MetaHeuristics
Previous Projects

Thermal performance improvment advisor

The efficiency of electricity generation plants based on fossil or nuclear energy is sensitive to degradation in performance of steam-cycle components such as reheaters, condensers, feedwaterheaters, deaerators and the various turbine stages. Such degradation may be difficult to pinpoint e.g fouling in the feedwater heaters, noncondensible buildup, turbine blade erosion etc, even though its overall effect shows up as a covering of plant efficiency. The advisor developed is based on Metacomp, but contains a preprocessing module that takes measured temperatures, pressures and flow rates around each component (called unit) and devices a Real Unit Efficiency. Unit efficiencies are then combined with other variables, such as ambient conditions, cooling water temperature and control settings to train a neural networks to predict global efficiency (or output power). The network can be inverted then to identify the units that have degraded performance when there is a reduction in global efficiency, and a diagnosis can be made as to the reason for the for degradation.