Virtual Utility Plant를 위한 인공지능 기법 기반 데이터 보간 데이터 전처리 시스템 개발

Development of AI-based data interpolation data preprocessing system for Virtual Utility Plant

Today, the energy crisis has begun to materialize due to various factors, including wars and global climate issues. To address this, we introduced the concept of a Virtual Energy Utility Plant to reduce energy consumption in industrial complexes, which account for 57.29% of Korea’s total energy use. Reliable data collection is essential for the study of Virtual Utility Plants (VUPs) to ensure accurate analysis and prediction. In this paper, we devised a data generation technique that goes beyond interpolation by using LSTM to recover missing data and ensure continuous availability. Additionally, we introduced FLAG, a technique that generates data based on existing patterns even when sensors fail. The resulting model achieved an RMSE of 4.20 and an accuracy of 96.59%, and was able to make over 120 predictions per second on a low-power device. This led to a study on deploying edge devices in individual factories.