ON LINE POWER SYSTEM STATE ESTIMATION

Abstract

The real time implementation of state estimation algorithms is always desirable for controlling the operation of a power system, but it is limited by computation speed when conventional methods are used. In this work a neural network based method is developed for solving state estimation problem, with constraints emanating due to physical restrictions on the network parameters. The neural network method for solving state estimation problem can be implemented on dedicated neural network processors or reconfigurable hardware to outperform their software implementation. The feasibility of hardware implementation is demonstrated by implementing the algorithm on dedicated neural network processor (NNP) architecture, to solve a simple nonlinear programming problem. The prerequisite for state estimation is that the system under study must be observable i.e. there are sufficient measurements. A new method based on Hopfield neural network is developed to determine topological observability of power networks. The algorithm also determines where the meters should be placed in order to get an observable system. FACTS devices, now days are becoming essential part of transmission systems, for optimal use of transmission capacities, which offer several advantages, in the system, so there is requirement of such estimators which not only estimate the voltage magnitude and phase angle but also FACTS device control parameters. Thus the state estimation algorithms must incorporate FACTS devices. In this work FACTS devices are also included in the neural network based, state estimation algorithm.