prob_joint = normalize(prob_user * prob_usage)

[Mirjam-blokker]

in simulate, you find:
prob_joint = normalize(prob_user * prob_usage)

In the Matlab version I have coded this a bit differently. In case there is a prob_usage that is not only 1's (e.g. for the kitchen tap), this pdf is prevailing. Only when the users are not present, then a correction is made. This means that prob_user is first converted to 0 (not present) and 1 (present), but without the information on peak, night, away included.
Of course when prob_usage is only 1's, then prob_user is prevailing, and the information on peak, night and away is used.

[Yukun-Xie]

This problem can be solved by appling bool calculation.
Besides, the joint probability can be calculated outside the 'for' cycle and normalized in the cycle(reduce the number of operations).