Abstract
In recent years, due to the enormous enhancements in distributed computing systems, the application of distributed algorithms for control and optimization of Microgrids and large-scale power grids has been proposed. In particular, various consensus-based control strategies have been integrated into Microgrid controllers aiming to improve their dynamic performance and provide accurate active and reactive power sharing among interface converters. Similarly, several distributed optimization techniques have been implemented and tested for real-time assessment and monitoring in power transmission systems. In this sense, this thesis explores innovative distributed strategies for future power systems, including small scale generation systems as well as modernized large-scale power grids. We explore and implement two distributed techniques: i) a novel distributed control approach for enhancing the operation of isolated droop-controlled Microgrids, and ii) a robust distributed optimization algorithm applied to probabilistic available transfer capability assessment of power systems with penetration of renewable energy. First, we focus on distributed control strategies for droop-controlled Microgrids and introduce a novel approach: distributed pinning droop control. The proposed approach seeks to enhance the overall operation of the Microgrid by providing accurate active and reactive power sharing among parallel power converters. The distributed pinning droop control technique can be considered as a special leader-following consensus algorithm integrated into the traditional droop control--- used as a distributed methodology for accurate active and reactive power sharing in isolated converter-fed Microgrids. Under this proposed framework, both power converters and loads in the Microgrid are treated as intelligent agents with communication capabilities inside a multi-agent system. A great advantage of this proposed framework is that, since only a fraction of the agents in the network have access to the control reference, the required communication bandwidth and control cost can be significantly reduced without degrading the dynamical performance compared to previously proposed consensus-based droop control techniques. The proposed distributed pinning droop control mechanism is studied in a rigorous theoretical and experimental manner, including convergence criteria, the appropriate selection of the leader agents given arbitrary communication topologies and the implementation using standard multi-agent programming environments. In order to validate the correctness and applicability of the distributed pinning droop control framework, a multi-agent platform is developed for a test-bed Microgrid with multiple interface converters and loads. The test-bed system is located in northern Taiwan and the simulation results are conducted using a precise dynamical model of the system, including actual parameters of transmission lines, converter capacity and load requirements. Extensive numerical simulations under various operating conditions, including normal operations, heavy loading conditions, and plug-in and plug-out scenarios, are investigated using the SimPowerSystems toolbox and the Java Agent Development Framework. Second, we investigate distributed optimization algorithms and their application to on-line monitoring of Available Transfer Capability in transmission systems. In particular, we explore the Optimality Conditions Decomposition algorithm for probabilistic Available Transfer Capability assessment in power systems with penetration of intermittent wind power generation. First, the Available Transfer Capability assessment is mathematically formulated as a non-linear optimal power flow problem. Then, an iterative decomposition-coordination methodology based on Optimality Conditions Decomposition techniques is implemented for distributed probabilistic Available Transfer Capability assessment. In order to estimate the probability density function and empirical cumulative density function of the Available Transfer Capability, the Latin Hypercube Sampling method is utilized for obtaining samples of the integrated wind power sources. These obtained wind power samples are appended into the proposed decomposition-coordination approach in order to compute an accelerated Monte Carlo simulation of the Available Transfer Capability at the current system state. In order to provide a preliminary approximation of the range of variation of the Available Transfer Capability, three distributed algorithms to estimate the average-case, best-case, and worst-case scenarios are studied and developed. The correctness and applicability of the proposed approach are evaluated by conducting extensive numerical simulations in the standard IEEE 118-bus system. Numerical simulations are developed using JuMP --- a Julia-based modeling language for mathematical optimization. Various operating conditions of the test-bed system are used in order to investigate the robustness of the proposed distributed probabilistic Available Transfer Capability assessment.