Tree-Based Automated Machine Learning to Predict Biogas Production for Anaerobic Co-digestion of Organic Waste

Abstract

The dynamics of microbial communities involved in anaerobic digestion of mixed organic waste are notoriously complex and difficult to model, yet successful operation of anaerobic digestion is critical to the goals of diverting high-moisture organic waste from landfills. Machine learning (ML) is ideally suited to capturing complex and nonlinear behavior that cannot be modeled mechanistically. This study uses 8 years of data collected from an industrial-scale anaerobic codigestion (AcoD) operation at a municipal wastewater treatment plant in Oakland, California, combined with a powerful automated ML method, Tree-based Pipeline Optimization Tool, to develop an improved understanding of how different waste inputs and operating conditions impact biogas yield. The model inputs included daily input volumes of 31 waste streams and 5 operating parameters. Because different wastes are broken down at varying rates, the model explored a range of time lags ascribed to each waste input ranging from 0 to 30 days. The results suggest that the waste types (including rendering waste, lactose, poultry waste, and fats, oils, and greases) differ considerably in their impact on biogas yield on both a per-gallon basis and a mass of volatile solids basis, while operating parameters were not good predictors of yield at this facility.

Integration of anaerobic digestion with heat Pump: Machine learning-based technical and environmental assessment

“…Referring to the F-table these correspond to the probability value (p-value) of zero, suggesting that the GPR model correctly predicts the data trend (i.e., the correlation is highly significant). The predictive accuracy of the optimal model is competitive with those reported in the literature of ML-based AD modelling with R 2 in the range of 0.8 to 0.9 (Cruz et al, 2022;Long et al, 2021;Wang et al, 2021;Xu et al, 2021). This superior performance of the GPR model when compared to other models is attributed to the capability of probabilistic Gaussian processes to handle datasets with a high degree of variance (see Table 3).…”

“…Following the training procedure, several unseen trial cases (test data) are subjected to the ML model, based on which the accuracy of the model is evaluated. A wide range of ML models has been developed for predicting biogas production from AD processes, among which artificial neural network (ANN) (S ¸enol, 2021), recurrent neural network (RNN) (Park et al, 2021), random forest (RF) (Wang et al, 2021), support vector machine (SVM) (Long et al, 2021), and extreme gradient boost (XGBoost) (Xu et al, 2021) have been popular choices among researchers. Therefore, the work compares the accuracies of five different ML models for predicting biogas yield and methane content for AD processes.…”

Co-Processing Agricultural Residues and Wet Organic Waste Can Produce Lower-Cost Carbon-Negative Fuels and Bioplastics

“…Commercial or industrial sources, such as food processors, grocery stores, and restaurants, may produce more readily processable food waste streams. Additionally, a challenge for codigestion facilities is to ensure that all feedstocks are timely scheduled to deliver during weekdays and to reduce waste hauling time during weekends . While organic wastes are abundant and close to biorefineries, onsite storage can be used to mitigate this problem.…”

“…Concentrated animal feeding operations in the U.S. produce approximately 300 million metric tons of waste per year and result in the release of excess nutrients to the environment, causing human health and ecological damage. , Much of the recoverable dairy, beef, and swine manure is located in close proximity to current U.S. biorefineries and likely future locations. , For example, more than 80% of the total organic waste available in Iowa is manure . While more densely populated regions have municipal wastewater and other organic waste processing infrastructure that can be leveraged to treat a portion of this waste, rural communities are less likely to have such centralized infrastructure in place. Lignocellulosic biorefineries have the potential to share the costs and benefits of anaerobic digestion (AD) infrastructure in rural communities, thus mitigating methane emissions and enabling the use of otherwise stranded resources.…”

“…The onsite AD facility is an often-overlooked component of most advanced biorefinery designs, yet it has the potential to codigest a range of locally generated organic wastes, such as livestock manure and food processing waste, which are abundant in agricultural areas . Previous experimental studies and current industry practices have demonstrated the feasibility of codigesting lignocellulosic residues and organic wastes. ,, Taking advantage of codigestion, a pragmatic method that overcomes the challenges related to substrate properties and system optimization in a single-substrate AD process, can increase biogas production by supplementing the waste streams of ethanol production with manure and food waste . Codigestion also presents the opportunity to earn tipping fees as a revenue stream ($42–68 per wet metric ton of waste in the Midwestern U.S…”