Binary Neural Networks For Modular Multilevel Converters - 2023/2026

This study explores the use of Binary Neural Networks (BNNs) for controlling Modular Multilevel Converters (MMCs), which are key components in advanced power systems such as HVDC grids and renewable energy infrastructures. By employing binary weights and activations, BNNs offer a computationally efficient implementation, ideal for embedded systems with limited resources. When applied to MMC control, they enable fast and precise decisions regarding submodule switching, improving voltage stability and reducing harmonic distortion. Additionally, their lightweight nature allows direct hardware integration, facilitating real-time control.

The study demonstrates that BNNs can achieve performance comparable to traditional neural networks at a fraction of the computational cost. This integration promotes faster, more robust, and efficient control solutions, supporting the modernization of electrical infrastructure. It marks a step forward toward smarter and more sustainable converters for the power grid of the future.

Program: Joint Research Project Chile–Germany

Project Duration: 36 months

Principal Investigator: Félix Eduardo Rojas Lobos