A Lagrangian based nonlinear closed loop control algorithm : application to the advanced liquid metal reactor

In this dissertation, a closed loop nonlinear control algorithm for the on-line control of nuclear power plants has been developed. The algorithm is based on the Lagrangian version of Pontryagin's Maximum Principle (PMP). The two-point boundary value problem associated with the PMP formalism has been converted into an initial value problem. This conversion, valid only when a demand is prescribed, is an important contribution to the advanced control field since it allows the on-line operation of the control algorithm. The present Lagrangian based control algorithm has the capabilities to perform on line load following, plant state estimations, and the ability to treat unmeasurable plant parameters as uncertain parts in the controller model while taking into account the effects of the control mechanism response function on the generated control signals. The control algorithm has been proven to be stable in the small, robust against large perturbations, and once tuned, capable of handling a wide range of transients. In view of these properties, it is able to operate safely under degraded plant conditions. The advanced controllers developed in this dissertation exhibit the following features: plant state estimation, estimations of unmeasurable plant parameters, analysis of plant performance, and determination of the correct control action to meet the corresponding predetermined demand. The usefulness of the new control algorithm has been demonstrated by the development a complete set of real-time advanced controllers for the Multimodular Advanced Liquid Metal Reactor. A sensor diagnostics algorithm has been developed to provide on-line estimates of the rate constant of sensors during load-following transients.