Neural Network Technology as a Pollution Prevention Tool in the Electric Utility Industry

Abstract

This paper documents efforts by the Lower Colorado River Authority (LCRA) to pilot test the use of neural network technology as a pollution prevention tool for reducing stack emissions from a natural gas-fired power generating facility. The project was funded in part by a grant from the U.S. Environmental Protection Agency (EPA), Region VI. Combustion control is quickly becoming an emerging alternative for reducing utility plant emissions without installing costly "end of pipe" controls. The LCRA estimates that the technology has the potential to improve the thermal efficiency of a large utility boiler by more than 1 percent. Preliminary pilot test results indicate that a 0.5 percent improvement in thermal efficiency at the 430 MW gas-fired utility boiler will result in an estimated energy savings of 76,000 mmBtus and carbon dioxide (CO2) reductions of 4,079 tons per year. This paper describes the processes that were undertaken to identify and implement the pilot project at LCRA's Thomas C. Ferguson Power Plant, located in Marble Falls, Texas. Activities performed and documented include lessons learned, equipment selection, data acquisition, model evaluation and projected emission reductions.