Network-based voltage stability analysis of power systems

Abstract

Voltage stability has attracted increasing attention since widespread blackouts in the 1970's and 1980's related to voltage issues. Nevertheless, most existing studies focus on generator and load dynamics without explicitly considering the impact of network structures which have recently been recognized to influence system stability greatly. Especially with the growing penetration of renewables and increasing dynamic loads, along with the tendency for power system structures to become increasingly complex, the network structure is expected to become more significant. Therefore, it is critical to analyze voltage stability from the network-based view.

This thesis investigates the impact of network structures on voltage stability from different viewpoints and explores the application. First, the influence of network structures on long-term voltage stability (LTVS) considering loads with recovery dynamics is studied theoretically. A system model with load dynamics is established in terms of a network graph and the aggregate nonlinear recovery load model. Stability conditions are derived for static analysis of voltage stability and small-disturbance voltage stability (SDVS), which reveal the interactions between network graph, load demand and load characteristics and shed new light on voltage stability mechanisms. The impact of the penetration of RESs is also discussed.

Second, based on the system model mentioned above, SDVS is investigated further. The weighted load connectivity (WLC) is defined to indicate the stability of network structures. To avoid the impact of uncertain recovery time of loads, robust stability assessment is applied. The Monte Carlo method is used to study the relationship between WLC and SDVS for a broad range of network topologies and parameters. It is discovered from the simulation results that the network with a larger value of WLC has a stronger ability to maintain SDVS. To study the influence of RESs, the system model is improved by connecting an inverter-based distributed generator in parallel with each aggregated nonlinear dynamic load.

Third, the impact of network structures on short-term voltage stability (STVS) is explored using data-driven methods. An integrated graph metric set (IGMS) is proposed to characterize the network structure. The transient voltage severity index (TVSI) is applied to quantify the STVS performance. Then based on artificial neural networks (ANNs), a two-stage ANN-based probabilistic prediction (TSAPP) method is devised to establish the mapping between STVS and network structures.

Finally, an efficient data-driven method for distribution network reconfiguration (DNR) considering STVS is developed. An STVS evaluation network is customized from deep convolution neural networks (CNNs) and trained to learn the relationship between network structure and STVS performance. Then the well-trained STVS evaluation network is applied in DNR to find the optimal topology which is cost-effective and has a stronger ability to maintain STVS. The STVS of distribution systems with various network topologies is evaluated without resorting to time-domain simulations, and topologies with better STVS performances are integrated as a constraint into the DNR problem.

In conclusion, this thesis provides novel insights into network-based voltage stability analysis of power systems and investigates STVS enhancement in DNR.