Capacitor Voltage Measurement Using Artificial Neural Networks for Flying Capacitor Multilevel Converter

The flying capacitor multilevel (FCML) topology can apply low-cost, low-power semiconductor switches to higher-power applications and achieve better performance indicators. However, compared with the traditional two-level topology, the FCML topology requires additional adjustment of the flying capacitor voltage, which increases the cost and volume. This paper studies the sensorless capacitor voltage measurement technology of the FCML topology and proposes a capacitor voltage measurement based on an artificial neural network (ANN). This method first collects the operating data of the multi-level converter and then trains the neural network offline to estimate the voltage value of the flying capacitor online. The method proposed in this paper is combined with the model predictive control technology to show a good estimation effect on the capacitor voltage.