Hybrid statistical-machine learning ammonia forecasting in continuous activated sludge treatment for improved process control

Newhart, Kathryn B.; Marks, Christopher A.; Rauch-Williams, Tanja; Cath, Tzahi Y.; Hering, Amanda S.

doi:10.1016/j.jwpe.2020.101389

Cited by 19 publications

(8 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Leveraging the complexities of environmental systems as an asset [81] with SL, ML—or Artificial Intelligence (AI), in general—likely has numerous practical applications. As with post-wildfire soils, agricultural soils [82], biological wastewater treatment systems [83], and oceans [84] all have critical biogeochemical cycles that require greater understanding. Global microbial-biogeochemical systems [76, 85] offer a near inexhaustible supply of data density—possibly enabling the principles of biomedical ’personalized medicine’ to be extended to environmental science for the precision management of soil, water, and air.…”

Section: Discussionmentioning

confidence: 99%

Statistical learning and uncommon soil microbiota explain biogeochemical responses after wildfire

Honeyman

Fegel

Peel

et al. 2022

Preprint

View full text Add to dashboard Cite

Wildfires are a perennial event globally and the biogeochemical underpinnings of soil responses at relevant spatial and temporal scales are unclear. Soil biogeochemical processes regulate plant growth and nutrient losses that affect water quality, yet the response of soil after variable intensity fire is difficult to explain and predict. To address this issue, we examined two wildfires in Colorado, USA across the first and second post-fire years and leveraged Statistical Learning (SL) to predict and explain biogeochemical responses. We found that SL predicts biogeochemical responses in soil after wildfire with surprising accuracy. Of the 13 biogeochemical analytes analyzed in this study, 9 are best explained with a hybrid microbiome + biogeochemical SL model. Biogeochemical-only models best explain 3 features, and 1 feature is explained equally well with hybrid or biogeochemical-only models. In some cases, microbiome-only SL models are also effective (such as predicting NH4+). Whenever a microbiome component is employed, selected features always involve uncommon soil microbiota (i.e., the 'rare biosphere', existing at < 1% relative abundance). Here, we demonstrate that SL paired with DNA sequence and biogeochemical data predict environmental features in post-fire soils, though this approach could likely be applied to any biogeochemical system.

show abstract

Section: Discussionmentioning

confidence: 99%

Statistical learning and uncommon soil microbiota explain biogeochemical responses after wildfire

Honeyman

Fegel

Peel

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Leveraging the complexities of environmental systems as an asset [81] with SL, ML-or Artificial Intelligence (AI), in general-likely has numerous practical applications. As with post-wildfire soils, agricultural soils [82], biological wastewater treatment systems [83], and oceans [84] all have critical biogeochemical cycles that require greater understanding. Global microbialbiogeochemical systems [76,85] offer a near inexhaustible supply of data density-possibly enabling the principles of biomedical 'personalized medicine' to be extended to environmental science for the precision management of soil, water, and air.…”

Section: Discussionmentioning

confidence: 99%

Statistical Learning and Uncommon Soil Microbiota Explain Biogeochemical Responses after Wildfire

Honeyman

Fegel

Peel

et al. 2022

Appl Environ Microbiol

View full text Add to dashboard Cite

show abstract

“…We also compared the feature sets selected by the ridge regression method to an adaptive lasso feature selection method using ridge regression for initial parameter estimation . Adaptive lasso yielded the same six features for the lab data set.…”

Section: Methodsmentioning

confidence: 99%

“…We tested the effect of adding a periodic time-dependent variable that was fit to the day of the week. First, we assigned the day of the week an integer from 0 to 6 and then scaled the variable to the unit circle by dividing by 6 and multiplying by 2π to get d. Next, we tested 1−10 sine and cosine pairs (e.g., sine(id) and cosine(id) for i = 1−10) as new variables as described by Newhart et al 17 While the periodic model did improve the linear model predictions (data not shown), we ultimately decided to remove the daily periodic variables to include only deterministic variables that were independent of historical operations. That is, the historical biogas flow rate peaks on Thursday, but the trend is almost certainly due to the HSW feeding schedule (Figure S2).…”

mentioning

confidence: 99%

Feature Engineering and Supervised Machine Learning to Forecast Biogas Production during Municipal Anaerobic Co-Digestion

Schroer,

Just

2023

ACS EST Eng.

View full text Add to dashboard Cite

Municipalities with excess anaerobic digestion capacity accept offsite wastes for co-digestion to meet sustainability goals and create more biogas. Despite the benefits inherent to codigestion, the temporal and compositional heterogeneity of external waste streams creates operational challenges that lead to upsets or conservative co-digestion. Given the complex microbial bioprocesses occurring during anaerobic digestion, prediction and modeling of the outcomes can be challenging, and machine learning has the potential to improve understanding and control of co-digestion processes. Biogas flows are a surrogate for process health, and here, we predicted biogas production from historical data collected by a water resource recovery facility (WRRF) during normal operation. We tested a daily lab and operational data set (n = 1089 after cleaning) and a minute-by-minute supervisory control and data acquisition (SCADA) operational data set (n = 491,761 after cleaning) to determine if forecasting biogas flow for a 24 h time horizon is feasible without collecting additional data. We found that a multilayer perceptron (MLP) neural network model outperformed tree-based and multiple linear regression models. Using a high-resolution SCADA data set for the first time, we showed that MLP neural networks could predict biogas production with an adjusted coefficient of determination (R 2 ) of 0.78 and a mean absolute percentage error of 13.4% on a holdout test set. Adding daily laboratory analyses to the model did not appreciably improve the prediction of biogas flows. Feature engineering was essential to an accurate prediction, and 11 of the 15 most important features in the SCADA model were calculated from raw SCADA outputs. In summary, this paper demonstrates that minute-scale SCADA information collected at a municipal co-digestion facility can forecast biogas production, as a first step toward a digital twin model, without additional data collection.

show abstract

Hybrid statistical-machine learning ammonia forecasting in continuous activated sludge treatment for improved process control

Cited by 19 publications

References 19 publications

Statistical learning and uncommon soil microbiota explain biogeochemical responses after wildfire

Statistical learning and uncommon soil microbiota explain biogeochemical responses after wildfire

Statistical Learning and Uncommon Soil Microbiota Explain Biogeochemical Responses after Wildfire

Feature Engineering and Supervised Machine Learning to Forecast Biogas Production during Municipal Anaerobic Co-Digestion

Contact Info

Product

Resources

About