Time series processing is an essential aspect of wind turbine health monitoring. Despite the progress in this field, there is still room for new methods to improve modeling quality. In this paper, we propose two new approaches for the analysis of wind turbine health. Both approaches are based on abstract concepts, implemented using fuzzy sets, which summarize and aggregate the underlying raw data. By observing the change in concepts, we infer about the change in the turbine's health. Analyzes are carried out separately for different external conditions (wind speed and temperature). We extract concepts that represent relative low, moderate, and high power production. The first method aims at evaluating the decrease or increase in relatively high and low power production. This task is performed using a regression-like model. The second method evaluates the overall drift of the extracted concepts. Large drift indicates that the power production process undergoes fluctuations in time. Concepts are labeled using linguistic labels, thus equipping our model with improved interpretability features. We applied the proposed approach to process publicly available data describing four wind turbines. The simulation results have shown that the aging process is not homogeneous in all wind turbines.