양자역학 기반의 신경망을 이용한 가스 터빈 에너지 수율 예측 프레임웍

초록

The company produces electric energy using gas turbine, a carbon neutralitytechnology, as an eco-friendly power generation method. Gas turbines are mainlyused in engines that require large power, such as aircraft and train generators, andare rotary power engines in which electrical energy is extracted from the flow ofhigh-temperature, high-pressure gas. When producing electrical energy usingnatural gas, companies must accurately predict turbine energy yield (TEY) and useelectrical energy efficiently. However, data collected from sensors attached to gasturbine contain noise and have low variance. These characteristics degrade theanalysis performance using neural networks and affect the accuracy of predictions. Therefore, to overcome this issue, this study models the weight update method inthe deep neural network (DNN) learning process as a quantum mechanics-basedstochastic process, taking into account the noise in the data, and proposes aframework to predict accurate TEY. To prove the superiority of the proposedframework, it is compared with existing DNN algorithms.

키워드