Exercise 2

In this exercise, you will try to use FMT to answer different questions about a strategic model and a tactical model through a re-planning process.

Thus, this exercise will teach you to:

• Create a re-planning script
• Re-plan with multiple models, and varying special FMT keywords
• Analyze re-planning results to answer different questions

To do this exercise, you must have FMT installed on your computer, along with the example files from the FMT GitHub repository.

• 💾 Download and install FMT

• 💾 Download the FMT example files

You must also have Python installed on your computer. If this is not done, refer to this page.

Part 1: Creating a re-planning script

The re-planning process often involves running many models, many times (to make iterations).

It is therefore very useful to create a Python script that can be called through a console, a terminal, or a command prompt.

Even better : it is possible to create a Python script that will read arguments given to it when it is called with the command prompt. This will allow to re-use this script for different re-planning scenarios or purposes.

Try to create the following script in Python language:

• Importing Models, Parser, Core, Version and Parallel from FMT.
• Read the following arguments (it is recommended to use the argparse module)
  • -pri which specifies the location of the primary file of the models that will be used
  • -str which indicates the name of the strategic scenario
  • -tac which indicates the name of the tactical scenario
  • -sto which indicates the name of the stochastic scenario
  • -out indicates a comma-separated list of output values of interest to be extracted from the model results
  • -fol which indicates the folder where the output files of the script will be put
  • -rep which indicates the number of iterations/replicas to do
  • -len which indicates the number of periods of the strategic model
  • -rel which indicates the number of periods over which the re-planning will be done
  • -thr which indicates the number of CPU cores to use for parallelization

Next, the script will need to:

• Read the model via the primary file
• Read the different scenarios
• Create the strategic/global model using the scenario name provided by -str.
  • Make the global model run for the duration provided by -len.
  • Make the global model use the number of threads given by -thr for parallel resolution
• Create the stochastic model using the scenario name provided by -sto.
  • Make the stochastic model run for one period
• Create the tactical model using the scenario name provided by -tac.
  • Make the tactical model run for one period
  • Make the global model use the number of threads specified by -thr for parallel resolution
• Add the outputs selected by the -out argument in a list from the model outputs
• Prepare the re-planning task through the Parallel.FMTreplanningtask() function by providing to the function :
  • The strategic model
  • The stochastic model
  • The tactical model
  • The list of desired outputs
  • The folder where the outputs will be put
  • The format of the outputs (here, .csv)
  • All other arguments needed for the function
• Run the rescheduling task through the Parallel.FMTtaskhandler() function

When you are done, or if you are stuck, you can view the correction of this exercise for Python code by clicking on the “Reveal correction” button below.

from FMT import Models
from FMT import Parser
from FMT import Core
from FMT import Version
from FMT import Parallel
import os,argparse


if __name__ == "__main__":
    ap = argparse.ArgumentParser()
    ap.add_argument("-pri", "--primary", required=True,help="Fichier primaire")
    ap.add_argument("-str", "--strategic", required=True,help="Nom du scenario strategique")
    ap.add_argument("-tac", "--tactic", required=True,help="Nom du scenario tactique")
    ap.add_argument("-sto", "--stochastic", required=True,help="Nom du scenario stochastique")
    ap.add_argument("-out", "--outputs", required=True,help="Liste d'outputs separe d'une virgule")
    ap.add_argument("-fol", "--folder", required=True,help="Dossier de sortie")
    ap.add_argument("-rep", "--replicates", required=True,help="Nombre de replicas")
    ap.add_argument("-len", "--strategiclength", required=True,help="Nombre de periode du strategique")
    ap.add_argument("-rel", "--replanninglength", required=True,help="Nombre de periode de re-plannification")
    ap.add_argument("-thr", "--threads", required=True,help="Nombre de threads")
    ap.add_argument("-com","--compresstime", nargs="+",help="Compress time sous la forme 20,5 40,8")
    args = vars(ap.parse_args())
    if Version.FMTversion().hasfeature("OSI"):
        modelparser = Parser.FMTmodelparser()
        path = args["primary"]
        scenarios = [args["strategic"],args["stochastic"],args["tactic"]]
        models = modelparser.readproject(path, scenarios)
        globalmodel = Models.FMTlpmodel(models[0], Models.FMTsolverinterface.CLP)
        globalmodel.setparameter(Models.FMTintmodelparameters.LENGTH,int(args["strategiclength"]))
        globalmodel.setparameter(Models.FMTintmodelparameters.NUMBER_OF_THREADS,1)
        if "com" in args:
            for firstperiod,value in [tuple(int(cmpt.split(",")[0]),float(cmpt.split(",")[1])) for cmpt in args["com"]]:
                globalmodel.setcompresstime(firstperiod,int(args["strategiclength"]),value)
        stochastic = Models.FMTnssmodel(models[1],0)
        stochastic.setparameter(Models.FMTintmodelparameters.LENGTH, 1)
        localmodel = Models.FMTlpmodel(models[2], Models.FMTsolverinterface.CLP)
        localmodel.setparameter(Models.FMTintmodelparameters.LENGTH, 1)
        localmodel.setparameter(Models.FMTintmodelparameters.NUMBER_OF_THREADS,1)
        selectedoutputs=[]
        for output in globalmodel.getoutputs():
            if output.getname() in args["outputs"].split(","):
                selectedoutputs.append(output)
        outputlocation = args["folder"]
        layersoptions=["SEPARATOR=SEMICOLON"]
        replanningtask=Parallel.FMTreplanningtask(globalmodel,stochastic, localmodel, selectedoutputs, outputlocation, "CSV", layersoptions,int(args["replicates"]),int(args["replanninglength"]),0.5, Core.FMToutputlevel.totalonly)
        handler = Parallel.FMTtaskhandler(replanningtask,int(args["threads"]))
        handler.setquietlogger()
        handler.conccurentrun()
    else:
        print("FMT needs to be compiled with OSI")

Part 2: Creation of scenarios for re-planning

Preparation

After downloading the FMT example files (see above), open the folder Models/TWD_land/Scenarios/Globalex1.

This Globalex1 scenario is a scenario that we will use as strategic model or global model.

Take time to look at the contents of its files; in particular, look at its actions, objectives and constraints.

Do the same with the Localex1 model, which we will use as the tactical model or local model

Question 1

Open a windows command prompt, or a powershell. Use the command cd to go to the Examples/Python/replanning folder. Then use the following command to start re-planning:

python Replanner.py -pri ../../Models/TWD_land/TWD_land.pri -str Globalex1 -tac Localex1 -sto Globalnofire -out OVOLREC,OSUPREC,OSUPPL,VOLINVENT -fol replanning_EX1 -rep 10 -len 20 -rel 20 -thr 2

This command tells the script created in the first part that we want to :

• Use the model found at ../../Models/TWD_land/TWD_land.pri
• Use the scenario Globalx1 as a strategic model
• Use the Localex1 scenario as a tactical model
• Use the Globalnofire scenario as a stocastic model
• Create output tables for the OVOLREC, OSUPREC, OSUPPL, and VOLINVENT outputs
• Put these tables in the replanning_EX1 folder
• Make 10 iterations
• Solve the strategic model for 20 periods
• Do the re-plannification on 20 periods
• Use 2 CPU cores to make iterations in parallel

Once the command has run, look at the output in the Examples/Python/replanning/replanning_EX1 folder.

With a stock-out tolerance of 5%, what is the probability of holding OVOLREC until period 10 in the tactical model?

Reveal correction

To look at the probability of keeping OVOLREC above a 5% value drop, we just have to look at the outputsdrift.csv file. This file shows us the probability that the different outputs selected in the tactical model remain above or below a certain distance (drift) from the strategic model.

We can open the file, and go to the lines that concern OVOLREC and a drift of 0.05 (5%), namely lines 462-481.

Looking at these rows, we can see that from period 7 onwards, OVOLREC has a probability of 0 of not falling by more than 5% compared to the strategic model (third column LowerProbability); this probability remains at 0 for all subsequent periods. Thus, it appears that the probability of keeping OVOLREC above a 5% decline until period 10 is 0.

These values can be represented with a graph via the data in outputsdrift.csv :

Question 2

Now duplicate the Globalex1 folder and rename it Globalex2 (delete the existing folder if it already exists).

Then modify the TWD_land.opt file of Globalex2 to force the Globalex2 scenario not to exceed 20,000 m³ for the OVOLREC output.

*OBJECTIVE
_MAXMIN oVolRec 1.._LENGTH

*EXCLUDE
aCaribou 1.._LENGTH
afire    1.._LENGTH
arecup    1.._LENGTH

*CONSTRAINTS

oVolRec<= 20000 1.._LENGTH

_NDY(oVolRec) 1.._LENGTH


*FORMAT MOSEK

Then run a re-planning with the strategic model Globalex2 and the tactical model Localex1, with the following command:

python Replanner.py -pri ../../Models/TWD_land/TWD_land.pri -str Globalex2 -tac Localex1 -sto Globalnofire -out OVOLREC,OSUPREC,OSUPPL,VOLINVENT -fol replanning_EX2 -rep 10 -len 20 -rel 20 -thr 2

Once the command has run, look at the output in the Examples/Python/replanning/replanning_EX2 folder.

With a stock-out tolerance of 5%, what is the probability of holding OVOLREC until period 10 in the tactical model?

Is there a gain from question 1?

Question 3

Now duplicate the Localex1 folder and rename it Localex3 (delete the existing folder if it already exists).

Then modify the TWD_land.opt file of Localex3 to force this tactical model to follow the solution of the strategic model with a weight of 1000, using the keyword _SETGLOBALSCHEDULE(weight).

*OBJECTIVE
_MAX ovolrec 1.._LENGTH _SETGLOBALSCHEDULE(100)

*EXCLUDE
aCaribou 1.._LENGTH
afire    1.._LENGTH
arecup    1.._LENGTH

*CONSTRAINTS

oVolRec<= 40272 1.._LENGTH _SETFROMGLOBAL(1)


*FORMAT MOSEK

Then run a re-planning with the strategic model Globalex1 and the tactical model Localex3, with the following command:

python Replanner.py -pri ../../Models/TWD_land/TWD_land.pri -str Globalex1 -tac Localex3 -sto Globalnofire -out OVOLREC,OSUPREC,OSUPPL,VOLINVENT -fol replanning_EX3 -rep 10 -len 20 -rel 20 -thr 2

Once the command has run, look at the output in the Examples/Python/replanning/replanning_EX3 folder.

With a stock-out tolerance of 5%, what is the probability of holding OVOLREC until period 10 in the tactical model?

Is there a gain from question 1?

Question 4

Now duplicate the TWD_land._opt file from Localex1 to the Localex4 folder (delete the existing file if it already exists).

Then modify the TWD_land.opt file in Localex4 to force the tactical model to follow the replica tables (PEUPLEMENT1.csv and PEUPLEMENT2.csv) for harvesting in the stands peuplement1 and peuplement2 (use _REPLICATE(). In doing so, you must force the tactical model to do a certain level of harvesting in particular spatial features of the model (peuplement1 and peuplement2) based on files you provide to the model (PEUPLEMENT1.csv and PEUPLEMENT2.csv).

*OBJECTIVE
_MAX ovolrec - _Penalty(_All) 1.._LENGTH

*EXCLUDE
aCaribou 1.._LENGTH
afire    1.._LENGTH
arecup    1.._LENGTH

*CONSTRAINTS

oVolRec<= 40272 1.._LENGTH _SETFROMGLOBAL(1)

oVolRecPEUPLEMENT1 <= 0 1.._LENGTH _Goal(G1,9999) _REPLICATE(PEUPLEMENT1.csv)
oVolRecPEUPLEMENT1 >= 0 1.._LENGTH _Goal(G2,9999) _REPLICATE(PEUPLEMENT1.csv)
oVolRecPEUPLEMENT2 <= 0 1.._LENGTH _Goal(G3,9999) _REPLICATE(PEUPLEMENT2.csv)
oVolRecPEUPLEMENT2 >= 0 1.._LENGTH _Goal(G4,9999) _REPLICATE(PEUPLEMENT2.csv)

*FORMAT MOSEK

Then run a re-planning with the strategic model Globalex1 and the tactical model Localex4, with the following command:

python Replanner.py -pri ../../Models/TWD_land/TWD_land.pri -str Globalex1 -tac Localex4 -sto Globalnofire -out OVOLREC,OSUPREC,OSUPPL,VOLINVENT,OVOLRECPEUPLEMENT1,OVOLRECPEUPLEMENT2 -fol replanning_EX4 -rep 10 -len 20 -rel 20 -thr 2

Once the command has run, go to the Localex4.csv file in the EX4_Scheduling folder, and look at the results of the OVOLRECPEUPLEMENT1 and OVOLRECPEUPLEMENT2 outputs. In particular, you can create a pivot table in a program such as Microsoft Excel to obtain a graph showing the evolution of these measures over time, with a curve for each iteration.

At the same time, open the files PEUPLEMENT1.csv and PEUPLEMENT2.csv, and observe their contents. Each column corresponds to values that the tactical model was supposed to use for a iteration, with each row containing a value for a given period.

When comparing the contents of Localex4.csv and PEUPLEMENT1.csv and PEUPLEMENT2.csv, what do you observe?

Reveal correction

We can see that the tactical model has indeed followed the harvest levels defined in the PEUPLEMENT1.csv and PEUPLEMENT2.csv files for the output OVOLREC in stands 1 and 2.

In particular, we can see that OVOLRECPEUPLEMENT1 drops to 0 for period 4 in all iterations, as expected in PEUPLEMENT1.csv. We can also see that iteration 4 keeps a value of OVOLRECPEUPLEMENT1 at 10 most of the time, while the other iterations keep a value of 20.

We can observe a drop to 0 for OVOLRECPEUPLEMENT2 at period 9 in all iterations, as expected in PEUPLEMENT2.csv.

Question 5

Now duplicate the Localex1 folder and rename it Localex5 (delete the existing folder if it already exists).

Then modify the TWD_land.opt file of Localex5 to force this tactical model to ignore the 90-100ha planting constraint from period (or to follow it only in period 1) using the keyword _REIGNORE(period).

*OBJECTIVE
_MAX ovolrec - _Penalty(_All) 1.._LENGTH

*EXCLUDE
aCaribou 1.._LENGTH
afire    1.._LENGTH
arecup    1.._LENGTH

*CONSTRAINTS

oVolRec<= 40272 1.._LENGTH _SETFROMGLOBAL(1)

osuppl >= 90 1 _Goal(G1,9999) _REIGNORE(2)
osuppl <= 100 1 _Goal(G2,9999) _REIGNORE(2)

*FORMAT MOSEK

Then run a re-planning with the strategic model Globalex1 and the tactical model Localex5, with the following command:

python Replanner.py -pri ../../Models/TWD_land/TWD_land.pri -str Globalex1 -tac Localex5 -sto Globalnofire -out OVOLREC,OSUPREC,OSUPPL,VOLINVENT -fol replanning_EX5 -rep 10 -len 20 -rel 20 -thr 2

Once the command has run, compare the evolution of the planting level output OSUPPL in the tactical model Localex1 (in the planning_EX1 folder created in question 1) with the evolution of OSUPPL in the tactical model Localex5 (in the planning_EX5 folder). As before, you can use a pivot table in a program like Microsoft Excel to obtain a graph showing the evolution of these measures over time, with a curve for each of the iterations.

What differences do you observe between Localex1 and Localex5?

Reveal correction

We can see that the tactical model Localex1 manages to sustain the level of OSUPPL for a long time in a stable way, as shown on this graph:

In comparison, in Localex5, we can see that the level of OSUPPL is much more variable and unstable. It quickly reaches values of 0, due to the fact that the model constraints on it are ignored after period 2, as shown on this graph:

Hence, it seems that the local model did ignore the constraints OSUPPL.

Question 6

Now duplicate the Localex1 folder and rename it Localex6 (delete the existing folder if it already exists).

Then modify the TWD_land.opt file of Localex6 to force this tactical model to achieve 100% of the harvest level (OVOLREC) of the strategic model, with a 1% variability allowed for exceeding the value (so at least 100%, and at most 101%) with the keyword _SETFROMGLOBAL(weight).

*OBJECTIVE
_MAX ovolrec - _Penalty(_All) 1.._LENGTH

*EXCLUDE
aCaribou 1.._LENGTH
afire    1.._LENGTH
arecup    1.._LENGTH

*CONSTRAINTS

oVolRec<= 40272 1.._LENGTH _SETFROMGLOBAL(1)

osuprec <= 40272 1.._LENGTH _Goal(G1,9999) _SETFROMGLOBAL(1.0)
osuprec >= 40272 1.._LENGTH _Goal(G2,9999) _SETFROMGLOBAL(0.99)

*FORMAT MOSEK

Then run a re-plannibg with the strategic model Globalex1 and the tactical model Localex6, with the following command:

python Replanner.py -pri ../../Models/TWD_land/TWD_land.pri -str Globalex1 -tac Localex6 -sto Globalnofire -out OVOLREC,OSUPREC,OSUPPL,VOLINVENT -fol replanning_EX6 -rep 10 -len 20 -rel 20 -thr 2

Once the command has run, look at the output in the Examples/Python/replanning/replanning_EX6 folder.

Now compare the probability of maintaining a similar harvest as in Exercise 1 (with Localex1) by looking at the contents of outputsdrift.csv in replanning_EX1 and replanning_EX6.

What differences do you observe?

Question 7

Now duplicate the Globalex1 folder and rename it Globalex7 (delete the existing folder if it already exists).

Then modify the TWD_land.opt file of Globalex7 to force the Globalex7 scenario to achieve at least 99% of the OVOLREC harvest level of the tactical model.

*OBJECTIVE
_MAXMIN oVolRec - _Penalty(_All) 1.._LENGTH

*EXCLUDE
aCaribou 1.._LENGTH
afire    1.._LENGTH
arecup    1.._LENGTH

*CONSTRAINTS

osuprec <= 40272 1.._LENGTH _Goal(G1,9999) _SETFROMLOCAL(1.0)
osuprec >= 40272 1.._LENGTH _Goal(G2,9999) _SETFROMLOCAL(0.99)

_NDY(oVolRec) 1.._LENGTH


*FORMAT MOSEK

Then run a re-planning with the strategic model Globalex7 and the tactical model Localex1, with the following command:

python Replanner.py -pri ../../Models/TWD_land/TWD_land.pri -str Globalex7 -tac Localex1 -sto Globalnofire -out OVOLREC,OSUPREC,OSUPPL,VOLINVENT -fol replanning_EX7 -rep 10 -len 20 -rel 20 -thr 2

Once the command has run, look at the output in the Examples/Python/replanning/replanning_EX7 folder.

Now observe how the planting levels (output OSSUPL) compare between those of question 1, and those of this re-planning by looking at the outputsdrift.csv files of replanning_EX1 and replanning_EX7.

How does the maintenance of the level of OSSUPL compare to question 1?