demanding real-time applications.

Simulation results indicate a substantial efficiency gain. Switching losses in the power transistors can be reduced by 70 to 95 percent. Depending on driving conditions, this translates into a noticeable increase in vehicle range, while also lowering heat generation in the inverter. Reduced thermal stress allows for smaller cooling systems and more compact inverter designs, with overall volume reductions of 20 to 50 percent. The gentler switching behavior also eases the load on power transistors, potentially extending their service life.

The AI-based control algorithm has already reached an advanced stage of development. Once finalized, Porsche Engineering intends to offer the technology as a software-based solution. Delivered as software libraries, it can be integrated into existing control units with relatively minor hardware modifications, making it suitable both for model updates and for entirely new electric vehicle platforms.