Data Analytics Methods to Measure Plant Outage Resilience

Every 18 or 24 months, nuclear power plants (depending on plant configuration, pressurized- or boiling-water reactor, respectively) undergo an outage period where the plant is taken offline and a large number of maintenance and surveillance activities (that cannot be performed while plant is running) are performed in typically 2-3 weeks. Planning a plant outage is very challenging since all the activities must be performed in the shortest amount of time given available resources (typically contractor crews hired for the duration of the outage). Consequently, plant outages can be costly due the loss of power generation and crew costs and so there is a need to maximize resource usage in the outage planning phase and reduce the risk of outage delays. This paper addresses these needs by providing a set of analytical methods designed to analyze a plant outage schedule and identify critical elements based on available resources (time and crews). These methods are based on natural language processing and optimization algorithms. In this respect, we have developed two classes of methods: one that focuses on the time resource and how variability in the time to complete outage tasks may impact outage delays and one that minimizes the risk of outage delays by integrating available resources to assess when daily activities should be performed.