Electricity demand and renewables penetration are set to increase worldwide over the coming decades as part of the global decarbonisation effort. As a result, distribution networks are expected to face challenges related to increased peaks and undesirable voltage excursions. Hence, significant network reinforcements may be required over the next decades. However, a very significant challenge in realizing this transition is the increased uncertainty that surrounds future distributed generation and load connections in terms of size, location and timing. This uncertainty inadvertently will give rise to the prospect of inefficient investments and stranded assets given that current planning practices remain deterministic. It follows that new planning frameworks are needed that allow the quantification of option value and achieve reduction of stranding risk by encouraging cost-efficient strategic investments through smart technologies under both endogenous and exogenous sources of uncertainty.
This thesis presents multi-epoch stochastic optimization models, for the distribution network planning problem, that consider a set of investment options with different techno-economical characteristics so as to reflect the multitude of choices available to planners in a realistic setting characterized by endogenous or exogenous uncertainty. These optimization models are rendered tractable through the use of novel decomposition schemes that effectively help manage the associated increased computational burden. The corresponding simulation results validate that smart technologies constitute valuable options for enabling cost effective integration of distributed generation units and underline the importance of early investment in such assets under decision-dependent uncertainty. In addition, the results emphasize that deterministic approaches systematically undervalue the flexibility that smart assets provide, thereby posing a barrier to the advent of the flexible smart grid paradigm.