Predicción de la temperatura Predicción de la temperatura de la de la contención primaria de una central nuclear contención primaria de una central nuclear mediante redes neuronales

Abstract

In a nuclear power plant, Drywell atmosphere temperature is an

important operating parameter, for which technical specifi cations

in each power plant set a limit in normal operation. This limit is an

initial condition in the analysis of safety requirements. Specifi cally,

in Santa María de Garoña’s power plant, this temperature has

an upper limit in normal operation of 58 °C. The cooling system

installed to remove the heat produced inside the Drywell consists

mainly of refrigeration units (3), pumps (2), and heat exchangers

or HVH equipments (5). The pump in service drives the coolant

(demineralized water) previously cooled by refrigeration units in

service (2), which fl ow through the battery of the 5 heat exchangers,

cooling Drywell atmosphere, which is inerted with nitrogen. Thus,

the heated coolant returns to the refrigeration units completing the

circuit. In case of unavailability of any HVH in service, it would occur

a slight increase in average Drywell temperature.

This article shows the use of neural networks to predict the evolution

of Drywell atmosphere temperature against the variation of different

variables that infl uence it, such as service water temperature

(cold sink of the system), the setting of the compressors of the

refrigeration units, the core fl ow and the number of HVH working.

Neural networks used in this article incorporate all historical

knowledge stored about the different ways of cooling system

operation for the Drywell atmosphere temperature, and they show

high accuracy in predicting that temperature. Thereby, they are

a useful tool for taking decisions over the operating point of the

system.