Steam Generator control in Nuclear Power Plants by water mass inventory

Control of water mass inventory in Nuclear Steam Generators is important to insure sufficient cooling of the nuclear reactor. Since downcomer water level is measurable, and a reasonable indication of water mass inventory near steady-state, conventional feedwater control system designs attempt to maintain downcomer water level within a relatively narrow operational band. However, downcomer water level can temporarily react in a reverse manner to water mass inventory changes, commonly known as shrink and swell effects. These complications are accentuated during start-up or low power conditions. As a result, automatic or manual control of water level is difficult and can lead to high reactor trip rates. This paper introduces a new feedwater control strategy for Nuclear Steam Generators. The new method directly controls water mass inventory instead of downcomer water level, eliminating complications from shrink and swell all together. However, water mass inventory is not measurable, requiring an online estimator to provide a mass inventory signal based on measurable plant parameters. Since the thermal-hydraulic response of a Steam Generator is highly nonlinear, a linear state-observer is not feasible. In addition, difficulties in obtaining flow regime and density information within the Steam Generator make an estimator based on analytical methods impractical at this time. This work employs a water mass estimator based on feedforward neural networks. By properly choosing and training the neural network, mass signals can be obtained which are suitable for stable, closed-loop water mass inventory control. Theoretical analysis and simulation results show that water mass control can significantly improve the operation and safety of Nuclear Steam Generators.