1 Abstract--A power system disturbance scenario of particular interest is that which occurs when one or more generators in a distributed power system is ‘taken out’. Under such circumstances the load becomes especially great on the remaining generators and the speed of rotation will rapidly decrease changing the frequency of a mains signal. Fast and accurate change in frequency rate estimation algorithms are used for disturbance detection, to allow enough time for introduction of appropriate control measures that will prevent system failure. The main application would be in the load shedding scheme. Some recent work has suggested estimating the change in frequency by use of the ambiguity function and cubic phase function algorithms. The method presented in this paper estimates mains power signal frequency drifts with the use of digital phase locked loop (DPLL). Simulations show that DPLL based methods can estimate the frequency drift with a lower mean square error (MSE) than existing maximization methods. Index Terms—Interconnected power systems, Power system monitoring, Frequency estimation. I. INTRODUCTION he changes in a generator’s rotating speed and the frequency (of the generated power signal) are dependent on the power supplied. Significant changes in main power signal frequency are caused by generator overloads. Power is shared between neighboring generators. Thus in an event of an overload, the frequency in connected generator site will change. The site with the overload is identified as the one with the maximum frequency drift. Thus accurate instantaneous frequency rate (IFR) estimation algorithms are required to detect the disturbance. Then appropriate control procedures can be applied to prevent system failure by ensuring an adequate level of shedding in the right areas. The speed of an algorithm is also important because the frequency rate estimates can be used to accelerate the load shedding schedule when there is a large system disturbance. Mark Glickman is with the School of Engineering Systems, Queensland University of Technology, Brisbane, Australia (e-mail: m.glickman@student.qut.edu.au) Shui-cheong Kam is with the School of Engineering Systems, Queensland University of Technology, Brisbane, Australia (e-mail: s.kam@qut.edu.au) Zahir Hussain is with the School of Electrical and Computer Engineering, RMIT University, Melbourne, Australia (e-mail: zahir.hussain@rmit.edu.au) The analytic version of the power signal for unit sampling is given by: ( ) ( ) ) ( 2 2 1 0 0 n z e b n z w n a n a a j r + = + + (1) or: ( ) ( ) ) ( sin 2 2 1 0 0 n z n a n a a b n z w r + + + = (2) where = 0 a Phase (radians), = 1 a Frequency (radians/s), = 2 a Frequency drift (radians/s 2 ), = 0 b Amplitude, = n Discrete time and = ) (n z w White Gaussian noise (WGN). It is assumed that the WGN is caused by local load variations. A. Polynomial Maximization Algorithms A well known method for estimating the instantaneous frequency is the Wigner-Ville Distribution (WVD) [1-4]. Discrete Wigner-Ville Distribution (DWVD) is given by [2]: ( ) ( )  - = - = L L l l j r z e l n K n W r ω ω 2 , 2 , (3) where: ( ) ( ) ( ) ( ) ( ) l w l w l n z l n z l n K r r r - - + = * * , = l Discrete time shift, = L Maximum time shift and ( ) l w = window sequence. Windows are used because WVD is a noncausal operation [2]. DWVD can be used to estimate the instantaneous frequency by finding the maximum frequency, ω for each time sample. Rate of change in frequency estimate is the derivative of the instantaneous frequency. An alternative approach is the Discrete Ambiguity Function (DAF) maximization [5]: ( ) ( ) ()  - = - + = τ ω ω N n n j r r z e n z l n z l DAF r 0 * , (4) The Use of Digital Phase Locked Loops for Estimation of Instantaneous Frequency Rate in Distributed Power Networks M. Glickman, Member, IEEE, S.-c. Kam, Member, IEEE, Z. Hussain, Senior Member, IEEE T