基于RBF神经网络的铁路沿线短时风速预测方法

Prediction Method for Short-Time Wind Speed along Railway Based on RBF Neural Network

对实测风速数据进行Kalman滤波,去除实测风速数据的偏差;通过归一化处理,消除数据中的冗余成分;针对RBF神经网络的预测误差会随着时间的推移而增大的问题,采用滚动式训练方法在线训练RBF神经网络;用训练好的RBF神经网络进行风速预测,再对预测结果进行反归一化处理,得到最终的预测风速。仿真结果表明,运用基于RBF神经网络的铁路短时风速预测方法对短时风速进行预测,最大相对误差仅为5.92%,可满足铁路防灾安全监控系统中风速预测子系统的要求。

The measured wind speed was processed with Kalman filter algorithm to eliminate deviations.The redundancies in the measured data were removed through normalization processing.Then,RBF neural network was online trained by using the rolling training method to deal with the problem that the prediction error of RBF neural network would increase as time went on.Finally,the wind speed was predicted by using the well-trained RBF neural network.The final forecasted wind speed was then obtained by anti-normalizing the output of RBF neural network.The simulation results show that the maximum relative error is only 5.92% using the proposed railway short-time wind speed prediction algorithm based on RBF neural network,which can satisfy the requirements of the wind forecasting subsystem in railway disaster prevention and safety monitoring system.