This study investigates the effectiveness of a time-frequency analysis of photovoltaic signals with transients caused by shading on the solar panel. The prototype experiment for the signals acquisition was performed by simulating the desired disturbances on the photovoltaic panel. In this direction, the signals were analyzed by a scheme called variable short-time Fourier transform, that has a varying window size over time that is adjusted by local kurtosis given by the fourth normalized cumulant. Thus, a large window size is applied in regions where the kurtosis is relatively high, improving the resolution of time-frequency image. The technique was used on a synthetic signal and current-voltage photovoltaic signals. The results showed that the variable short-time Fourier transform was able to identify transients and classify them over time.