7 - Wave energy forecasting

author = "Gordon Reikard",

title = "7 - Wave energy forecasting",

editor = "George Kariniotakis",

booktitle = "Renewable Energy Forecasting",

publisher = "Woodhead Publishing",

pages = "199--217",

year = "2017",

series = "Woodhead Publishing Series in Energy",

keywords = "genetic algorithms, genetic programming, Converters, Forecasting, Physics models, Simulation, Time series models, Wave energy, Wave farms",

isbn13 = "978-0-08-100504-0",

DOI = "

doi:10.1016/B978-0-08-100504-0.00007-X",

URL = "

http://www.sciencedirect.com/science/article/pii/B978008100504000007X",

abstract = "This chapter overviews the major developments in wave energy forecasting. The literature on wave forecasting falls into two major groups, physics-based and time series models. Physics models use the energy balance equation, which solves the wave action balance as a function of source and sink terms. In deep water, these include forcing by wind, nonlinear wave-wave interactions, and dissipation by white capping. In shallow water, they also include shoaling and bottom friction. There are several large physics models in operation, WAVEWATCH III, the European Commission for Medium-range Weather Forecasts Wave model, and Simulating Waves Near shore. Time series methods include regressions, neural networks, and newer methods such as genetic programming and artificial intelligence. Comparisons of the two approaches have found that time series models predict more accurately over short horizons, but at horizons beyond the first few hours, physics models are more accurate. The primary measure of wave energy is the flux, a function of the wave height squared and the period. However, in the matrices associated with leading converter designs, the power output is a nonlinear function of the wave height and the period, leveling off above a given threshold, and declining for higher values of the period. The resulting power flow is smoother than the flux. Simulations of wave farms have found that geographic dispersal further reduces the amount of random noise, making the power flow smoother and more predictable than buoy data",