I’d like to model repeating peaks of various periodicity of a time series as a curve. Here’s the general scenario: A device under measurement experiences reasonably regular voltage spikes every N minutes/hours. These spikes vary in amplitude by their periodicity. These spike cycles come and go.
A device may experience a low amplitude spike every 8 minutes and a medium amplitude spike every 40 minutes.
This 8 minute cycle may go on for a few hours/days and then never return.
A new voltage spike cycle may appear at any time. For example, in addition to the 8 and 40 minute spike cycle a 12 minute cycle may appear and continue for a while.
Autocorrelation does a reasonably good job at finding the spike cycles once they appear. Given the spike cycles are of different amplitude and periodicity I’d like to project them forward and visualize them as a curve where peaks would indicate time periods when several coincident spikes are projected to occur. I.e. times of high coincident spike cycle convergence. I can therefore be prepared days/weeks in advance for the voltage spike anomalies.
Question: What would you suggest as effective methods to model and visualize these spike cycles (maybe as a curve) and project them forward in time.