Dynamic response improvement for multi-terminal DC system with AI-designed adaptive dynamic reference control

Abstract

The significant penetration of renewable energy sources (RES) makes the AC/DC hybrid system a typical low-inertia and poor-damping system. The multi-terminal direct current system (MTDC) is critical in connecting multiple AC systems and transferring bulk power among different AC regions. Conventional control often involves a tradeoff between response time and damping oscillations. To solve these issues, an AI-designed adaptive dynamic reference (ADR) control is investigated in this paper. It is designed with a controllable settling time and is able to mitigate the power oscillation during step change of power reference. In detail, an ADR module is utilized to generate the reference signal, which is input to dual closed-loop control. Raw data is generated from the simulation to build a data-driven surrogate model to map ADR parameters directly to settling time and overshoot in dynamic response. Once the surrogate model is trained, it is able to evaluate VSC’s dynamic performance within 0.003s, which is much faster in comparison with online simulation. A meta-heuristic optimization method is then adopted to find the optimal control parameters based on the surrogate model. The effectiveness of the proposed ADR control and its AI-aided parameter design is validated by the real-time digital simulation and hardware-in-loop experiment.

Keywords

• Adaptive dynamic reference
• Dynamic response
• Parameter design
• Voltage source converter