(38c) Multi-Stage Stochastic Optimization Framework for Design of Integrated Biorefineries Under Uncertainty

Multi-stage Stochastic Optimization Framework for Design of Integrated Biorefineries under Uncertainty

As research is exploring possibilities for developing renewable bio-based fuels and chemicals, there arises a need for development of  strategies which can design sustainable value chains that can be scaled up efficiently and provide tangible net environmental benefits from renewable energy utilization. Viability of these technologies can be gauged by detailed techno-economic analysis based on rigorous models. Modeling strategic and operational level decision processes of an enterprise which produces renewable-based fuels and chemicals, using engineering and financial tools, can provide valuable insight into the inter-play of technology and process interactions for different product streams, hence providing valuable insights into the correlation between different processing routes which can in turn lead to substantial cost savings for the enterprise.

Several contributions have appeared over the last few years in order to manage the complexity of decision making process for designing profitable renewable energy production systems. Many of the proposed studies in the literature use deterministic modeling approaches which assume that all the parameters are known in advance. However, common to early stages of process design is the lack of certain information that will introduce variability into the decision-making problem.A biorefinery of the future will be plagued by uncertainty in raw material prices and availability, product demands and prices, and technological evolution.

Our work tries to develop a comprehensive decision support tool, within which long and short term decision tasks of an enterprise are optimized under the presence of uncertainties. The proposed methodology is based on an iterative framework that utilizes systems-based strategic planning and optimization in conjunction with detailed mechanistic modelling, simulation and optimization of process. Specifically, this study addresses how uncertainty in different levels of decision making process impacts uncertainty on profitability projections on emerging technologies for energy production. In strategic planning model, the process is formulated as a stochastic mixed integer linear programming (MILP) model which incorporates a stepwise capacity expansion strategy by defining binary variables for capacity increments at each time period in the planning horizon instead of establishing the whole capacity during the first planning year. The model is represented by linear equations for mass and energy balances to describe physical flows of materials across the system nodes and financial flows that result from the system design and material movements. Market uncertainty is incorporated to the decision support framework through stochastic programming in strategic model. A binomial tree is used to generate scenarios for uncertain parameters and a dynamic programming approach is used to evaluate the long-term decisions at each time step. Additionally, to reflect and control the variability of performances associated with each specific scenario, an explicit risk measure is appended and assessed to obtain a solution which reflects the decision-maker’s preference. The output of the strategic model includes optimal design of production capacity of the plant for the planning horizon by maximizing the expected net present value (NPV).

The results are then fed to the second stage of the optimization algorithm. The second stage, which optimizes the operating conditions of the plant, consists of three main steps including simulation of the process in the simulation software (nonlinear modeling), identification of critical sources of uncertainties through global sensitivity analysis, and employing stochastic optimization methodologies to optimize the operating condition of the plant under uncertainty. One of the characteristics of our approach is the incorporation of the complex kinetics of bio-reactions in the simulation model. An iterative dynamic data exchange between process simulation model and developed kinetic models is embedded as part of the process simulation. The final results of the proposed framework include strategic capacity plan, optimal NPV, and optimal operating conditions of the process. The potential process scheme that is evaluated using the aforementioned architecture involves simultaneous production of bioethanol and bio-succinic acid (biochemical pathway) with a centralized utility facility. Succinic acid is produced as a co-product of cellulosic ethanol with utilization of fermentation-related carbon dioxide emissions.