Significant wave height and energy flux estimation with a Genetic Fuzzy System for regression

This paper proposes a regression Genetic Fuzzy System (GFS, FRULER) for a problem of sea wave parameters estimation from neighbor buoys, with application on wave energy systems. FRULER is a recently proposed, three-staged algorithm that combines an instance selection method for regression, a multigranularity fuzzy discretization of the input variables and an evolutionary algorithm to generate accurate and simple Takagi-Sugeno-Kant (TSK) fuzzy rules. We have applied FRULER to a real problem of significant wave height and energy flux prediction in one buoy of the West Coast of the USA (California), from values of other two neighbor buoys. In the case of the significant wave height, FRULER is able to obtain a robust prediction with only three rules, which in addition are fully interpretable, since they clearly separate swell situations from wind-sea in the prediction. In both cases, the variables used in the significant wave prediction are completely different and can be identified as relevant for the specific case (swell or wind-sea). In the case of the energy flux prediction, the interpretation of the rules provided by FRULER is more difficult, since eight rules are necessary to obtain the prediction. Even in this case, several rules can be clearly classified as swell predictors, and the rest of the rules describe local wind situation of waves. This study shows that the GFSs are useful tools to obtain robust and interpretable predictions in ocean wave parameter estimation problems.