Study, find books with endgame puzzles. If you receive the uscf magazine, then check out Benko's Bafflers, or something like that, those are good puzzles. So yeah, studying from different books, and working with endgame puzzles. Sometimes you just need to think of a strategy to win/draw, or put up a good fight.
Edit: Also, don't always trust a chess engine with endgames! They're not always accurate.
How to improve my endgame play?

Well for example lookup games in the databases with similair pattern's/positions. What plan do strong players use in the the similair situation/position.
Another effective way which i use to analyse my own otb games is to take a critical position in my end game, and play the same position a few times against a computer with a different plan/idea. Afterwards i analyse what did work and didn't and try to come to conclusions why it did/didn't work. Then test your conclusions with a chess engine to see if you were right/wrong.
Finally put those important end game positions in your personal database so you can look/play them again in the future and remeber them.

Getting books like Silman's or Dvoretsky's is good, and practicing against a person or computer is also very good. For example technical things such as the defensive set up for rook and king vs rook king and pawn. When the pawn is on your 4th rank do you know the defense, if your rook can't get to the 3rd rank do you know that one, and it changes if the pawn can advance to the 3rd, do you know the conditions of a successful defense then? It can be tedious but a book and practice pay off.
Other than that though you seem to be asking about practical endgames (" . . . but in practical games I need help") so other than technique and specific positions that you'll find in a book (which are important) I'd like to try and answer that.
First of all, it's hard to overestimate the importance of activity. In endgames, because of the reduced material, activity is much more important than it was in the opening or middlegame. For example if your knight is out of play in the middle game it may hurt you a bit, but in the endgame that may be as much as half your total force. And so at this time more than ever, losing a few pawns here or there will take a back seat to activity (most famously in rook endgames). The more valuable the piece, the more important it is you keep it active and by the way in this sense (in general) rooks > the king > minor pieces. So if have a choice of which piece you'll use for passive defense (such as a blockader) in general you'll chose a king over a rook and a minor piece over a king. Therefore if you're king is stuck defending you'd want to replace it with a minor piece and if you can tie a the enemy's king up with your bishop then the remaining two pieces (your king vs their bishop) will give you almost a winning advantage (again in general).
So although I'd say zugzwang is a main theme of endgames, tying pieces to defense is a close second. So other than activity you have to be acutely aware of targets. Fixing a pawn target on a4 (by playing a5 as black) is a fundamental idea. Trading pawns in such a way that the enemy pawns become split (isolated instead of one pawn island) is another.
With pawns we can talk about both targets and in relation to mobility. It took me a while to realize that in general, for the endgame pawns prefer to be on their home square for two reasons. The first reason is the farther advanced they are, the more they're subject to attack.
In this diagram we see black has an ideal position as the base of his pawn chains, his anchor pawns, are still on his 2nd rank making them much easier to defend. In an endgame the further up your pawns, the greater the liability for their defense. In this bare bones structure, white's weakness on d4 puts him at a disadvantage.
The 2nd reason pawns like their home square is one of the primary themes of any endgame which is zugzwang. Pawn on their home square can burn a tempo as necessary (by choosing to jump two spaces) to pass the move to the opponent. The more pawns you have with this option, the more likely you'll be able to force a zugzwang.
In the following position regardless of who is to move, white will easily win as black will be put in zugzwang. This position contains a general tid bit I also found not mentioned in endgame books (maybe too obvious to mention, but a nice rule of thumb for those that don't know) which is if a king can get on the same rank as a loose enemy pawn, then that pawn is lost. White to move will give the move back with 1.h3 h5 2.h4 while if it's black to move after 1...h5 2.h4 wins. If there were more pawns whoever has the most still on their home squares would have such an advantage.
So it follows when choosing between moving a pawn on it's original square or one already moved, save your home pawns. Some of this may already be obvious to you, but you'll find many novices when they're feeling lost in an endgame (or right before entering an endgame) needlessly hurt their position by moving pawns.
So pieces activity, (a bishop tying a king to defense, everything else being equal, can be a winning advantage) pawns as targets (keep your anchors as far back as possible) and pawns as extra tempo (pawns on home squares) as well as zugzwang are some of the basics of endgame play for any endgame position.

Thanks, that was very interesting, especially about the unmoved pawns being an advantage in producing zugzwang.

Since you like attacking chess, you should just mate them in the middlegame and avoid all that endgame nonsense.

Since you like attacking chess, you should just mate them in the middlegame and avoid all that endgame nonsense.
I try!

OrangeHonda is righht, like i try to do, calculate 6-10 moves ahead to see if i can force zugzwang before my opponent, in some endgame situations you have to calculate far.. but its far easier calculating in endgame then middlegame (middlegame i usually calculatel ike 3 moves)

What you need is an endgame STRATEGY book. Puzzle books and books like Silmans endgame course only deal with TECHNICAL endgames which is a whole different thing. Ofcourse you will need to know the basic technical endgames like R+pawn vs R etc so you know what to aim for in your endgame strategy. Unlike mastering technical endgames it can take time to learn playing strategical endgames because there are no clear cut plans and rules like in technical endgames how to convert an advantage. Check out some good endgame strategy books and play out some typical ones and you'll get better at it.
Im just like you, i mess up the endgame but im very good mid gamer. The key is to know when to transform from a mid game to a end game that favors you. I have lost so many games because i never transformed to a good endgame when i had the chance. As a result i got a bad endgame, lost it and thinking im bad at endgames. I would recommend you to learn who is the better side in typical endgames ( for example rock+pawn vs rock is usualy a draw, bishop+rookpawn is draw if the queening square is opposite colour of the bishop) so you dont have to calculate the whole endgame before your transition from mid game. You dont really have to know exactly how to win them. You got plenty of time for that when you have reached your favorable endgame.
Hi, I'm a B player and like open attacking games, but I often mess up the endgame, like when each side has king + one or two pieces + several pawns. I know all the basic endgame theory at the level of Averbakh's Essential Chess Endings, but in practical games I need help. What things can I do to improve my endgame? Thank you.