Abstract:
In order to determine the ongoing status of an entire power plant or one of its components, the procedure generally adopted is to examine the process variables involved. In most process analysis systems, however, these process variables are dealt with in isolation from each other. The article gives an overview about new possibilities of process visualization and process analysis in power plants using the SOM approach. Furthermore some aspects about on-line and off-line analysis of process trajectories will be discussed, because one major point is to analyse the similarity of a reference curve (which means a good reference start-up or shut-down of the plant) and a test curves (test start-up or shut-down).
With it the article describes a non-linear, generally applicable method for analysing complex, multi-dimensional process states in a power plant by means of Self-Organizing Maps. This method of process analysis thus extends current methods for process monitoring, since it is generally applicable, and no longer merely examines each process variable in isolation, but evaluates their interrelationships as an integrated whole, as well as providing the appropriate display options. The procedure can be used both for static (steady-state) and for dynamic (transient) power plant phenomena.