On Comparing Six Optimization Algorithms for Network-based Wind Speed Forecasting
作者单位：School of AutomationSoutheast University Key Laboratory of Measurement and Control of Complex Systems of EngineeringMinistry of EducationSoutheast University
会议名称：《第37届中国控制会议》
会议日期：2018年
学科分类：08[工学] 0807[工学-动力工程及工程热物理]
基 金：supported by Postgraduate Research and Practice Innovation Program of Jiangsu Province
关 键 词：Wind speed forecasting Back-propagation neural network Forecasting accuracy Optimization algorithms
摘 要：Network-based wind speed forecasting has played an important role in the power system. The network parameters optimization is an important issue, and different optimization algorithms are believed to result in different forecasting accuracies. In this paper, six network parameters optimization algorithms, including Gradient descent, Momentum, Ada Grad, RMSprop, Adam, and Adadelta, are implemented and compared in the application of wind speed forecasting. As a case study, this paper uses a wind speed data obtained from Ningxia, China. The performance is evaluated by three metrics, namely, mean absolute error（MAE）, root mean square error（RMSE）, and mean absolute percentage error（MAPE）. The experiment results show that, Adam algorithm and RMSprop algorithm achieve better forecasting accuracy and less training time than the other optimization algorithms. This study can be a guide to the selection of optimization algorithms on wind speed forecasting problems for researchers.