There is no concept of strand in bigWig files, they only hold positions with values associated with them. If you want to have strand-specific bigWig files then you'll need to make a bigWig file for each strand. You can do that with
bamCoverage from deepTools, using the
genomeCoverageBed is using a different understanding of strand than what you want. It's splitting individual reads by their orientation, so since you presumably have paired-end data, the mates from each pair will end up in different files. What you want instead is a tool that will treat mates the same and count them (or not) depending on the orientation of the fragment from which they arose.