[1M3-GS-13] AI application: Social data and prediction

[1M3-GS-13-04] Online Steam Quantity Prediction Using Time-series Data of Waste Incineration Plant

〇Yuto Ichikawa¹, Ken’ichiro Shimada¹, Daichi Kimura¹, Tomonori Izumitani¹ (1. NTT Communications Corporation)

Keywords:Online Prediction, Machine Learning

Mainly in the manufacturing and plant industries, sensor data prediction using machine learning from sensor data such as temperature and pressure is getting increasing attention. However, if the distribution of feature values involved in prediction fluctuates dynamically, the prediction accuracy tends to be unstable depending on the selection of the learning data and the test data of the machine learning model. This paper proposes an online prediction method that adaptively estimates the number of components of the PLS regression model and learns only the latest data. The proposed online prediction approach is applied to the prediction of steam volume in waste incineration plant and its performance is compared to that of the existing method with the root mean square error and correlation coefficient to verify the superiority of this approach.