Decision theory for fault diagnosis in electric power systems (Q1090597)

In state estimation of electric power systems, anomalous measurement diagnosis is essential for guaranteeing a reliable database. More specifically, the identification of anomalous data is a very important but also very difficult task, especially in the case of multiple interacting ones. This work focuses primarily on the latter problem. The proposed solution is based on the classical hypothesis testing of decision theory and its application to the random vector of measurement error estimates; the latter are obtained via a perfect observations estimator which is proven to be optimal. Furthermore, strategies able to treat any type of bad measurements are considered; the identification strategy under an upper bounded \(\beta\) risk is chosen and extensively exploited. The developments concern reliability aspects as well as on- line computational requirements for which many algorithmic improvements and useful formulae are derived. A practical example, coming from actual implementation experience, illustrates the overall methodology and shows the consistently good performances of the method.