Risk assessment and mitigation strategies for the sustainable development and deployment of CO2 utilization technologies

Abstract

Carbon Dioxide Utilization (CDU) technologies offer realistic solutions for reducing the greenhouse gas (GHG) emissions in the manufacturing sector, which are responsible for 33% of mankind’s carbon footprint. While the current demonstrations of CO2 conversion have so far been technically promising, from the industrial point-of-view, there exists significant risk-based barriers that hamper wider-spread adoption. In the development of early-stage CDU technologies, generally weak knowledge-bases and high technological uncertainties mean that sustainability with respect to economic and environmental performance criteria cannot be guaranteed. Consequently, the decision to best allocate limited R&D resources is both highly risky and non-trivial. For the commercial deployment of late-stage technologies, constantly changing market dynamics, dependence on renewable energy, and the lack of clear climate policies and regulatory frameworks mean that the long-term feasibility is highly dubious. While capital investments costs are high and are incurred “here and now”, uncertain profits are only accumulated gradually throughout the technology lifetime.

To better facilitate industrial adoption, this dissertation proposes strategies that can assess and hedge risk in the development and deployment of CDU technologies. In the methodology for risk-based uncertainty assessment for early-stage technologies, key sustainability hurdles are identified via Bayesian classification, based on stakeholder-set sustainability decision rules. Class likelihoods are used to quantify the risk, and the method can be scaled to two or more sustainability criteria via Bayesian decision networks. This method addresses the pitfall of conventional GSA-based methods, which strictly analyzes variance contributions and is limited to a single output. The methodology was applied to two CDU technologies that are being developed as part of the Carbon-to-X R&D project: CO2 hydrogenation to formic acid and CO2 biofixation to algal bioplastic feedstocks. In each application study, key risk-based hotspots were identified and sustainable development strategies have been proposed.

For late-stage commercial deployment, a multi-period deployment model combining real options theory (ROT) and reinforcement learning (RL) is proposed (ROT+RL). Under this formulation, stakeholders are able to exercise flexible options such as sequential plant expansion (Option to Expand) and delayed start (Option to Delay). By being able to deploy in multiple periods in response to future uncertainties, the formulation itself is risk-averse. On the other hand, conventional valuation methods are mostly single period, which conducts a discounted cash flow analysis of the CDU project and makes a go/no-go decision based on the net present value (NPV). As part of the proposed ROT+RL model, methods for characterizing time-variant uncertainties using stochastic processes in discrete time, and learning algorithm selection based on two different sequential expansion approaches are also discussed. The multi-period deployment methodology was applied to the CO2 hydrogenation to methanol process. For the baseline scenario consisting of 8% risk-adjusted discount rate, 10 exercisable periods for real options, and a CO2 credit on saved emissions of 80 €/ton CO2eq, the results showed that 37% higher project value was achieved with a 58% higher final installed. Simultaneously, a higher cumulative saved CO2 emissions was achieved.