Short-term Interval Prediction of Wind Speed Using LSTM with Improved Grid Search

As a clean and inexhaustible renewable energy, wind power attracts more and more attention. However, the intermittence and randomness of wind speed make the utilization of wind power challenging. These troubles can be alleviated via wind speed forecasting that can provide comprehensive future information about wind speed uncertainties. Compared with traditional point forecast methods, interval forecast is able to quantify uncertainties in renewable energy production effectively. In this paper, the idea of prediction intervals (PIs) is employed to capture the uncertainties of wind speed. Specifically, a novel interval forecast model based on long short-term memory (LSTM) neural networks is proposed to construct PIs with the lower and upper bound estimation (LUBE) method, which is a powerful nonparametric forecast approach. To tune the parameters of LSTM prediction model, the improved grid search algorithm is introduced as the optimization tool. The effectiveness of the proposed model and algorithm is demonstrated by a series of experiments based on a real world wind speed dataset, and the comparative results show the superiority of the model.