Construction of generic and adaptive restoration strategies

Abstract

Electric power systems are among the most critical infrastructures of modern societies. The interruption of electric supply may cause great adverse impact on the economy of modern societies. For the sustainable development of modern societies, power systems have been undergoing a paradigm shift towards more environmentally-friendly, economic, reliable and resilient modern power grids. Power system restoration has been identified as one of the most important enabling technologies to achieve one of the aforementioned desirable features, i.e., resilience, of modern power grids.

The contribution of this thesis is to apply both new types of components and the state-of-the-art optimization/computation technologies to establish restoration strategies, considering a large spectrum of technical constraints in power systems. This thesis extends the methodology entitled “Generic Restoration Milestones” (GRMs) by implementing the following milestones, including Serve Load in Area, Synchronize Electrical Islands, Establish Transmission Grid, and Form Black_Start_Non_Black_Start Building Blocks. The implementation of each milestone is based on configurable optimization models considering new types of components. Thus, the restoration strategies obtained by these models are generic for various power system characteristics, and adaptive to actual outage scenarios.

First, a methodology to establish and validate a complete load restoration strategy is proposed. This methodology models the load restoration as a multi-stage decision-making process considering the operational constraints of generating units. At each stage, a mixed-integer nonlinear load restoration model (MINLR) is formulated to maximize load pickup subject to comprehensive realistic constraints. A complete load restoration strategy is obtained by solving a number of MINLR models in series. A branch-and-cut (B&C) solver is constructed to solve MINLR models efficiently. The theoretical basis of this B&C solver, including the convexification of power flow equations and the applicability of cutting planes, is established. The optimal generation dispatch is then modelled as a dynamic optimal power flow (DOPF) model and solved by a direct solution method based on the interior point algorithm. The frequency stability and transient stability of the obtained load restoration strategy is validated by an efficient time-domain simulators based on Graphics Processing Units.

Second, a novel methodology entitled “virtual synchroscope” is proposed. Assuming a full observability obtained from Phasor Measurement Units (PMUs), the proposed methodology compares voltage phasors at the synchronizing points of two electrical islands with the accurately time-stamped measurements. Thus, a smooth synchronization can be implemented. A sufficient condition for full observability of static state estimation with PMUs is provided.

Finally, the application of fast cut back (FCB) thermal units and high-voltage direct-current (HVDC) transmission systems in power system restoration is investigated. The benefits of FCB units as blackstart sources are quantified by a bi-level optimization algorithm to minimize the blackstart duration. A non-divergent power flow model with HVDC systems based on nonlinear programming is established to quantify the feasibility of using HVDC links in power system restoration.

In summary, this thesis has developed anarray of optimization models and efficient computational methods to establish generic and adaptive restoration strategies. The proposed models and methods can be extended for online applications to realize resilient power grids.