Constrained neural control for the adaptive tracking of power profiles in a triga reactor

In this study, a control scheme to accomplish the tracking of power profiles in a TRIGA Reactor is presented. This scheme permits the fulfillment of the inverse period constraint. Additionally, it eliminates the need for a physical model of the plant. Closed-loop identification of the nuclear system is carried out based only on external reactivity and neutron power by a differential neural network. This network utilizes a new learning law by which it is possible to guarantee the boundedness for weights and identification error. Once a neural model of the reactor is obtained, inverse period constraint can be expressed as a new constraint on the control input (external reactivity). However, an error term must be calculated to determine the new admissible set of control signal. This difficulty is overcome by using the sliding mode technique. Finally, a new control law is proposed. The effectiveness of this procedure is illustrated by numeric simulation.