A Precise Tidal Level Prediction Method Using Improved Extreme Learning Machine with Sliding Data Window

Abstract

To improve the tidal prediction accuracy, a prediction scheme is proposed by using improved Extreme Learning Machine (IELM) based on a sliding data window. The changes of tidal level are complex processes which are influenced by not only the movement of celestial bodies but also the non-periodic meteorological factors such as wind, air pressure, and water temperature. Harmonic analysis is a traditional tidal level analysis and forecasting method. However, it cannot take into account the impact of time-varying factors. Extreme Learning Machine is a single-hidden-layer feed-forward network (SLFN) with extremely fast learning speed and good generalization performance. The algorithm of ELM has the good fitting ability for nonlinear processes. In order to achieve tidal level prediction with high accuracy, the model of improved extreme learning machine with sliding data window (SDW) is proposed, referred to as Improved Extreme Learning Machine with Sliding Data Window (IELM-SDW). The SDW is used to reflect the current changing dynamics of the tidal level. The improvement of ELM applies temporal difference (TD) learning algorithm. According to the actually received information, adjust the forecasting results. Measurements of tidal level data from Adak, Atka and Juneau are used for the simulation experiment. Comparison simulations are conducted with approaches of the conventional ELM and harmonic analysis. Comparison results demonstrate the effectiveness of the proposed IELM-SDW.