Resilient Operation of Active Distribution Networks via Self-learning Smart Devices

This dissertation focuses on developing Artificial Intelligence (AI)-based and self-healing control techniques to enhance the resiliency of active distribution networks for upcoming power grid challenges. In the first stage of this work, a high bandwidth primary control layer is developed to achieve an ultra-fast predictive controlled dual active bridge converter interfaced grid-following inverter for voltage and frequency support. The primary control layer is developed by a novel model predictive self-healing control (MPSC) scheme. This control technique heals intrinsic drawbacks in commonly used control approaches by decreasing the potential errors in the control processes. However, the frequency restoration process needs more advanced techniques due to the high nonlinearity of the active distribution networks such as power electronic dominated grids (PEDG). Therefore, an artificial intelligence-based power reference correction (AI-PRC) mechanism is developed to address the shortcomings of frequency restoration of the state-of-the-art virtual synchronous generator (VSG)-based or droop-based grid following inverters (GFLIs) and grid forming inverters (GFMIs) via re-defining GFLI role at grid-edge. A detailed analytical validation is provided that shows control rules in PEDG intrinsically follow the underlying dynamic of the swing-based machines to extend its stability boundary. Considering this fact, comprehensive transient and steady state-based mathematical models are used for constructing the learning database of the proposed AI-PRC mechanism. Subsequently, a neural network is trained by Bayesian Regularization Algorithm (BRA) to realize the proposed AI-PRC for GFLIs. The proposed training approach can deal with all grid characteristics alterations and uncertainties. Thus, this approach incorporates all PEDG’s effective variables that shape its dynamic response during transient disturbances. Several simulations and experimental case studies were provided that evaluate the functionality of the proposed AI-PRC for GFLIs towards enhancing transient response and resiliency of PEDG. The provided evaluations demonstrate significant improvement in frequency restoration in response to transient disturbances. Moreover, the proposed control technique is exploited as a shadow controller in the case that the attacker aims to threaten the entire grid stability via stealthy attacks. Some stealthy attack scenarios are investigated on the 14-bus PEDG, and the results have proven the effectiveness of the proposed approach in fast supporting of the grid in the event of stealthy attacks, thus the grid resiliency is enhanced in this case as well. Due to the high importance of power grid resiliency, in the final stage of this work, an intrusion detection system (IDS) is developed to provide another layer of security that monitors grid dynamics and vital variables in other time scales. The groundwork of this technique is based on a load forecasting procedure that benefits from an artificial intelligence approach. In more details, an anomaly detection technique based on a condition monitoring vector and ultra-short demand forecasting is designed and developed for achieving the above-mentioned goals. The designed IDS is more robust against attack scenarios that could bypass other primary control layers. Thus, the proposed approach enables grid operators to take proper and prompt actions for providing a secure operation of the grid.