基于回声状态网络的海浪谱实时预报方法

Real-time forecast method of wave spectrum based on echo state network

为实现在非线性较强情况下的海浪谱预报,以船载测波雷达实船测量的海浪谱数据为基础,将经验模态分解(EMD)和回声状态神经网络(ESN)相结合,对海浪谱的实船实时预报方法进行了研究。所提方法利用经验模态分解对子波谱密度值时间序列进行分解,并对分解后各分量分别应用回声状态神经网络进行预报。将预报结果进行叠加,可以得到子波谱密度的预报值,进一步可合并得到整个海浪谱信息。结果表明:该方法可以有效解决非线性较强情况下预报效果变差的问题。方法可为船舶实时掌握海浪谱信息,提高船载测波雷达系统的实用性提供一定的基础。

In order to realize forecast of ocean wave spectrum in relatively strong-nonlinearity, empirical mode decomposition (EMD) and echo state network (ESN) was combined to study on real time forecast method of wave spectrum, with the spectrum data measured by onboard X-band radar. EMD was used to decompose the measured nonlinear time series of wavelet spectral density value, and the decomposition results can forecast using ESN. Forecast results can be summed to reconstruct forecast result of different wavelets, and then the frequency spectrum. Results showed that, this method can solve the problem that the forecast results getting worse in relatively strong-nonlinearity condition. This method can provide the basis for people onboard know the real-time data of wave spectrum and for improving practicability of X-band radar.