Employing Bayesian networks for energy transition pathways across Swiss Cities, Midlands, and Alps

Abstract In the context of Switzerland’s ambitious Energy Strategy 2050, solar PV is poised to play a central role. However, traditional energy modelling approaches often overlook the social dimension underlying solar PV adoption. This paper demonstrates the potential of using Bayesian Networks (BNs) to integrate the social dimension into energy modelling. BNs can handle uncertainty, incorporate population-level survey data, and simulate dynamic adoption scenarios. Using survey responses from 4,466 individuals across Switzerland, we construct data-driven BNs through structure learning and parameter learning across three regions, namely the Swiss cities, midlands, and Alps. While the structure of the networks remains consistent, the conditional probabilities reveal region-specific adoption dynamics for solar PV adoption In cities, heat pump ownership emerges as the dominant factor, while in the Midlands adoption reflects a balanced mix including electric car ownership and perceived economic viability. In the Alps, single-family house and electric car ownership are most influential. We also tested three scenarios. Scenario 1 generated the largest increases, especially in the Alps and midlands. Scenario 2 produced a modest increase, with the strongest relative effect in the midlands, while scenario 3 was most impactful in cities. We demonstrate that Bayesian Networks are useful for modelling socio-technical energy transitions. Our findings support the need for targeted, regionally differentiated policy interventions to accelerate solar PV adoption in Switzerland.