peak signal detection in a time series is non-trivial. It's simple if a window is defined as in this simple one-line solution in numpy or using scipy.signal.find_peaks however, to do this with noisy data without defining a window this becomes very tricky very quickly per here and here
Plotly has a very simple example here and this blog post does an excellent job of explaining now it works.
this extensive study with an accompany
github repo which recommends peakutils