Modified Nearest Level Control and Model Predictive Control on Compensation of the Number Rounding Algorithm and Sub-module Voltage for MMC-HVDC

Abstract

This paper proposes a modified nearest level control (NLC) method and a model predictive control (MPC) method with a reduced number of considered states for modular multilevel converter (MMC) to compare their effects on the compensation of the number rounding algorithm and the voltage sorting algorithm. The NLC method is suitable for MMC-HVDC with large number of sub-modules (SM). And the proposed MPC method for the MMC-HVDC system utilizes weighting factors to configure the cost function in combination of the DC side current control and the valve side converter voltage control with a reduced number of states. Number rounding method for the sub-modules with the capacitor voltage sorting algorithm are two methods which are normally considered for the modified NLC and MPC methods to track the referenced voltage in order to guarantee the voltage balancing effect among different capacitors of SMs. However, there are always two inherent types of errors bringing with the two mentioned methods: errors resulted from the number rounding algorithm and errors generated from the deviation of the capacitor voltage of SMs from the average capacitor voltage respectively, leading to a significant difference between the output arm voltage and the input reference voltage. In this paper, two compensated methods brought from modified NLC and MPC methods respectively are proposed to compare their effects on improving the performance to significantly dampen the deviation of the output converter voltage at valve side from the referenced one, as to decrease the harmonics to avoid the oscillation and even to optimize the converter losses.