Multiloop PI Controller Optimization and Low-Frequency Oscillation Analysis in VSC-HVDC Systems

Abstract

Voltage-source-converter-based high-voltage direct-current (VSC-HVDC) transmission is increasingly used for renewable-energy integration, asynchronous-grid interconnection, and dynamic support of weak AC networks. However, fast converter control dynamics, reduced system inertia, and interaction between the inner and outer control loops can influence electromechanical damping and may intensify low-frequency oscillations (LFOs) under weak-grid operating conditions. This paper presents an integrated study of multiloop proportional–integral (PI) controller optimization and measurement-based LFO analysis for a two-terminal VSC-HVDC system. A detailed small-signal model was developed in the synchronous $D$ – $Q$ reference frame, including converter current dynamics, DC-link dynamics, and the associated outer control loops for active-power/DC-voltage and reactive-power/AC-voltage regulation. The PI controller gains are tuned using a multiloop optimization framework such that a common set of gains provides acceptable damping over different rectifier/inverter operating modes and AC-system strengths. The proposed control study was complemented by a Phasor Measurement Unit(PMU) / Wide Area Monitoring System (WAMS)oriented oscillation-monitoring framework. Because measurement noise can distort modal estimates, the statistical nature of the measured signal is first examined using empirical cumulative distribution function (ECDF), theoretical Gaussian cumulative distribution function (CDF)and quantile–quantile (Q–Q) analysis. After verifying the noise characteristics, wavelet-based denoising using a Daubechies wavelet was applied before the modal estimation. The dominant oscillatory modes are then extracted using a modified Prony method and compared with the subspace-based data-driven identification. Simulation studies under different effective short-circuit ratio (ESCR) values demonstrate that coordinated multiloop PI tuning improves damping the performance, whereas the proposed preprocessing and modal-identification chain improves the reliability of LFO estimation from noisy PMU-like measurements.