Thank you tospo.
The problem is that not each number can be combined with each evaluation name. Therefore I do need a config file. In fact I happened to populate the hash %module reading the csv-file in the following way:

open(my $fh, "<", "parameter.csv") or die "$!";
my $header = <$fh>;
while(<$fh>)    {
    my @line = split(/;/, $_);
    my $t = $line[0];
    ${module}{$t}{name} = $line[1];
    my $n = $line[2];    
    ${module}{$t}{indicator}{$n}{ind} = $line[3];
    ${module}{$t}{indicator}{$n}{yscala} = $line[4];
}
close $fh;
[download]

The file names in the gnuplot command are generated afterwards on the fly whereby only the names are created which are allowed by config file.
After that I used the code from wfsp from Re^2: How to declare variables per loop in the following way:

my @cmds;
for my $number (sort keys %{${module}{$topic}{indicator}}){
  my $cmd = build_command($number, %args);
  push @cmds, $cmd;
}

for (my $i=0; $i<@cmds; $i++) {
        if($i == 0 or $i % 2 == 0)    {
            $cmds[$i] = "$multiplot\n\n$cmds[$i]";
        }
}

for (my $i=0; $i<@cmds; $i++) {
        if($i % 2 == 1 or $i == $#cmds)    {
            $cmds[$i] = "$cmds[$i]\n\n unset multiplot\n\n";
        }
}

my $slurpline;
for (my $i=0; $i<@cmds; $i++) {
        $slurpline.=$cmds[$i];
}
[download]

The gnuplot command itself is very short now:

open my $P, "|-", "gnuplot" or die;

printflush $P qq[
                
        $postscript
        
        $slurpline
        
            
]; 

close $P;
[download]

Thanks again!
VE