Wind Speed Prediction with Spatio Temporal Correlation using Convolutional and Spiking Neural Networks

Abstract

Due to inherent intermittency, reliable wind speed prediction with space-time correlation is a crucial task for several practical applications, including grid operation and energy trading. This paper aims to address this problem through the use of Artificial Neural Networks (ANNs). The proposed model is an integration of Convolutional Neural Networks (CNNs) and Spiking Neural Networks (SNNs) to obtain space-time characteristics of wind speed. CNNs are utilised for capturing the spatial features of the data and SNNs are used to find the relationship among wind speed and the learnt spatial features. Experimental results on real data demonstrate that the underlined model efficiently retrieves space-time correlation. It performs better than the traditional models including machine learning techniques, such as multilayer perceptron, decision tree and support vector regression. The prediction horizon for testing ranges from 5 minutes to 1 hour and the proposed model shows improvements for all tested horizons compared using two measures, namely the array mean absolute percentage error (A-MAPE) and array root mean square error (A-RMSE).