Prediction of NO2 Emission Concentration via Correlation of Multiple Big Data Sources Using Extreme Learning Machine

Abstract

Increase of electricity demand and urbanization process has caused more power plants to be built to meet the demand of electricity. However, development of power plant will cause environmental issue for its surrounding. Necessary measures need to be taken to ensure social and environmental sustainability. Among the requirements in Malaysia, discharge of air pollution emission of a gas- or distillate-fired power plant has to comply with air pollution level as described in the Malaysian Ambient Air Quality Standards ((MAAQS) 2013 and the Environmental Quality (Clean Air) Regulations 2014. Pertaining to the environmental requirements, this paper is to investigate the ability of a regression based artificial intelligence tool, namely Extreme Learning Machine (ELM) in correlating multiple sources of big data sets and subsequently predicting the air pollution emission level from the chimney of a Combined Cycle Gas Turbine (CCGT) power plant. This emission data is later being used to ensure the clean air regulatory requirement is fulfilled. The big data sources that have been used in this work are meteorological data, terrain and land use data, historical emission data and power plant parameters particularly related to the point source emitter. With the correlation of multiple big data sources, Extreme Learning Machine (ELM) is then trained for the prediction of emission rate at certain targeted areas, which are classified as air sensitive receptors (ASR) surrounding the power plant. Nitrogen dioxide (NO2) is the key emission that has been studied in this paper due to its criticality towards environment. A standalone application program has been developed to employ ELM based big data analytics tool for the prediction of NO2 pollution emission. The output of ELM is analyzed to ensure the emission at ground level of ASR is maintained within allowable limit.