The increased penetration of Distributed Energy Resources (DERs) within power networks is bringing challenges, an important one being the potential voltage excursions within the system that must be mitigated, as voltage must be maintained within statutory range at all times and at any node of the system by regulation. This paper proposes a scalable framework based on machine learning techniques (ML) to assess voltage excursion risks node by node and derive the related marginal probabilities in response to any net-loads under various DER penetration scenarios. The framework is then used to quantify the resulting financial impact of voltage excursion in large-scale networks. Therefore, this novel end-to-end risk framework supports decision making in the planning phase of networks in response to any intermittent DER penetration scenario. We show through simulations that the framework is both scalable to high-dimensional systems and efficient to handle vast number of scenarios. In our simulations, the use of ML technique enables to lower the computing time by a factor of 800 compared to load flow solving, while maintaining an accuracy ≥ 95%, enabling the assessment of vast number of scenarios.