A SIMPLE NEURAL-NETWORK FOR ESTIMATING EMISSION RATES OF HYDROGEN-SULFIDE AND AMMONIA FROM SINGLE-POINT SOURCES

Neural networks have shown tremendous promise in modeling complex prob lems. This work describes the development and validation of a neural n etwork for the purpose of estimating point source emission rates of ha zardous gases. This neural network approach has been developed and tes ted using experimental data obtained for two specific air pollutants o f concern in West Texas, hydrogen sulfide and ammonia. The prediction of the network is within 20% of the measured emission rates for these two gases at distances of less than 50 m. The emission rate estimation s for ground level releases were derived as a function of seven variab les: downwind distance, crosswind distance, wind speed, downwind conce ntration, atmospheric stability, ambient temperature, and relative hum idity. A backpropagation algorithm was used to develop the neural netw ork and is also discussed here. The experimental data were collected a t the Wind Engineering Research Field Site located at Texas Tech Unive rsity in Lubbock, Texas. Based on the results of this study, the use o f neural networks provides an attractive and highly effective tool to model atmospheric dispersion, in which a large number of variables int eract in a nonlinear manner.