Distribution free Quasi-Recurrent Neural Network for Short-Term Wind Speed Interval Prediction
Forecasting wind speed accurately is an extremely difficult task and less accurate forecasts may sometimes make the power systems vulnerable. Interval predictions on the other hand may help in maximizing the usage of integrated wind energy, as well as to reduce the adverse effects of the uncertainties introduced by the random fluctuations of wind, to the power systems. In this paper, we propose a novel wind speed interval prediction model by integrating prediction interval (PI) characteristic based loss function (QD) into a quasi-recurrent neural network (QRNN). QRNN provides faster training advantage while QD enables distribution assumption free direct PI generation, thus advantageous for real world complex wind data. The proposed model is evaluated for eight different datasets distributed among two wind fields. Our computational experiments reveals that the method generated narrow intervals with high coverage and thus achieved an improvement of 26% in coverage width criterion over the traditional models. Keywords - Wind Speed forecasting, Interval Prediction, Quality-driven loss function, Quasi-recurrent neural network.