Abstract

The city of Copiapo in Chile, close to the Hernan Videla Lira smelting copper plant, often suffers episodes of high SO2 concentration levels that seriously affect the well-being of its inhabitants. A system forecasting high SO2 levels was developed in order to alert the people of Copiapo and to give the plant engineers time to alter the operating conditions and thereby prevent the occurrence of these episodes or at least minimize their magnitude. The system uses a fuzzy clustering technique to process on-line data streaming from five monitoring stations every minute. The system is capable of forecasting SO2 levels precisely two hours in advance during the summer season, when SO2 concentrations are usually low. However, during winter, when SO2 levels are high, the system underpredicts maximum values.