Karhunen-Loeve (PCA) based detection of multiple oscillations in multiple measurement signals from large-scale process plants

In the perspective of optimizing the control and operation of large scale process plants, it is important to detect and to locate oscillations in the plants. This paper presents a scheme for detecting and localizing multiple oscillations in multiple measurements from such a large-scale power plant. The scheme is based on a Karhunen-Loeve analysis of the data from the plant. The proposed scheme is subsequently tested on two sets of data: a set of synthetic data and a set of data from a coal-fired power plant. In both cases the scheme detects the beginning of the oscillation within only a few samples. In addition the oscillation localization has also shown its potential by localizing the oscillations in both data sets.