(398e) Designing Robust Energy Policies for Low-Carbon Technology Adoption

Increasing the adoption of alternative technologies is vital to ensure a successful transition to net-zero emissions in the manufacturing sector. Yet there is no model to analyse technology adoption and the impact of policy interventions in generating sufficient demand to reduce cost. Such a model is vital for assessing policy-instruments for the implementation of future energy scenarios. The design of successful policies for technology uptake becomes increasingly difficult when associated market forces/factors are uncertain, such as energy prices or technology efficiencies. In this work we formulate a novel robust market potential assessment problem under uncertainty, resulting in policies that are immune to uncertain factors. We demonstrate two case studies: the potential use of carbon capture and storage for iron and steel production across the EU, and the transition to hydrogen from natural gas in steam boilers across the chemicals industry in the UK. Each robust optimisation problem is solved using an iterative cutting planes algorithm which enables existing models to be solved under uncertainty. By taking advantage of parallelisation we are able to solve the nonlinear robust market assessment problem for technology adoption in times within the same order of magnitude as the nominal problem. Policy makers often wish to trade-off certainty with effectiveness of a solution. Therefore, we apply an approximation to chance constraints, varying the amount of uncertainty to locate less certain but more effective solutions. Our results demonstrate the possibility of locating robust policies for the implementation of low-carbon technologies, as well as providing direct insights for policy-makers into the decrease in policy effectiveness that results from increasing robustness. The approach we present is extensible to a large number of policy design and alternative technology adoption problems. Figure 1 provides an overview into the work.

Technology investment cost distributions from numerous engineering studies of nuclear, biomass, solar and wind technologies demonstrate that technology change and learning effects are highly uncertain. The presence of uncertainty makes the availability, adoption, and diffusion of alternative technologies slower and more discontinuous than would be if these decisions were made using nominal models. In practice, stakeholders passively seek to realise unknown uncertainties by waiting for them to instantiate, resulting in slower adoption. Traditional deterministic optimisation and social planning models have been criticised for being optimistic in their handling of uncertainties and technologically too pessimistic by neglecting reduction in technology cost over time from increased demand and spillover effects. These stated downsides have contributed to an increasing number of nonlinear stochastic and robust problem formulations to account for these known uncertainties.

The contributions of this work are three-fold. We first present novel robust market potential assessment model to study and analyse technology adoption in the manufacturing sector considering uncertainty in all the model parameters and considering important effects such as technology learning. The robust market potential assessment model is the first to exploit how market size together with policy interventions can create sufficient demand to trigger and sustain technology cost reduction to accelerate uptake. Secondly, we apply a state-of-the-art iterative robust optimisation algorithm and subsequently investigate robust solutions to the model, which immunise a solution against all aspects of uncertainty. This approach enables us to maintain the original model formulation, including nonlinear effects such as the learning rate of a technology, whilst providing an approximate robust solution.

We investigate the effect of worst-case uncertainty on the model, providing conclusions regarding the adoption of an alternative technology such as Carbon capture and storage (CCS), and transition from natural gas to hydrogen under different scenarios. Finally, by making assumptions as to the form of uncertainty we are able to reduce the conservatism of the robust solution at the expense of increased risk. The insights gained by reducing the conservatism of the problem provides direct insights for policy makers, including the probability a solution will be made infeasible under some assumptions. The approach we present results in an approximate solution to worst-case uncertainty for a given nonlinear model. We believe this approach is more beneficial than making simplifying assumptions to gain an exact solution to worst-case uncertainty, particularly for technology adoption related policy optimisation. Our approach is general enough to be extended to different technologies and fuels.

CCS is crucial for emissions abatement in iron and steel production. CCS is crucial to achieving carbon neutrality targets, however, the high capital, operation, and maintenance costs as well as the large uncertainties of oil or methanol prices hinder its development. Integrated steel making using a blast furnace and basic oxygen furnace are the most common routes to make steel. In this case study we investigate the market effects on the uptake of carbon capture and storage within the iron and steel industry. The policy instrument designated as decision variables is the carbon tax. The objective is to minimise the difference between the existing cost of capture, and the new cost of capture determined from increased demand in CCS due to policy intervention. The final nominal problem contains 66 constraints, 64 variables, and 36 uncertain parameters. The data collected for this research includes the cost of capture for CCS technologies applicable to integrated steel plants, the emissions reduction potential of these technologies, the number of integrated steel plants in the EU, the location of each plant, the steel production capacity of the plants, and the carbon intensity of steel. A post combustion capture from blast furnace with 63% capture rate using 3% MEA solvent has been demonstrated for use in integrated steel plants with TRL 6 – 8, though this capture rate is generally not precisely known. Post-combustion involves removal of carbon dioxide from flue gas generated from a combustion process. It is usually carried out through the chemical or physical absorption of carbon dioxide into a solvent. In this case study we present 32 iron and steel plants across the EU using MEA Post Combustion capture from blast furnace with an existing cost of capture ranging from 59.6- 116.3$/t.

Rather than make simplifying assumptions within the model to enable reformulation, we take the view that for this case, maintaining the original model and applying an approximate approach for robust optimisation to immunise solutions against uncertainty provides a greater value. To solve the robust market potential assessment problem, we apply the Blankenship & Faulk algorithm. Algorithm 1 presents this approach.

This method allows us to maintain the original form of the market potential assessment model, important for considering nonlinear terms under uncertainty such as learning rate. Algorithm 1 provably converges to the robust solution of a problem, under the assumption that the lower-level problems are solved globally. This ensures that robust constraints are not violated, regardless of the value of uncertain parameters. As global optimisation solvers do not scale well for large problems, we do not solve lower levels globally, and instead, use a local solver with a multi-start heuristic to solve pessimisation sub-problems. While this provides no guarantees on the optimality of the subproblems, it gives us tractability in solving these real-world formulations. The subproblems solved between lines 5 and 11 in Algorithm 1 may be solved in parallel enhancing the performance of the algorithm. In some large problems, the number of constraints added to the upper-level problem becomes unsustainable, and redundant constraints must be `dropped' from the upper-level problem to ensure its tractability. However, in the current work, we find this unnecessary. For illustration purposes, Figure 2 demonstrates the behaviour of the Blankenship & Faulk algorithm applied to a 2-dimensional robust optimisation problem with a single semi-infinite constraint.

Figure 3 demonstrates how the level of uncertainty of each individual parameter affects the robust solution quality. We subsequently change the nature of uncertain parameters, enabling us to trade-off solution feasibility with quality. Here we provide the ability for policy-makers to choose an optimal policy design entailing their perspective of risk. Figure 4 demonstrates the outcome of trading off solution quality with risk. In this work we provide results and analysis across both case studies, providing recommendations for practitioners, and decision-makers in dealing with nonlinear uncertain energy policy problems.

We present a methodology for reducing the conservatism of robust future energy policy models. We demonstrate how our methodology can result in significantly better solutions that are statistically likely to remain feasible. Our results demonstrate the possibility of locating and communicating robust policies for the implementation of low-carbon technologies, as well as providing direct insights for policy-makers into the decrease in policy effectiveness that results from increasing robustness.