Tip 1: profile. There's no real point in optimising code which already contributes negligibly to the runtime. I like the venerable Devel::NYTProf but other profilers are available.

I haven't analysed your code in any detail but this just jumps right out at me:

    for (@lines2) { s/A2022Local/ACTUAL;FY22;Working_Central;Input;Loc
+al_YTD/g }
    for (@lines2) { s/A2022AjConsoLocal/ACTUAL;FY22;Working_Central;Ad
+j_Conso;Local_YTD/g }
    for (@lines2) { s/A2022InEur/ACTUAL;FY22;Working_Central;Input;Eur
+_ACT2022_Rate_YTD/g }
    for (@lines2) { s/A2022AjConso/ACTUAL;FY22;Working_Central;Adj_Con
+so;Eur_ACT2022_Rate_YTD/g }
    for (@lines2) { s/A2022TxB22/ACTUAL;FY22;Working_Central;Input;Eur
+_BUD2022_Rate_YTD/g }
    for (@lines2) { s/A2022TxB22AjConso/ACTUAL;FY22;Working_Central;Ad
+j_Conso;Eur_BUD2022_Rate_YTD/g }

    for (@lines2) { s/A2021Local/ACTUAL;FY21;Working_Central;Input;Loc
+al_YTD/g }
    for (@lines2) { s/A2021AjConsoLocal/ACTUAL;FY21;Working_Central;Ad
+j_Conso;Local_YTD/g }
    for (@lines2) { s/A2021InEur/ACTUAL;FY21;Working_Central;Input;Eur
+_ACT2021_Rate_YTD/g }
    for (@lines2) { s/A2021AjConso/ACTUAL;FY21;Working_Central;Adj_Con
+so;Eur_ACT2021_Rate_YTD/g }
    for (@lines2) { s/A2021TxB22/ACTUAL;FY21;Working_Central;Input;Eur
+_BUD2021_Rate_YTD/g }
    for (@lines2) { s/A2021TxB22AjConso/ACTUAL;FY21;Working_Central;Ad
+j_Conso;Eur_BUD2021_Rate_YTD/g }

    for (@lines2) { s/A2020Local/ACTUAL;FY20;Working_Central;Input;Loc
+al_YTD/g }
    for (@lines2) { s/A2020AjConsoLocal/ACTUAL;FY20;Working_Central;Ad
+j_Conso;Local_YTD/g }
    for (@lines2) { s/A2020InEur/ACTUAL;FY20;Working_Central;Input;Eur
+_ACT2020_Rate_YTD/g }
    for (@lines2) { s/A2020AjConso/ACTUAL;FY20;Working_Central;Adj_Con
+so;Eur_ACT2020_Rate_YTD/g }
    for (@lines2) { s/A2020TxB22/ACTUAL;FY20;Working_Central;Input;Eur
+_BUD2020_Rate_YTD/g }
    for (@lines2) { s/A2020TxB22AjConso/ACTUAL;FY20;Working_Central;Ad
+j_Conso;Eur_BUD2020_Rate_YTD/g }

    for (@lines2) { s/B2022Local/BUDGET;FY22;Working_Central;Input;Loc
+al_YTD/g }
    for (@lines2) { s/B2022AjConsoLoc/BUDGET;FY22;Working_Central;Adj_
+Conso;Local_YTD/g }
    for (@lines2) { s/B2022AjTBLocal/BUDGET;FY22;Working_Central;Adj_C
+onso;Local_YTD/g }
    for (@lines2) { s/B2022InEur/BUDGET;FY22;Working_Central;Input;Eur
+_BUD2022_Rate_YTD/g }
    for (@lines2) { s/B2022AjConso/BUDGET;FY22;Working_Central;Adj_Con
+so;Eur_BUD2022_Rate_YTD/g }
    for (@lines2) { s/B2022AjTB/BUDGET;FY22;Working_Central;Adj_Conso;
+Eur_BUD2022_Rate_YTD/g }
[download]

You are looping over the same array 24 times to run 24 very similar search/replace operations. It should be much more efficient (not to mention maintainable) to combine these into one loop and ideally one search/replace operation. But if your profiling indicates that the slow part is elsewhere, concentrate on that first.