There are only two ways to reduce BQ query cost using only the where clause: partitioning and clustering. Judging by your dataset name, I suspect that your dataset only contains data relevant to March 5th. Unfortunately, BQ only supports day partitioning, so it won't help your case. There is no way to partition by any time block smaller than day. That being said, it may make your data cleaner if you simply create one table and partition it by day instead of creating a new dataset each day.
The other option, clustering, involves specifying a field and providing all potential values. For example, if you had a field "Color" of type String, you could set potential values as "Red", "Green", and "Blue", and therefore, reduce query size to a third.
If you're just looking to reduce cost and find yourself repeating a similar query on a specific table a lot, you can query it once and copy the results to a new table, which would essentially limit your query to a specified time range.
That being said, in general, try to follow BQ's recommendations, such as only selecting columns that you're interested in.