Energy system planners and decision makers rely on Energy System Optimization Models in assisting long-term decisions that ensure robust real-world energy system designs that deliver the energy transition goal. Optimization usually provides a single 'optimal' outcome which misrepresents the underlying uncertainties and the large set of possible futures. In this research a method is proposed with which model-owners can be provided insight into the impact of uncertainties on (Energy System Design) Optimization Model outcomes by producing insights regarding the model behavior across model runs under uncertainty. The proposed method consists of three steps: 1) Uncertainty Characterization, 2) Exploratory Modelling, and 3) Results Analysis. The proposed method is applicable to Energy System Design Optimization Models specifically, and to Design Optimization Models in general. This general applicability is constrained to (Energy System) Design Optimization Models where the 'optimized' outcome, the design, can be formatted as a (high dimensional) vector which contains the value for all potential design components, including the zero values, in a fixed ordering over all experiment runs. When applied to an existing Design Optimization Model, this method should help to answer two questions: ‘How do the (Energy System) Designs vary resulting from underlying uncertainties?’; ‘What (Energy System) Design trade-offs are driven by which underlying uncertainties?’. Step 1) consists of the identification of uncertain model-parameters and the assignment of a mathematical representation of their uncertainty. In step 2) the characterized uncertainties are integrated into the model under analysis with the Exploratory Modeling and Analysis (EMA) Workbench. Each exploratory modelling case is composed of n model simulations which results in n experiments. Each experiment represents a unique set of uncertain parameter value combinations and of outcomes containing the model behavior resulting from that specific experiment input space. To answer the first question, clusters of experiments resulting in similar energy system design are identified with a novel approach of cosine distance-based agglomerative hierarchical clustering with complete linkage. To characterize the cluster designs, the total outcome is aggregated to case-design specifics and related back to the underlying uncertainty input with subspace partitioning. The second question is answered with sensitivity analysis and subspace partitioning techniques to characterize design elements that are of interest and to identify trade-offs between these design elements resulting from the underlying uncertainties. As a proof of concept of the applicability and functionalities of the proposed method, it is applied to an Energy System Optimization Model that aims to aid decision-making regarding integrated energy system design and operation in urban areas. This revealed that model-owners can use the insights to: identify model specifications (or the lack thereof) that are determining for the model output, and possibly take measures to limit these effects, and to offer their clients (possibly decision-makers) strategy advice.