Network-based stability analysis of high renewable-penetrated power systems

Abstract

Stability issues for power systems with high penetration of renewable energy sources (RESs) are of great concern. These intermittent RESs have been found to pose challenges for the resulting low-inertia systems. However, most research to date has focused on detailed modelling of node dynamics, while the consideration of both node dynamics and network structures is still in its infancy. In this thesis, an in-depth network-based investigation on these new stability issues is conducted, and progress has been made in four key aspects. First, an investigation of small-disturbance stability (SDS) (considering both voltage and angle) is conducted for the feasibility of all converter-interfaced generation systems. An overall framework is first established for studying the influence of the network topology, inverter dynamics, and load dynamics on system SDS. Then, an algorithm for stability analysis that enables sensitivity calculation and parameter optimization with high computational efficiency is designed. The results show how parameter selection of dynamic variables plays a major role in SDS. Finally, a parameter adjustment algorithm for improving SDS is provided. Second, an investigation of transient stability is conducted, with a focus on cutset properties. Two indexes – the cutset index (CI) and the improved cutset index (ICI) – are investigated and compared for identifying the vulnerable cutset as well as estimating the critical energy for determining the stability region. Stability analysis is then explored with different RESs penetration levels from 0% to 100% with different network size, topology, and dynamics. Simulation results show that the increased penetration of RESs can affect system stability and the specific influence depends on the penetration levels, system structures, and the RESs replacement. Third, apart from an individual investigation of impact factors, studies of the small-disturbance angle stability (SDAS) of high renewable-penetrated power systems with multi-scale-free (MSF) properties are further conducted. Two types of parameter distribution are assessed and quantified, through which a relationship between SF network topology and SF parameters distribution is identified. The results show that network topology, system dynamic parameters, and their synergistic effects all matter in determining system SDAS. Guidance for coordination of various parameters to improve SDAS is also provided. Finally, alongside the stability assessment, a further stability prediction is explored based on topological features with XGBoost, a decision tree-based ensemble machine learning model. The XGBoost model is trained and pruned based only on topological features for SDS prediction. To quantify the influence of network topology on SDS, thirty topological features are preselected as potential factors. To sort and filter the feature importance, a state-of-the-art machine learning method, SHAP values, is applied. The well-trained and pruned model is finally tested on both typical large-scale complex networks and power system testing cases. All the prediction results present favourable efficiency and accuracy. In summary, the obtained results in this thesis pave the way for the network-based research of stability issues in the high renewable-penetrated power systems. The investigation also fills research gaps by combining modelling complexity and grid topology variation. Future research can focus on more detailed network-based modelling and parameter optimization.