Uncertainties in the day-ahead forecasts for load and wind energy availability are considered in a reliability unit commitment problem. A two-stage stochastic program is formulated to minimize total expected cost, where commitments of thermal units are viewed as first-stage decisions and dispatch is relegated to the second stage. Scenario paths of hourly loads are generated according to a weather forecast-based load model. Wind energy scenarios are obtained by identifying analogue historical days. Net load scenarios are then created by crossing scenarios from each set and subtracting wind energy from load. A new heuristic scenario reduction method termed forward selection in recourse clusters (FSRC) is customized to alleviate the computational burden. Results of applying FSRC are compared with those of a classical scenario reduction method, fast forward selection (FFS) by evaluating the expected dispatch costs when the commitment decisions derived from each subset of scenarios are applied to the whole scenario set. In an instance down-sampled from data of an Independent System Operator in the U.S., the expected dispatch costs for both scenario reduction methods are similar, but FSRC improves reliability.