Exponential weighted methods for forecasting the hourly fine particulate matter (PM2.5) concentrations with seasonal cycle

• Main Authors: Bo-You Chen, 陳柏佑
• Other Authors: Ji-Cheng Wu
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/61191264538713150903

碩士 === 國立屏東科技大學 === 工業管理系所 === 100 === Due to people's living standards are improving, the number of motor vehicles grew rapidly, and the impact of industrialization, resulting in urban air quality has become worse. The quality of air has direct relationship with one’s health. Many studies had supported that fine particulates (PM2.5) suspended in the air are harmful to the human respiratory system and could further lead to severe cases of bronchitis. It had become an international trend to use the measurement of these fine particulates as the regulatory strategy of air quality control. Previous studies on fine particles, ozone and other air pollutants had focused on chemical properties of the air contaminants and basic statistical analysis. There are researches that based their prediction on the concentration of contaminants, but only limited to prediction of average concentration of the year, quarter, month, and day. Reports on air quality will aid people who are allergic to air pollutants and patients who suffer from bronchitis to take early precautions. This study used published data obtained from the Environmental Protection Administration website that collected the concentration of PM2.5 at every hour at stations in Ping-Tung city from January 1st, 2007 to December 31st, 2011. The research tried to construct two PM2.5 concentration prediction models based on the concept of moving average. Differed from diffusion and time series models, the research adopted longitudinal viewpoint to sort the historical data. From the time sequence plots, the data suggested that the concentration of PM2.5 cycles hourly though the day; this cycle continues throughout the months. As of result, the seasonal months were considered as an index and the seasonal factors were eliminated. Exponentially weighted moving average (EWMA) model I and II were used to predict the hourly PM2.5 concentration in Ping-Tung city throughout the day. The minimal mean absolute deviation and minimal one-step ahead Mean Squared Prediction Error were used as a criterion to determine the optimal smoothing parameter * in model I and II, respectively. Finally, the hourly PM2.5 concentration in Ping-Tung city from January 1st to April 30st, 2011 were predicted using the EWMA model I and II with the best smoothing parameter. The results showed that predicted values generally consistent with the real values.