Resilience-aware management of active distribution networks

Abstract

The electric power system is one of the greatest engineering achievements in the history of mankind. Electricity is integral to every part of our economy and society. Therefore, it is essential to make the electricity grid more resilient in facing different extreme events and black outs. In this work, four problems have been investigated to study and improve the resiliency of distribution networks. The first one focuses on the problem of power line congestion which can negatively harm the economy and different equipment in the grid. Two neural network models are used to predict where the congestion might happen and what would be the cause of it. Using these predictions, the problem can be alleviated in time and the resiliency of the grid will be improved. The second problem discusses power management of the distribution network under the occurrence of an extreme event. The problem is formulated as a Markov Decision Process using different agents and is solved using two Reinforcement Learning algorithms, namely, Q-Learning and Value Iteration. This approach is then tested on a benchmark system and the results show a remarkable improvement of the resiliency. The third problem studies the stability and power sharing of parallel inverters in a multi inverter-fed system. A small signal model of the power controller is studied. Further, the system's nonlinear dynamic equations are derived using accurate mathematical models. The system model is then trimmed and linearized around its operating point and the system's control parameters are optimized using Grey Wolf Optimization. Finally, improvement of stability and power sharing are verified by running time domain simulations. The last problem investigates the optimal siting and sizing of energy storage systems in a multi-microgrid system to improve the resiliency. A two-stage optimization method is used to solve the nonlinear and non-convex problem. The first stage involves an Optimal Power Flow and the second stage uses Genetic Algorithm. Further an investment cost based on the sensitivity analysis is introduced to improve the resiliency even further. The effectiveness of this ESS placement is tested on a benchmark system and validated using a fault scenario.