As more renewable resources are integrated into the transmission system, and the power system operates closer to its capacity, congestion conditions become less predictable and locational marginal prices more volatile. The increased congestion and price uncertainties pose significant challenges to system operators and market participants. This research aims to develop new computationally tractable techniques for short-term probabilistic forecasting of real-time locational marginal price in a power grid. In particular, the developed methods provide the probability distribution of the locational marginal prices and congestion patterns up to several hours ahead of the operation time. For system operators and market participants, such techniques can be used for congestion management, system planning, risk management, and demand response.