Study Of Weather Forecasting And Gas Diffusion/Source Localization Based On Machine Learning

Environmental protection is an significant part of China’s sustainable development strat-egy.Among them,environmental monitoring technology provides a large amount of mean-ingful data support for environmental management and decision-making,and it is an important part of environmental protection.This paper focuses on two significant issues in environmen-tal monitoring:weather forecasting and polluting gas diffusion/source localization.We use sensor technology(including photoelectric sensor)and machine learning to improve the accu-racy of existing weather forecasting methods and enhance the intelligent level of monitoring in industrial parks.We first studied the implementation of a weather forecast system based on machine learn-ing.A prediction framework called E-STAN-MLP was innovatively proposed.Prediction is made by the meteorological element data observed by the ground weather station sensors and the Ruitu numerical forecast system data developed by the Institute of Urban Meteorology(IUM),China Meteorological Administration in Beijing.Taking 24 meteorological stations in Beijing as an example,the forecasting accuracy of the four meteorological elements(temperature,hu-midity,wind speed,and wind direction)has been greatly improved.In addition,this paper established a gas diffusion simulation data set based on Fluent software for an industrial park in Zhejiang Province.We have studied several sub-problems:the first is the optimization of sensor location placement;the second is to use the machine learning algorithm to directly locate the pollution source based on the sensor’s observed concentration data;the third is to establish a gas diffusion model and combine the particle swarm optimization algorithm to indirectly locate the pollution source.In this industrial park,the accuracy of locating the pollution source to specific factory reached over 95%,and the accuracy of locating the pollution emission source reached 80%.