Neurocontrol of Pressurized Water Reactors in Load-Follow Operations

Abstract

The neurocontrol technique was applied to control a pressurized water reactor (PWR) in load-follow operations. Generalized learning or direct inverse control architecture was adopted in which the neural network was trained off-line to learn the inverse model of the PWR. Two neural network controllers were designed: One provided control rod position, which controlled the axial power distribution, and the other provided the change in boron concentration, which adjusted core total power. An additional feedback controller was designed so that power tracking capability was improved. The time duration between control actions was 15 min; thus, the xenon effect is limited and can be neglected. Therefore, the xenon concentration was not considered as a controller input variable, which simplified controller design. Center target strategy and minimum boron strategy were used to operate the reactor, and the simulation results demonstrated the effectiveness and performance of the proposed controller.