Changed delay times of cycle loads to obtain annual demand and cycles

[cprotopa]

I noticed that for cold appliances, which are cycling loads, using the given cycle length and delay (between cycles) resulted in too many cycles in a year, and consequently too high annual demand (compared to reference from Richardson's CREST model). Assuming that what is more important is the annual consumption, and taking as given inputs the power and cycle length, I adjusted the delay time such that the annual consumption fits the desired one.
There is still some variation among households taken into account by sampling the cycle length and delay around the mean values.
This is not the one and only correct way to do it, but I think it's more clean and respects the values given in Appliances.py.

[SilkeVerbruggen]

I see that you both changed the delay time in the appliances.py and in residential.py. It seems to me a bit strange that you increase the delay time in one part and decrease it in the other.
Is it not possible to just increase the delay time in the appliances.py file, in order to get the same amount of cycles? or did I completely misinterpreted this?

[cprotopa]

In Apppliances.py I indeed changed the "nominal" values of the delay, so that on average the annual demand is correct. However, to avoid that all fridges in all houses work exactly the same, the actual delay time for each created household is sampled from a normal gauss distribution with mean the nominal self.delay and standard deviation self.delay/10. The std used to be self.delay/4, and I changed to have a narrower spread, but with no particularly good justification actually (up for discussion, if anybody has a better suggestion).
Furthermore, to avoid that all start approximately at the same time I initiate the starting time that is left as left = random.gauss(delay/2, delay/4), thus somewhere in the middle of the delay period. Also up for discussion, as it's only an assumption.

[SilkeVerbruggen]

Okay I understand. I think the change of 'left' is logical. I just don't know about the change in 'delay' as it seems a bit randomly chosen now. But I understand what you did and I think you can merge this to the main branch.