Modeling, analysis and mitigation of the small-signal instability in future power systems with renewables integration via AC/DC connections

Abstract

Renewable energy exploitation induces new power system stability issues caused by interactions between inverter-based resources (IBRs) and AC/DC electrical connections. Confirmed in many real-world incidents, doubly-fed induction generator-based wind parks (DFIG-WPs) can interact with both series-capacitor compensated and weak (low short circuit ratio) AC grids, and full-size converter-based WPs (FSC-WPs) can interact with weak AC grid. To mitigate instabilities of two mainstream WPs with AC connections, this thesis proposes a traditional model-driven method and a modern data-driven method. The traditional model-driven method is based on impedance-based stability assessment (IBSA) in the frequency domain, which starts from obtaining impedance models of IBRs and converter stations. Analytical impedance models of grid-connected VSC, DFIG and FSC-based wind turbine (FSC-WT) and the WP side modular multi-level converter (WPMMC) are established considering the AC-DC coupling phenomenon and the MMC multi-harmonic property that are partially ignored in existing works. Frequency scanning is then conducted in electro-magnetic (EMT) simulations to validate the accuracy of the proposed analytical impedance models. The instability of DFIG-WPs takes place in the sub-synchronous range in a series-capacitor compensated grid and is mainly affected by the rotor-side converter (RSC) control parameters. However, the weak grid instability is found in the super-synchronous range with its mirror frequency in the sub-synchronous range and is affected by both RSC and grid-side converter (GSC) control parameters. Recommendations are given for instability prevention and/or mitigation through DFIG control parameter modification based on the guidance of the stability contour analysis. The weak grid instability of FSC-WPs also takes place in the super-synchronous range rather than sub-synchronous range and may occur even above the double fundamental frequency under certain parameter conditions. Hence, there is a need to reconsider the classification of weak grid instability in both DFIG-WPs and FSC-WPs. In addition, FSC­-WT control parameters are adjusted for instability mitigation according to the impacts on resonance frequency and stability margin. The thesis also investigates the difference and boundary of the weak grid instability phenomenon between RMS and EMT simulations to prepare for the machine learning applications. The promising and effective solution for instability mitigation is the supplementary damping controller (SDC). However, most SDCs cannot handle large resonance frequency shifts caused by external grid condition changes. Deep reinforcement learning-based agent (DRL-Agent) is utilized to design an instability-immune SDC adaptive to shifted resonance frequencies in weak grid instability. This thesis coordinates the fast root mean square (RMS) and accurate EMT simulations as a computationally efficient training and testing scheme. The efficacy of the trained DRL-Agent to mitigate weak grid instability is tested in unseen scenarios out of the training dataset. The performance of the DRL-Agent is then improved step-by-step through modifying the reward function and hyper-parameters. In future power systems, various types of HVDC converters, IBRs, and conventional power plants integration through AC/DC connections will bring challenges with complex, highly nonlinear, and time-varying properties. A combination of both model-driven and data-driven methods would provide explainable and trustable solutions.