Benchmarking energy efficiency of water treatment plants: Effects of data variability