Feedwater Flow Estimation in Nuclear Power Plants Using Neural Networks

摘要

Most of Pressurized Water Reactors(PWRs) utilize venturi meters to measure the steam generator feedwater flow rate. However, the venturi meters are known to be susceptible to the fouling due to the corrosion products that are entrained in the feedwater. This fouling results in an overestimation of the feedwater flow rate due to an increased pressure drop across the venturi. Consequently, the reactor thermal power is also overestimated and operators are forced to derate the plant in order to operate within the licensed power limits. The average derating ranges from 1% to 2% of full power. In this study, the steam generator feedwater flow rate was estimated using the artificial neural networks, which can evaluate the flow rate regardless of changes in the measurement system or physical state of the feedwater. The wavelet denoising was also applied for enhanced information extraction in the data processing.