Energy systems worldwide require rapid diffusion of granular energy technologies like solar photovoltaics, heat pumps, or electric vehicles. This thesis investigates how to create probabilistic long-term projections of already growing and still emerging granular energy technologies on municipality to global levels under varying policies and contextual conditions. It develops the modeling framework PROWIDE to assess the performance of different modeling set-ups using hindcasting and probabilistically project future diffusion. Investigated projections cover nine technologies, 2,148 Swiss municipalities, 65 countries worldwide, three types of policies, and various extents of historical data. The results highlight the merits of five methodological innovations that improve accuracy and reliability of projections: probabilities, flexible modeling frameworks, model weighting, parametric integration of contextual conditions, and historical analogs. Looking forward, the projections show most often medium to low likelihood of reaching required capacities for net-zero greenhouse gas emissions and call for changes in contextual conditions that extend beyond policies.