Performance Enhancement of HVDC-VSC Using FFNN-Based PI Control Under Hybrid Renewable Integration

Abstract

This study introduces an advanced utilization of artificial neural networks (ANNs) in the feed-forward control of high-voltage direct current voltage source converters (HVDC-VSC) connected to hybrid renewable energy systems (HRES). The inherent nonlinear characteristics of renewable sources, combined with the fast-switching dynamics of power electronic converters, create significant challenges for conventional controllers in ensuring stable grid operation and efficient energy transfer. To overcome these limitations, a feed-forward neural network (FFNN)-based scheme is developed to improve DC-link voltage regulation, optimize the coordination of active and reactive power, and enhance overall system stability. The proposed FFNN is trained offline with extensive datasets that reflect diverse operating scenarios, enabling it to capture the complex relationships among renewable generation, grid variations, and converter control requirements. By predicting control actions in advance, the method reduces dependence on traditional feedback loops, thereby achieving faster transient response and greater resilience to disturbances. Simulation outcomes confirm that the FFNN controller surpasses the conventional proportional–integral (PI) approach in terms of reduced settling time, minimized steady-state error, and improved power quality, which makes it a feasible solution for modern grid applications.