The document presents a method called Constrained Support Vector Quantile Regression (SVQR) for conditional quantile estimation and probabilistic forecasting. SVQR formulates quantile regression as a support vector machine problem with non-crossing constraints. It solves the dual problem, which is a quadratic program, to obtain quantile estimates. The method is tested on wind power forecasting using features from wind speed and direction data. Results show SVQR outperforms benchmark persistence, climatology and uniform distribution models. Future work on combining SVQR with neural networks for improved feature learning is also discussed.