Mathematics can supply a bit of analysis before brute force computation is required.
We can count the number of problems easily enough. If the digits are all distinct (and nonzero), then there will be $C(9,4)$ of those, so all distinct digit problems number 126. Note that this considers the order of the digits unimportant since you seem to all their rearrangment in your Question above.
It remains only to count the problems which involve a certain number of duplicate digits. For example the same digit can occur four times in just 9 problems of that type. One digit occurring three times with a different fourth digit accounts for 9*8 = 72 problems.
Otherwise some digit occurs twice (but no more than twice). If the problem has two "twin" digits, it can happen in 72/2 = 36 ways. If the problem has one duplicate pair and two other distinct digits, this happens in 9*8*7/2 = 252 ways.
Totaling these possibilities gives 126+9+72+36+252 = 495.
I see no shortcut to deciding apriori how many of these have an admissible way of expressing 24 (using just the four basic arithmetic operations and parentheses). Certainly a program could be written to explore all the possibilities. For this purpose we can dispense with any question of parentheses by using Polish notation (evaluating such by a stack order of operation to see if 24 is the result). Note that with rearrangements we are back to considering how many ways the four arithmetic operators can be inserted qua Polish notation among any four single digit operands.
We can identify five possible sequences of operators and digits that are allowable. In the strings below, each
A represents a place for an arithmetic operator and each
D represents a place for a digit.
That is, if three binary operators are used to combine single digit operands, then only these five patterns are feasible to get a single result. All five patterns admit $4^3 = 64$ substitutions for the operators and $9^4 = 6561$ substitutions for the digits. Therefore the total number of expressions to check is at most 2099520. I say "at most" because some pruning will naturally occur as a result of finding solutions for problems without checking all possible expressions for a given set of digits. Further some of the patterns are equivalent because of the commutativity of addition and multiplication, but it would likely take magnitudes more time to program for any reduction in search space on that basis than it would to just exhaust the five patterns above