A Tidal Current Speed Forecasting Model based on Multi-Periodicity Learning

TL;DR

This work tackles the challenge of forecasting tidal current speed by modeling multi-periodicity and local periodicity using a Wavelet-Enhanced Convolutional Network (WECN). It integrates discrete wavelet transform-based period extraction with a TimesNet-inspired 2D convolutional backbone and adaptive aggregation, augmented by Tree-structured Parzen Estimator hyperparameter optimization. Empirical results on Orkney tidal data show that WECN achieves state-of-the-art 10-step forecasts (MAE ≈ 0.025) and robust performance for longer horizons, demonstrating improved stability and accuracy over a diverse set of baselines. The approach offers a practical path to higher penetrations of tidal energy by providing more reliable short- to mid-term tide forecasts and a generalizable framework for multi-periodicity time series.

Abstract

Tidal energy is one of the key components in increasing the penetration of renewable energy. High tidal energy penetration into the electrical grid depends on accurate tidal current speed forecasting. Model inaccuracies hinder forecast accuracy. Previous research primarily used physical models to forecast tidal current speed, yet tidal current variations influenced by the orbital periods of celestial bodies make accurate physical modeling challenging. Research on the multi-periodicity of tides is crucial for forecasting tidal current speed. We propose the Wavelet-Enhanced Convolutional Network to learn multi-periodicity. The framework embeds intra-period and inter-period variations of one-dimensional tidal current data into the rows and columns, respectively, of a two-dimensional tensor. Then, the two-dimensional variations of the sequence can be processed by convolutional kernels. We integrate a time-frequency analysis method into the framework to further address local periodic features. Additionally, to enhance the framework's stability, we optimize the framework's hyperparameters with the Tree-structured Parzen Estimator. The proposed framework captures multi-periodic dependencies in tidal current data. Numerical results show a 10-step average Mean Absolute Error of 0.025, with at least a 1.18% error reduction compared to other baselines. Further ablation studies show a 1.4% reduction in Mean Absolute Percentage Error on the data with artificially added periodic fluctuations.

Figures (8)

• Figure 1: Evaluation of Levelized Cost of Energy (LCOE) and Cumulative Installed Capacity in the UK. ±1 SD indicates data variation within one standard deviation around its expected value.
• Figure 2: Flowchart of the Period Extraction Module, illustrating the process of transforming multivariate time series inputs into the main period ($P$) and frequency ($f$).
• Figure 3: The WECN captures intra-period and inter-period variations and uses 2D convolutional kernels to process both types of variations. The time series case in the figure is selected from the used tidal current speed dataset.
• Figure 4: Overall architecture of WECN.
• Figure 5: The multi-step forecasting process of WECN.