Learning, fast and slow: a two-fold data-based model adaptation algorithm for processes under varying operating conditions

Abstract

This article proposes a novel two-fold learning framework to maintain the accuracy of data-based process models, typically affected by two types of uncertainty. Out-of-domain uncertainty arises when the modelled process operates under conditions not represented in the training dataset, in-domain uncertainty results from real-world plant variability. To handle out-of-domain uncertainty, a slow learning component learns system dynamics under unexplored operating conditions; it consists of an ensemble of neural network models, featuring (i) a combination rule that weights individual models based on the statistical proximity between their training data and the current operating condition, and (ii) a monitoring algorithm based on statistical control charts that supervises the ensemble's performance and triggers the offline training and integration of a new model when a new operating condition is detected. To address in-domain uncertainty, a fast learning component continuously compensates in real time for the mismatch of the slow learning model using Gaussian processes. The proposed methodology is tested on an energy process system referenced in the literature, demonstrating that the combined use of slow and fast learning components improves model accuracy compared to standard adaptation approaches.