Event-triggered dynamic state estimation based on set membership filtering for power systems

Thumbnail Image
Boragolla, B. N. Weerasinghe Mudiyanselage Rashmi Amadini
Journal Title
Journal ISSN
Volume Title
In the future, dynamic state estimation (DSE) will be an important function in monitoring and control of smart power grids. In this context, data communication networks and phasor measurement units (PMU) that are currently being deployed in smart power grids, will play a major role. However, very high data rates of PMUs will necessitate event-triggered multi-area state estimation (MASE) to ease the burden on the communication networks. In MASE, a large power grid is divided into sectors and local state estimates of each sector are communicated to a monitoring center for fusion and decision making. Traditional DSE algorithms such as Kalman filter (KF) are not well suited for event-triggered state estimation and new approaches will be required. The goal of this thesis is to investigate the applicability of a lesser-known class of algorithms known as set membership (SM) filtering to MASE. These algorithms have the important property known as data selective update. In the context of MASE, this property will allow the communication of sector-based estimates to the fusion center, only when the measurements observed by sensors within a sector are informative, that is, indicative of the existence of an abnormality such as a fault condition. The contribution of this thesis consists of two parts. In the first part, a simple to implement SM algorithm incorporating data-selective updates is presented in detailed, and it's properties are investigated through a numerical study. It is shown that, despite sparse updates, the estimation accuracy of the SM algorithm is comparable to the traditional extended KF (EKF) algorithm. In the second part, a comprehensive case study involving event-triggered state estimation in a single machine infinite bus system with a synchronous machine is presented. The simulation results show that, except during transient and fault conditions, the SM algorithm presented in this thesis rarely performs complete state updates, thus saving communication burden in a MASE context. The estimation accuracy, however, remains comparable with the EKF. As an important avenue for future research, methods of robust initialization of the SM algorithm is identified. It is shown that improper initialization can affect the ability of the SM algorithm to respond to fault conditions.
Event-triggered state estimation,Dynamic state estimation,Power systems,Set membership filter