Simulation of the Ozone Concentration in Three Regions of Xinjiang, China, Using a Genetic Algorithm-Optimized BP Neural Network Model

Accurate ozone concentration simulation can provide a health reference for people’s daily lives. Simulating concentrations is complex task because near-surface production determined by combination of volatile organic compounds (VOCs) and NOx emissions, atmospheric photochemical reactions, meteorological factors. In this study, we applied genetic algorithm-optimized back propagation (GA-BP) neural network, multiple linear regression (MLR), BP random forest (RF) algorithm, long short-term memory network (LSTM) to model in three regions Xinjiang, China (Urumqi, Hotan, Dushanzi districts) the first time inputting wind speed, humidity, visibility, temperature, direction. The results showed that average relative errors simulations Urumqi, districts were (61%, 14%, 16%), MLR (97%, 23%), RF (39%, 11%, 14%), LSTM (50%, 12%, GA-BP (16%, 4%, 6%) significance coefficients R2 (0.73, 0.65, 0.83), (0.68, 0.62, 0.74), (0.85, 0.80, 0.88), (0.78, 0.74, 0.85), (0.92, 0.93, 0.94), respectively, with simulated values being closest true values. among 100 samples same direction data, highest 173.5 μg/m3, 114.3 228.4 respectively. theoretical basis effective control regional pollution urban areas (Urumqi), dusty (Hotan), industrial (Dushanzi) Xinjiang.