Dirichlet Process-based Prediction of Gasholder Level with Self-learning Scheme in Steel Enterprises

Yu Zhang,Long Chen,Zhongyang Han,Jun Zhao,Wei Wang

DOI: https://doi.org/10.1109/cac57257.2022.10055800

2022-01-01

Abstract:Gasholder level prediction is very important for the workers to perform energy scheduling in steel enterprises. The gasholder level model is time-varying due to environmental disturbance. Since the traditional gasholder level prediction are often based on off-line model without the consideration of the change of working conditions, the prediction results may be inaccurate. This paper proposes a gasholder level prediction method with self-learning based on Dirichlet process, to deal with the multi-conditions structure of gas system. The Dirichlet process is introduced as a priori distribution so as to automatically decide the number of feature clusters, and then realize the automatic clustering algorithm for multiple working conditions of gasholder. Compared with other state of the art prediction methods, the advantage of this method is that the number of clusters need not to be specified in advance. The experimental results show that the method in this paper achieves better performance in the accuracy of gasholder prediction.