Closure of “A Unified Approach to Transient Stability Contingency Filtering, Ranking, and Assessment”

We are indebted to Professor S. Rovnyak for his appreciation and thoughtful comments. The questions raised in his discussion give us the opportunity to clarify some key issues of the FILTRA technique and, more generally, to elaborate further on the SIME method and its two-faceted approach to transient stability assessment and control. In the following we consider his questions in the same order as they appear in the discussion. The first three questions refer to the filtering task of FILTRA, achieved by Block 1 in Fig. 1 of the paper. 1) We agree that it might be beneficial to sacrifice reliability for the sake of computational efficiency, especially since, as you correctly observe, many uncertainties may exist about the system parameters and modeling, the system state, and the very definition of the contingencies. Note, however, that some of these uncertainties fade as the system gets closer to real-time operation. Remember also that SIME does not introduce any additional uncertainties with respect to the time-domain program that it drives; of course, a good statistical approach could be more appropriate for handling uncertainties. More generally, the tradeoff between accuracy and computational efficiency depends on the power system under consideration as well as the operational practices in use. The structure of Block 1 in Fig. 1 may be designed so as to comply with various requirements. For example, reference [15] of the paper uses a two-step procedure, where the first step relies on the classical power system modeling together with a tailor-made ultra-fast transient stability program. Obviously, such modeling and programs, which are found to give sound-though approximate-assessment for the system simulated in [15], would be unacceptable for other power systems (e.g., for the Hydro-Québec one). However, this tradeoff between accuracy and computing performances is not necessarily a real issue. Indeed, contingency filtering is an easily parallelizable process, where computations distributed among, say, “c” computers divide the computing time by almost c; this may allow one to consider as many contingencies as deemed a priori interesting without sacrificing to accuracy. 2) We also agree that we could use an adaptive OMIB model to exploit its predictive capabilities, thus speeding up the computations. This may be achieved via the 2-machine equivalent proposed in [1] or via the OMIB derived in the context of the Emergency SIME discussed in 6). 3) Admittedly, in order to discard first-swing stable contingencies one could think of “pure” time-domain simulations without using SIME, thus making it easier to implement. However, an advantage of SIME is that for first-swing unstable contingencies it provides a stability margin which can be extrapolated (or interpolated) with the margin provided by the “ranking-assessment block” (Block 2 in Fig. 1) in order to identify potentially harmful and harmless contingencies