Comparing Approaches to Probabilistic Design of ABAC Aeration Systems

The traditional WRRF design approach has been to design for the combined maximum load of several constituents during worst-case environmental conditions. Additionally, factors of safety may be applied to account for the uncertainty of various design assumptions. This can lead to over sizing of infrastructure and equipment or under estimation of existing process capacity. Recognizing that aeration systems (blowers and oxygen transfer systems) are notorious for being oversized, a recent project at the Denver Metropolitan Wastewater Reclamation District (MWRD) Robert W. Hite Treatment Facility (RWHTF) developed a basis of design for an aeration system based on a probabilistic approach. By taking a probabilistic approach it should be possible to simultaneously account for actual system variability and assumed uncertainty, thereby limiting conservatism of traditional design. Monte Carlo (MC) analysis is a commonly applied probabilistic approach whereby variability or uncertainty of model inputs can be propagated through mechanistic models resulting in probability distributions as outputs; effectively creating probabilistic shell around deterministic models (Benedetti, 2009). Probabilistic design approaches have been taken in our industry (Belia, 2012; Fries, 2010; McCormick, 2007; Gray, 2012). McCormick (2007) proposes a 20% reduction in the design blower capacity was possible by taking an MC approach and Gray (2012) determined that a 27% reduction in blowers capacity was possible by using an MC approach when designing a system using ammonia-based aeration control (ABAC) system. With the publication of the DOUT STR in 2020, it is likely that these techniques will see more application. Despite the benefits, properly executing a Monte Carlo analysis can be challenging, with possible pitfalls to consider and unique challenges in interpreting results. In this technical assessment, an actual aeration study will be used as a lens to compare MC analysis with an alternative Brute Force (BF) search approach and to answer the question of how these different approaches impact resulting basis of design?