• Energy Research

Machine learning has proven to be a powerful tool for accelerating biofuel development. Although numerous models are available to predict a range of properties using chemical descriptors, there is a trade-off between interpretability and performance. Neural networks provide predictive models with high accuracy at the expense of some interpretability, while simpler models such as linear regression often lack in accuracy. In addition to model architecture, feature selection is also critical for developing interpretable and accurate predictive models. We present a method for systematically selecting molecular descriptor features and developing interpretable machine learning models without sacrificing accuracy. Our method simplifies the process of selecting features by reducing feature multicollinearity and enables discoveries of new relationships between global properties and molecular descriptors. To demonstrate our approach, we developed models for predicting melting point, boiling point, flash point, yield sooting index, and net heat of combustion with the help of the Tree-based Pipeline Optimization Tool (TPOT). For training, we used publicly available experimental data for up to 8351 molecules. Our models accurately predict various molecular properties for organic molecules (mean absolute percent error (MAPE) ranges from 3.3% to 10.5%) and provide a set of features that are well-correlated to the property. This method enables researchers to explore sets of features that significantly contribute to the prediction of the property, offering new scientific insights. To help accelerate early stage biofuel research and development, we also integrated the data and models into a open-source, interactive web tool.

Machine learning has proven to be a powerful tool for accelerating biofuel development. Although numerous models are available to predict a range of properties using chemical descriptors, there is a trade-off between interpretability and performance. Neural networks provide predictive models with high accuracy at the expense of some interpretability, while simpler models such as linear regression often lack in accuracy. In addition to model architecture, feature selection is also critical for developing interpretable and accurate predictive models. We present a method for systematically selecting molecular descriptor features and developing interpretable machine learning models without sacrificing accuracy. Our method simplifies the process of selecting features by reducing feature multicollinearity and enables discoveries of new relationships between global properties and molecular descriptors. To demonstrate our approach, we developed models for predicting melting point, boiling point, flash point, yield sooting index, and net heat of combustion with the help of the Tree-based Pipeline Optimization Tool (TPOT). For training, we used publicly available experimental data for up to 8351 molecules. Our models accurately predict various molecular properties for organic molecules (mean absolute percent error (MAPE) ranges from 3.3% to 10.5%) and provide a set of features that are well-correlated to the property. This method enables researchers to explore sets of features that significantly contribute to the prediction of the property, offering new scientific insights. To help accelerate early stage biofuel research and development, we also integrated the data and models into a open-source, interactive web tool.

ABSTRACTThe 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 co-digestion (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 are not useful predictors in a carefully operated facility.

ABSTRACTThe 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 co-digestion (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 are not useful predictors in a carefully operated facility.

AbstractCrop yield modeling is critical in the design of national strategies for agricultural production, particularly in the context of a changing climate. Forecasting yields of bioenergy crops at fine spatial resolutions can help to evaluate near‐term and long‐term pathways for scaling up bio‐based fuel and chemical production, and for understanding the impacts of abiotic stressors such as severe droughts and temperature extremes on potential biomass supply. We used a large dataset of 28,364 Sorghum bicolor yield samples (uniquely identified by county and year of observation), environmental variables, and multiple approaches to analyze historical trends in sorghum productivity across the USA. We selected the most accurate machine learning approach (a variation of the random forest approach) to predict future trends in sorghum yields under four greenhouse gas (GHG) emission scenarios and two irrigation regimes. We identified irrigation practices, vapor pressure deficit, and time (a proxy for technological improvement) as the most important predictors of sorghum productivity. Our results showed a decreasing trend of sorghum yields over future years (on average 2.7% from 2018 to 2099), with greater decline under a high GHG emissions scenario (3.8%) and in the absence of irrigation (4.6%). Geographically, we observed the steepest predicted declines in the Great Lakes (8.2%), Upper Midwest (7.5%), and Heartland (6.7%) regions. Our study demonstrates the use of machine learning to identify environmental controllers of sorghum biomass yield and predict yields with reasonable accuracy. These results can inform the development of more realistic biomass supply projections for bioenergy if sorghum production is scaled up. © 2020 Society of Chemical Industry and John Wiley & Sons, Ltd

AbstractCrop yield modeling is critical in the design of national strategies for agricultural production, particularly in the context of a changing climate. Forecasting yields of bioenergy crops at fine spatial resolutions can help to evaluate near‐term and long‐term pathways for scaling up bio‐based fuel and chemical production, and for understanding the impacts of abiotic stressors such as severe droughts and temperature extremes on potential biomass supply. We used a large dataset of 28,364 Sorghum bicolor yield samples (uniquely identified by county and year of observation), environmental variables, and multiple approaches to analyze historical trends in sorghum productivity across the USA. We selected the most accurate machine learning approach (a variation of the random forest approach) to predict future trends in sorghum yields under four greenhouse gas (GHG) emission scenarios and two irrigation regimes. We identified irrigation practices, vapor pressure deficit, and time (a proxy for technological improvement) as the most important predictors of sorghum productivity. Our results showed a decreasing trend of sorghum yields over future years (on average 2.7% from 2018 to 2099), with greater decline under a high GHG emissions scenario (3.8%) and in the absence of irrigation (4.6%). Geographically, we observed the steepest predicted declines in the Great Lakes (8.2%), Upper Midwest (7.5%), and Heartland (6.7%) regions. Our study demonstrates the use of machine learning to identify environmental controllers of sorghum biomass yield and predict yields with reasonable accuracy. These results can inform the development of more realistic biomass supply projections for bioenergy if sorghum production is scaled up. © 2020 Society of Chemical Industry and John Wiley & Sons, Ltd

Technoeconomic analysis (TEA) is an approach for conducting process design and simulation, informed by empirical data, to estimate capital costs, operating costs, mass balances, and energy balances for a commercial scale biorefinery. TEA serves as a useful method to screen potential research priorities, identify cost bottlenecks at the earliest stages of research, and provide the mass and energy data needed to conduct life-cycle environmental assessments. Recent studies have produced new tools and methods to enable faster iteration on potential designs, more robust uncertainty analysis, and greater accessibility through the use of open-source platforms. There is also a trend toward more expansive system boundaries to incorporate the impact of policy incentives, use-phase performance differences, and potential impacts on global market supply.