States prediction for solar power and wind speed using BBA-SVM

Abstract

For solar power and wind speed prediction, the uncertainty and randomness of prediction model or parameters make it a challenging task to optimise the accurate output. This study presents a novel backward bat algorithm (BBA) for the parameter tuning of the support vector machine (SVM). Then, the BBA-SVM approach is used to predict the solar power and wind speed output in different situations. The salient feather of the novel BBA-SVM is that an improved flying principle is developed by adopting the backward flying mechanism, which enhances the randomly searching ability and thus avoids the local optimum effectively. Compared to traditional SVM methods, the BBA-SVM gains higher training accuracy, shorter training time, and better prediction performance. Take the solar power output in a sunny day as the validation case, the real data sets from Australia are adopted for comparative simulations, demonstrating the priority of the BBA-SVM against some other SVMs like the grid searching SVM, bat algorithm SVM, and generic algorithm aided BBA-SVM.