Context. The identification of proton core, proton beam, and alpha particles in solar wind ion measurements is usually performed by applying specific fitting procedures to the particle energy spectra. In many cases, this turns out to be a challenging task due to the overlapping of the curves. Aims. We propose an alternative approach based on the statistical technique of clustering, a standard tool in many data-driven and machine learning applications. Methods. We developed a procedure that adapts clustering to the analysis of solar wind distribution functions. We first tested the method on a synthetic data set and then applied it to a time series of solar wind data. Results. The moments obtained for the different particle populations are in good agreement with the official data set and with the statistical studies available in the literature. Conclusions. Our method is shown to be a very promising technique that can be combined with the traditional fitting algorithms in working out difficult cases that involve the identification of particle species in solar wind measurements.
