Distributed Neural Network Observer for Submodule Capacitor Voltage Estimation in Modular Multilevel Converters

Modular multilevel converters (MMCs) have become one of the most popular power converters for medium/high-power transmission systems and motor drive applications. Standard control schemes for MMCs use a voltage measurement per submodule (SM) to balance the capacitor voltages and govern the MMC. Consequently, the control system requires a significant amount of sensors and the effective communication of sensitive data under relevant electromagnetic interference (EMI), impacting the reliability and cost of the MMC. This work presents a distributed neural network (DNN) observer inspired by a general predictor-corrector structure for estimating the capacitor voltages at each SM. The proposed observer predicts each SM capacitor voltage using a standard average model. Then, each prediction is corrected and denoised by a neural network of reduced computational complexity. As a result, the proposed observer reduces the number of required voltage sensors per arm to only one and filters the high-frequency noise without noticeable delay in the estimated SM capacitor voltages for both transient and steady-state operations. Experiments conducted in a three-phase MMC with 24 SMs confirm the effectiveness of the proposed DNN observer.

Automated summary

Modular multilevel converters (MMCs) are popular power converters for medium/high-power transmission systems and motor drive applications. However, the standard control schemes for MMCs require a large number of sensors and the communication of sensitive data, which can be affected by electromagnetic interference (EMI), thus impacting the reliability and cost of the MMC. This paper proposes a distributed neural network (DNN) observer to estimate the capacitor voltages at each submodule (SM) of the MMC, reducing the number of required sensors and improving performance through noise filtering.