This study presents a Bayesian network meta-model approach for reasoning over uncertainty in techno-economic assessments of early-stage commercial-scale fusion pilot plants. Bayesian networks as meta-models offer distinct advantages, such as bi-directional inference and a probabilistic representation of knowledge under uncertainty. In recent work we presented a Bayesian network meta-model framework to aid the nuclear fusion community by facilitating decision making over uncertain knowledge. In this paper, we present an application of our framework to an applied industry case study for fusion developer Tokamak Energy. Through bi-directional reasoning, our results identify the feasible regions for plasma physics and engineering parameters that minimise capital expense and maximise heat and electricity production, meeting criteria set for the US Department of Energy Fusion Development Program. Using the Bayesian network meta-model insights, Tokamak Energy avoid making deterministic decisions, and can optimise engineering design points and allocate resources towards their pilot plant concept without relying purely on assumptions. The decision-making approach can generalise across analysis codes in the fusion community.
This paper presents a Bayesian network approach to aid decision-making for fusion pilot plant design. It allows reasoning over uncertain parameters to optimize engineering and plasma physics for cost-effective heat and electricity production, meeting criteria for the US Fusion Development Program.