A Dynamic Programming Model of U.S. Nuclear Power Plant Operations

This paper presents a dynamic programming (DP) model of an electric utility’s optimal policy for operating a nuclear power plant (NPP). The utility chooses the level of capacity utilization of the NPP as a function of signals about the NPP’s current operating state. In each period the utility must determine whether or not to operate the reactor, or shut it down for preventive maintenance or refueling, or to permanently close the plant. Maintenance performed during periodic refueling outages partially “regenerates” the NPP, reducing the risk of unplanned forced outages in succeeding periods. However since NPP’s are designed to satisfy base power load, refueling outages involve a substantial opportunity cost in terms of lost power generation. The utility faces a tradeoff between cost-efficiency, fuel-efficiency and plant safety and makes its decisions subject to exacting safety regulations by the U.S. Nuclear Regulatory Commission (NRC). Tightened safety regulations following the Three Mile Island accident in March 1979 have contributed to a near doubling of the mean duration of refueling outages from 8 weeks prior to TMI to over 14 weeks after TMI. Simultaneously the mean time between refuelings increased from approximately 12 months prior to TMI to nearly 18 months after TMI. Using monthly data on U.S. NPPs in the post-TMI era we estimate parameters of the utility’s profit function, the failure processes that lead to unplanned forced outages, and the parameters governing the duration of refueling outages. These parameters imply an endogenous distribution of operating spells and capacity utilization levels that depend on the NPP’s age, signals the operator receives about the NPP’s current operating state, and the duration since last refueling. The estimates of the DP model reveal that utilities appear responsive to NRC regulation insofar as they impute a very high cost to unplanned and forced outages. Utilities are also highly averse to causing unnecessary wear and tear on their NPP’s caused by stop/start operation of their NPP’s including planned and unplanned outages. Overall, the DP model yields very accurate predictions of nuclear power generation including the impact of the relatively rare event of NPP decommissionings. JEL Classification: C41–Duration Analysis