Power System Stability Analysis Using Wide Area Measurement System

Abstract

Advances in wide area measurement systems have transformed power system operation from simple visualization, state estimation, and post-mortem analysis tools to real-time protection and control at the systems level. Transient disturbances (such as lightning strikes) exist only for a fraction of a second but create transient stability issues and often trigger cascading type failures. The most common practice to prevent instabilities is with local generator out-of-step protection. Unfortunately, out-of-step protection operation of generators may not be fast enough, and an instability may take down nearby generators and the rest of the system by the time the local generator relay operates. Hence, it is important to assess power system stability over transmission lines as soon as the transient instability is detected instead of relying on purely localized out-of-step protection in generators.

This thesis proposes a synchrophasor-based out-of-step prediction methodology at the transmission line level using wide area measurements from optimal phasor measurement unit (PMU) locations in the interconnected system. Voltage and current measurements from wide area measurement systems (WAMS) are utilized to find the swing angles. The proposed scheme was used to predict the first swing out-of-step condition in a Western Systems Coordinating Council (WSCC) 9 bus power system. A coherency analysis was first performed in this multi-machine system to determine the two coherent groups of generators. The coherent generator groups were then represented with a two-machine equivalent system, and the synchrophasor-based out-of-step prediction algorithm then applied to the reduced equivalent system. The coherency among the group of generators was determined within 100 ms for the contingency scenarios tested. The proposed technique is able to predict the instability 141.66 to 408.33 ms before the system actually reaches out-of-step conditions.

The power swing trajectory is either a steady-state trajectory, monotonically increasing type (when the system becomes unstable), or oscillatory type (under stable conditions). Un- der large disturbance conditions, the swing could also become non-stationary. The mean and variance of the signal is not constant when it is monotonically increasing or non-stationary. An autoregressive integrated (ARI) approach was developed in this thesis, with differentiation of two successive samples done to make the mean and variance constant and facilitate time series prediction of the swing curve.

Electromagnetic transient simulations with a real-time digital simulator (RTDS) were used to test the accuracy of the proposed algorithm with respect to predicting transient in- stability conditions. The studies show that the proposed method is computationally efficient and accurate for larger power systems. The proposed technique was also compared with a conventional two blinder technique and swing center voltage method. The proposed method was also implemented with actual PMU measurements from a relay (General Electric (GE) N60 relay). The testing was carried out with an interface between the N60 relay and the RTDS. The WSCC 9 bus system was modeled in the simulator and the analog time signals from the optimal location in the network communicated to the N60 relay. The synchrophasor data from the PMUs in the N60 were used to back-calculate the rotor angles of the generators in the system. Once the coherency was established, the swing curves for the coherent group of generators were found from time series prediction (ARI model). The test results with the actual PMUs match quite well with the results obtained from virtual PMU-based testing in the RTDS. The calculation times for the time series prediction are also very small.

This thesis also discusses a novel out-of-step detection technique that was investigated in the course of this work for an IEEE Power Systems Relaying Committee J-5 Working Group document using real-time measurements of generator accelerating power. Using the derivative or second derivative of a measurement variable significantly amplifies the noise term and has limited the actual application of some methods in the literature, such as local measurements of voltage or voltage deviations at generator terminals. Another problem with the voltage based methods is taking an average over a period; the intermediate values cancel out and, as a result, just the first and last sample values are used to find the speed. This effectively means that the sample values in between are not used. The first solution proposed to overcome this is a polynomial fitting of the points of the calculated derivative points (to calculate speed). The second solution is the integral of the accelerating power method (this eliminates taking a derivative altogether). This technique shows the direct relationship of electrical power deviation to rotor acceleration and the integral of accelerating power to generator speed deviation. The accelerating power changes are straightforward to measure and the values obtained are more stable during transient conditions. A single machine infinite bus (SMIB) system was used for the purpose of verifying the proposed local measurement based method.