(486b) Regional Air Quality Improvement Via Startup Plan Optimization for Multiple Chemical Plants

Industrial area where chemical plants are heavily aggregated often experiences highly localized and transient air pollution events, with elevated concentrations of multiple air pollutants and ozone concentrations. One identified major source related to these pollution events is the flaring emissions from chemical plants. Flaring is crucial to the chemical plant safety. However, excessive flaring, especially the intensive flaring during chemical plant startups, emits huge amounts of pollutions and also results in tremendous raw materials and energy loss. Thus, flare minimization at each point source must be minimized. Note that flare minimization at each individual plant may not be sufficient for regional air quality to meet national ambient air quality standards (NAAQS). One problem is that a large number of industrial point sources are spatially distributed in an industrial area and their startup flare emissions exhibit significant temporal variability. Such variability may cause an unexpected coincidence of flare emissions and thus induce localized and transient air pollution events. Therefore, from the emission control point of view, the best practices for improving regional air quality should integrate both efforts on plant-wide flare minimization at every industrial point source, and regional-wide emission variability minimization.

In this paper, a multiple chemical plant startup plan model is proposed for regional air quality improvement. In the model, geometric locations of plants and pollution transportation are both taken into account. Each plant proposes its startup time window and pollution emission response. Multi-period feature is introduced in the model, with which the pollution location and concentration in each time period could be simulated and optimized. The objective of the model is to minimize the pollution fluctuation at both temporal and spatial scales. Special constraints are set to prevent localized and transient pollution events. The whole model is a mixed-integer nonlinear programming (MINLP) model. It can be built in GAMS environment, and solved by the solver of BARON.