A neural network based inverse system control strategy to decouple turbine power in multi-reactor and multi-turbine nuclear power plant

Abstract

In multi-reactor and multi-turbine nuclear power plants (MMNPP), several reactor units working in parallel provide steam to turbines that generate power for different purposes. Due to a common steam header that all turbines are connected to, changes in the power output of one turbine can influence the performance of other turbines (and their power output) when operating in load following scenarios. This may threaten the stable and safe operation of the plant. To avoid the coupling effect between the power outputs of turbines, this paper presents a composite control method to decouple turbine power from each other. This (power) decoupling control strategy is composed of a neural network based inverse system (NNIS) and a robust controller. The neural network inverse system approaches the inverse system of the original nonlinear system. The robust controller, containing PI controllers and integrators, is added to the NNIS to construct a closed loop decoupling controller with strong robustness. Simulation results indicate that the proposed decoupling controller has excellent power decoupling capabilities and robust performance as well as fast tracking capability.