If you don't need this to be portable to another RDBMS (or another major Postgres version) it is faster and simpler to use the catalog tables in pg_catalog instead of the standard information schema:
SELECT c.confrelid::regclass::text AS referenced_table
     , c.conname AS fk_name
     , pg_get_constraintdef(c.oid) AS fk_definition
FROM   pg_attribute a 
JOIN   pg_constraint c ON (c.conrelid, c.conkey[1]) = (a.attrelid, a.attnum)
WHERE  a.attrelid = '"Schema2"."TableB"'::regclass   -- table name
AND    a.attname  = 'A_Id'                           -- column name  
AND    c.contype  = 'f'
ORDER  BY referenced_table, c.contype DESC;
Returns:
| referenced_table | fk_name | fk_definition | 
| Schema1.TableA | b1_fkey | FOREIGN KEY ("B_id") REFERENCES "Schema1"."TableA"("A_id") | 
 
You only asked for the first column. I added two columns for context.
This returns all referenced tables by all foreign keys involving the given column name - including FK constraints on multiple columns.
The name is automatically schema-qualified if necessary according to the current search_path. It is also double-quoted automatically where needed.
Check out details of pg_constraint and pg_attribute in the manual. More about object identifier types, too.
Related: