Forecasting of ozone concentration using frequency MA-OWA model

Air pollution can cause the human health, plants growth and daily mortality in numerous studies over the past decade. Therefore, forecasting and analysis of air quality are important topics of research today. The causes of poor air quality are global warming, greenhouse effects, acid rain, etc. Air...

Full description

Saved in:
Bibliographic Details
Main Authors: Ching-Hsue Cheng, Sue-Fen Huang
Format: Conference Proceeding
Language:English
Subjects:
Air pollution
Air Quality
Carbon dioxide
Degradation
Frequency
Global warming
Humans
Hydrocarbons
moving average
Nitrogen
order weight average
ozone
Predictive models
Rain
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Air pollution can cause the human health, plants growth and daily mortality in numerous studies over the past decade. Therefore, forecasting and analysis of air quality are important topics of research today. The causes of poor air quality are global warming, greenhouse effects, acid rain, etc. Air pollution problems are related to the emissions of sulfur dioxides (SO 2 ), nitrogen dioxide (NO 2 ), suspended particulates (PM 10 ), ozone (O 3 ), carbon monoxide (CO), and unburned hydrocarbons (HC) and so on. O 3 adverse effects on human health has received intensively concern in recent years, It has been recognized as one of the principal pollutants that degrades air quality, thus we uses O 3 attribute to evaluate air quality. This study proposed a frequency moving average - order weight average (MA-OWA) model to forecast air quality by daily O 3 concentration. Due to O 3 data is belong to time series pattern, MA can simple calculation and OWA operator can aggregate multiple lag periods into single aggregated value by different situation parameters a. Frequency MA-OWA based time series model can efficiently and accurately forecast O 3 . To demonstrate of the proposed, air quality monitoring the urban sites of Hsinchu (Taiwan), is selected for the numerical experiment from 2007 was utilized. From the results, the proposed methods outperform the listing methods in RMSE and MAPE.
ISSN:1062-922X
2577-1655
DOI:10.1109/ICSMC.2009.5346212