To effectively manage a power system, it is important to predict the network power consumption, the peak times, and the peak value. For the wind power system, it is also crucial to analyze wind power and wind speed as a climatic variable power source, which is dependent on weather conditions. Unpredicted changes in wind power source, wind fluctuations, as well as, consumption and peak trends could increase the needs for spinning reserves and raise the production costs. These phenomena can be effectively described by a time series. Wind power times series are generally not stationary, with typically important seasonal components having with inter-annual variability. In non-stationary time series, the mathematical model of the underlying system changes through time. In this work, we have applied a change detection based on Boundary Variation (BV)-clustering on a wind speed time series. The results show the lower frequency variability of wind data through time. The objective of this research is to analyze a non-stationary time series such as wind speed to detect change points between regimes, while the model of time series in each regime is stationary. Change detection in non-stationary time series is a challenging problem as it is mathematically ill-posed. However, with BV-clustering, we can convert the problem into a convex optimization. This method does not rely on Gaussian/Markovian models to analyze the data and finds the optimal number of change points by the well-known information theory criteria.