NPV improvements results

[Aurashk]

I've implemented the suggestions for improving the npv calculation and for the 'simple' example it seems to be working more like you'd expect. copying the results here for a sanity check

• input folder (enable broken results and make A0_RES use npv objective)
  simple_test.zip

• current output
  simple_test_unchanged.zip

• output when we add 1e-14 to each activity coefficient in npv objective (PR add epsilon term to npv activity coefficients #994)
  simple_test_with_eps.zip

• output when we remove AFC term from the optimisation (PR refactor and remove capacity term from npv objective #997)
  simple_test_without_afc_term.zip

• output when have both of the above changes (completes without errors!)
  simple_test_both_changes.zip

[ahawkes]

Thanks sounds like this is working as expected. I am away at the moment but perhaps @tsmbland can take a look in the meantime - specifically are the values being produced for the metric as expected from manual calculation?

[tsmbland]

Had a go calculating the profitability index by hand for one example and it seems correct. This is for the simple model, setting all agent objectives to NPV, then looking at the WNDFRM candidate in 2030:

Capacity: 7.5828584109727935
Capital cost: 1000
Fixed operating cost: 30
Lifetime: 25
Discount rate: 0.1
-> capital recovery factor = 0.11016807219
-> annual capital costs per capacity unit = 110.16807219
-> annual fixed cost per capacity unit = 140.16807219
-> annual fixed cost = 1062.87464516

Variable operating cost per activity: 0.4
Total variable cost (sum(activity) * var cost): 34.49545599
Sum of timeslice revenue (sum(elc price * activity * output coeff)): 1224.935797
-> Total annual surplus: 1190.440341

=> Profitability index (annual surplus/ annual fixed costs) = 1.12001951163

Reported value: -1.1200195115677323 (noting that MUSE converts the number to negative internally so it can find the best by taking a minimum)