Real-time alarm management system for wide-area monitoring

Control center alarms for power systems usually arrive with low latency, quickly overwhelming operators. Previous approaches to dealing with alarm ordering include logical grouping (suppression and filtering) of alarms according to pre-defined rules and presentation of the resulting alarms to the operator. Knowledge-based approaches for alarm handling allow us to incorporate operator experience into the decision making process. However, there are limitations that impede their use in real-time systems. In this paper, we approach these limitations from two perspectives. First, we propose a mapping from knowledge representations such as network models (i.e., of a power system) and ontologies to the design of stream computing applications for efficient alarm processing. Secondly, we propose a new class of experiential, knowledge-based algorithms for ranking alarms and introduce a contextual ranking algorithm in this class. We also present initial evaluation results of this algorithm using a simulation-based metric.