Hello
I'm afraid that for a waveform that varies randomly (or cvasi-randomly) in time, the only solution to define its consumption would be to make averages of RMS values over large periods of time, as FvM said, such as pulse-to-pulse, but even that can't be set as a definite reference, only as a guess. For example, in the picture you provided you can set a current RMS period of 200ms, do reading (calculations) for, say, 50 values or 10sec, then average the readings to get an approximation of what you want. Of course, the longer the time-span, the better the approximation.
Vlad