Tuesday, July 28, 2009

Neural Networks 7/28/2009


I am currently in ST. Louis at a NEUCO user conference. NEUCO is a company that supplies neural network software and hardware that uses Learning Artificial Intelligence to provide performance improvements to power plant performance. The conference was set up to inform users about the different software enhancements that are currently available, and to see them in operation at a Power Plant. The NEUCO suite of performance solutions include, Combustion Opt, Performance Opt, Sootblower Opt, and Maintenance Opt.

Neural Networks sit on top of a Distributed Control System (DCS) and tunes the realtime data inputs from the electricity generating unit to improve unit performance. The nuts and bolts of the system works like this. Data from the DCS is retrieved by the neural network and the neural network trims (applies biases) to the data to push it above or below a predetermined value. The biases implemented by the neural network provides for better mixing of the fuel and air to the boiler thus improving the combustion process. Other tangential benefits of the neural network are improved Heat Rate (a measure of boiler efficiency), Nitrogen Oxides (NOx), CO and operating unit Opacity.

Neural networks when used in this application reduces power plant emission's to EPA designated standards. This is a plus for the environment and allows coal fired power plants to operate under the emissions limits that have been passed by congress and enforced by the EPA.

No comments:

Post a Comment