We present a qualitative search for ultra-fast outflows (UFOs) in excess variance spectra of radio-quiet active galactic nuclei (AGN). We analyse 42 sources from the Tombesi et al. (2010) spectroscopic UFO detection sample, and an additional 22 different sources from the Kara et al. (2016) variability sample. A total of 58 sources have sufficient observational data from XMM-Newton EPIC-pn and variability for an excess variance spectrum to be calculated. We examine these spectra for peaks corresponding to variable blue-shifted H- and He-like ion absorption lines from UFOs. We find good evidence for such outflows in 28% of the AGN sample and weak evidence in a further 31%, meaning that $\sim$ 30-60% of the AGN sample hosts such UFOs. The mean and median blue-shifted velocity is found to be $\sim$ 0.14c and 0.12c, respectively. Current variability methods allow for a fast, model-independent determination of UFOs, however, further work needs to be undertaken to better characterize the statistical significance of the peaks in these spectra by more rigorous modelling. Detecting good evidence for variable UFO lines in a large number of sources also lays the groundwork for detailed analysis of the variability timescales of the absorbers. This will allow us to probe their densities and hence distances from the central super-massive black hole.

We present simple XSPEC models for fitting excess variance spectra of AGN. Using a simple Monte-Carlo approach, we simulate a range of spectra corresponding to physical parameters varying, then calculate the resulting variance spectra. Starting from a variable power-law, we build up a set of models corresponding to the different physical processes that can affect the final excess variance spectrum. We show that the complex excess variance spectrum of IRAS 13224-3809 can be well described by such an intrinsic variability model, where the power-law variability is damped by relativistic reflection and enhanced by an ultra fast outflow. The reflection flux is correlated with that of the power-law, but not perfectly. We argue that this correlation is stronger at high frequencies, where reverberation lags are detected, while excess variance spectra are typically dominated by low frequency variability.