Linearized and distributed methods for power flow analysis and control in smart grids and microgrids

Optimization and control is a core part of Energy Management System (EMS), which receives the data from supervisory control and data acquisition (SCADA) system, analyzes the data centrally and provides decision actions to the system operators. Unlike conventional power system, smart grid and microgrid are more complex, dynamic and flexible, which requires a high level of computational intelligence, speed and flexibility. This dissertation presents a linearized model to analyze the optimization problem of smart grid, which can provide speed advantage as well as enough accuracy. In the meantime, a distributed multi-agent based control system is proposed in this dissertation for a flexible control of microgrids.;With the development of smart grid, power electronic control devices, such as Flexible AC Transmission Systems (FACTS), are introduced into power system. They can help system operators adjust real and reactive power flows, provide voltage support or regulate voltage. The way to determine the optimal size and location to install FACTS devices is a nonlinear optimization problem. Various nonlinear techniques have been proposed and developed to solve this optimization model, such as descent methods, Newton's methods, gradient projection methods, interior methods and so on. These nonlinear methods, to some extent, can provide accurate optimal solutions; however, they are usually computationally expensive when dealing with large power systems with tens of thousands of buses. And this computational speed sometimes cannot satisfy system operators' requirements. Therefore, many industrial applications have utilized a DC optimal power flow model which assumes a flat voltage magnitude over the system. This model can achieve the results very fast, but it sacrifices accuracy and reactive power information.;To reach a better trade-off between accuracy and speed, in the first half of this dissertation, it proposed a linearized power flow model for studying benefit of FACTS devices. This linear model can achieve better accuracy than DC power flow model and maintain reactive power information while the computational speed is not sacrificed.;In the meantime, the increasing penetration of renewable energy and its potential accommodation paradigm, microgrids, restrict traditional central control structure in terms of cost, flexibility and reliability. Distributed control is able to address these challenges in three aspects: more economic efficiency by utilizing low-cost devices; more flexibility in terms of time-varying and adaptive configurations or functions; and more robustness by continuing working in the presence of single-point failure.;For the power balance control, this dissertation first proposed a distributed multi-agent system without considering network losses and voltage regulation. In this proposed system, the information flows in parallel and results are obtained in a non-iterative way; therefore, this method achieves superior performance in terms of speed without any convergence issues. In the case where power losses are considerable and voltage regulation is expected, this dissertation proposed a distributed multi-agent control for power balance based on Guess method. The proposed power flow algorithm fully makes use of communication time, and updates state information synchronously among agents. Therefore it can also provide speed advantage over asynchronous methods.;For economic dispatch, this dissertation first proposed a distributed algorithm for microgrids without considering network constraints. This proposed economic dispatch is merit for fast convergence and is applicable for on-line control. In order to consider the network constraints, this dissertation presented a distributed multi-agent system which can consider the network constraints with full or partial observation of system state information. The proof of convergence is also presented in the dissertation.