Accurate modeling of the wind power curve is crucial for estimating production capacity, furthering proper planning, abnormal state detection, and integration with the power system. For such, data-driven approaches appear as appealing alternatives. However, wind-power turbine performance is susceptible to uncertainties and non-linearities, and although several techniques have been explored, none was established as more advantageous in all cases, whereas each can suffer from data pollution, insufficiency, or poor conditioning. Therefore, adjudicating the most suitable model for each specific application requires careful consideration. Among the most prevalent strategies in the literature, artificial intelligence-based approaches often achieve better results, but curve-fitting solutions can obtain comparably good models with lesser development complexity. This work studies piecewise linear fitting and machine learning-based methods. An automatic data filtering procedure based on median absolute deviation patterns also is used. Compared simulations use datasets for the case of interest, related to a project aimed at building a hybrid solar-wind power plant, with integrated battery storage features, by the re-purposing of one generating unit operating in a previously existing Brazilian wind farm. Both approaches yielded useful results, and although the machine learning application outcomes had a wider range, they typically presented around 10% better error metrics overall.