Effect of Partial Outage Representation on the Statistical Properties of Loss=of-Load Probability Estimates

Using a three-state Markov model to represent the available capacity of a generating unit, this paper examines the statistical properties of the Loss-of-Load Probability (LOLP) estimates of a system consisting of n such identical units. It is assumed that the units are observed for a total time of T units and the durations of each capacity state within this time-interval are recorded. The expected values and the upper confidence limits of the system LOLP are computed using (a) the full 3-state representation and (b) the corresponding equivalent forced outtage rate (EFOR) index. Numerical values are given to indicate that the expectation of LOLP using the EFOR model is larger than the corresponding value obtained from the three-state model. However, the difference between the expectations of the upper confidence limits for the two models is relatively smaller, and for high load values it actually changes sign. Thus this study suggests that depending on the details of the decision in view it may sometimes be preferable to pool the partial and full outage data (as in the EFOR model), rather than keep them distinct.