find-peaks | Algorithm to find peaks in a std : :vector float | Learning library

I am trying to take my time series data and isolate all data between points of 0, and then identify those intervals with the highest peaks. I am working in python.

Referring to this graph: time series data with peaks and valleys identified

Source: https://tcoil.info/find-peaks-and-valleys-in-dataset-with-python/

Noting that the first and last red valley points are at 0, I want to find a way to take time series data, identify all points at 0 on the y-axis, and then isolate the data in between. For the graph I linked to here, I would want to isolate all data between the first and last red valley point. I want to do this across an entire time series data set, where data between points of 0 on the y-axis are isolated. Now that those intervals are isolated (representing different events/cycles throughout the data), I want to record the highest point within each of these intervals. Then I want to find the intervals with the 5 highest peaks (one peak per each interval). Lastly, I want to output the interval (or range) that contains these top 5 peaks. For context, each of these intervals represents an event/cycle, and I want to find the most extreme. As such, I would want an output that tells me essentially that the most extreme event/cycle occurred between 3/5/20 and 3/24/20.

How can this be done in python? Would I need to smooth the data first? How would I go about isolating data between points of 0 on the y-axis? I am trying to figure out which direction to go in first, and do not have code yet.