ARTIFICIAL NEURAL NETWORKS FORECASTING OF THE PM10 QUANTITY IN LONDON CONSIDERING THE HARWELL AND ROCHESTER STOKE PM10 MEASUREMENTS

作者：Popescu M^*; Ilie C; Panaitescu L; Lungu M L; Ilie M; Lungu D

来源：Journal of Environmental Protection and Ecology, 2013, 14(4): 1473-1481.

• PM10
• forecasting
• pollution
• artificial neural network (ANN)
• London

This paper presents early results of a research that aims to forecast PM10 tiny particles (solid, liquid or gas) in the London Kensington area, considering measurements made 2 h in advance in Harwell and Rochester Stoke areas. Because it was considered a type of nonlinear problem (there is no mathematical relationship between used input and output data) the method was the modelling and simulation with artificial neural networks (ANN). Considered data were the wind speed, the wind direction and the amount of PM10 for the 2 areas (Harwell and Rochester Stoke) as inputs and the amount of PM10 in Kensington as output. The values represent the hourly measured data for 9 days: January 1st 2009 and January 9th 2009. Because the research presented in the present work is in early stages, the researchers defined the error values for training process smaller than 20%. The used ANN was feed forward with back propagation learning algorithm. After training and testing, values lower than 14% were obtained. The authors concluded that a future forecasting method for PM10 and other pollutants quantities was established. Thus, the results established a basis for further research where the method can be extended in order to forecast the possibility of pollution in populated areas with a relatively enough time to prepare the necessary contrameasures.

• 出版日期2013