Multi-Year Investment Pathways

Let's explore the multi-year investments in Tulipa. We will talk about discount approaches and different investment methods. The latter is an example for different levels of detail in Tulipa.

1. Set up

1. Paste the code below that add the packages and instantiate your enviroment (if you don't have it already)

using Pkg: Pkg       # Julia package manager (like pip for Python)
Pkg.activate(".")    # Creates and activates the project in the new folder - notice it creates Project.toml and
# Or enter package mode (type ']') and run 'pkg> activate .'
# Manifest.toml in your folder for reproducibility
Pkg.add("TulipaEnergyModel")
Pkg.add("TulipaIO")
Pkg.add("DuckDB")
Pkg.add("DataFrames")
Pkg.add("Plots")
Pkg.instantiate()

1. Paste the code below that loads the packages in the file

import TulipaIO as TIO
import TulipaEnergyModel as TEM
using DuckDB
using DataFrames
using Plots

2. The problem & explore the files

We are modeling two milestone years 2030 and 2050. The system has some initial wind capacity built in 2020, the model can choose to invest in wind in both milestone years.

There are two pairs of input-output files, we start with the simple one.

3. Discount parameters

Run TEM

connection = DBInterface.connect(DuckDB.DB)
input_dir = joinpath(@__DIR__, "my-awesome-energy-system/tutorial-6-simple-method")
TIO.read_csv_folder(connection, input_dir)
TEM.populate_with_defaults!(connection)
energy_problem = TEM.run_scenario(connection)

EnergyProblem:
  - Model created!
    - Number of variables: 290
    - Number of constraints for variable bounds: 290
    - Number of structural constraints: 432
  - Model solved!
    - Termination status: OPTIMAL
    - Objective value: 8.502460892530702e6
    - Objective breakdown:
      - assets_fixed_cost_aggregated_vintage_method: 3.8553e6
      - assets_fixed_cost_compact_vintage_method: 0.0
      - assets_investment_cost: 184970.34199707804
      - flows_fixed_cost: 0.0
      - flows_investment_cost: 0.0
      - flows_operational_cost: 4.462190550533624e6
      - storage_assets_energy_fixed_cost: 0.0
      - storage_assets_energy_investment_cost: 0.0
      - units_on_operational_cost: 0.0
      - vintage_flows_operational_cost: 0.0

There is a new file model-parameters.csv. It contains model-wide parameters, in this case:

connection = energy_problem.db_connection
row = only(collect(DuckDB.query(connection, "SELECT discount_rate, discount_year FROM model_parameters")))
discount_rate = row.discount_rate
discount_year = row.discount_year

2030

Check discounting parameters calculated internally by TEM.

df_objective = filter(:asset => ==("wind"), TIO.get_table(connection, "t_objective_assets"))[:,
    [:asset, :milestone_year, :investment_cost, :annualized_cost, :salvage_value,
     :investment_year_discount, :weight_for_asset_investment_discount, :weight_for_operation_discounts]]

df_asset = filter(:asset => ==("wind"), TIO.get_table(connection, "asset"))[:,
    [:asset, :technical_lifetime, :economic_lifetime]]

df = leftjoin(df_objective, df_asset, on = :asset)

4. Different levels of details: aggregated method vs compact method

Remember that we have wind built in 2020 - does it have the same profiles as 2030? Let's check it out.

plot()
wind_profiles = filter(row -> occursin("wind", row.profile_name) && row.milestone_year == 2030,
    TIO.get_table(connection, "profiles_rep_periods"))

for pname in unique(wind_profiles.profile_name)
    subdf = wind_profiles[wind_profiles.profile_name .== pname, :]
    plot!(subdf.value, label="$(pname), year 2030")
end
xlabel!("Time")
ylabel!("Capacity factor")

So...the wind built in 2020 has a worse profile. How does it play a role in the investment methods?

Aggregated method

Let's try the aggregated method first (vintage_method = "aggregated").

connection = DBInterface.connect(DuckDB.DB)
input_dir = joinpath(@__DIR__, "my-awesome-energy-system/tutorial-6-simple-method")
TIO.read_csv_folder(connection, input_dir)
TEM.populate_with_defaults!(connection)
energy_problem = TEM.run_scenario(connection)

EnergyProblem:
  - Model created!
    - Number of variables: 290
    - Number of constraints for variable bounds: 290
    - Number of structural constraints: 432
  - Model solved!
    - Termination status: OPTIMAL
    - Objective value: 8.502460892530702e6
    - Objective breakdown:
      - assets_fixed_cost_aggregated_vintage_method: 3.8553e6
      - assets_fixed_cost_compact_vintage_method: 0.0
      - assets_investment_cost: 184970.34199707804
      - flows_fixed_cost: 0.0
      - flows_investment_cost: 0.0
      - flows_operational_cost: 4.462190550533624e6
      - storage_assets_energy_fixed_cost: 0.0
      - storage_assets_energy_investment_cost: 0.0
      - units_on_operational_cost: 0.0
      - vintage_flows_operational_cost: 0.0

Check initial capacity - under the aggregated method, we will not be able to differentiate units built in other years (than milestone years), they will be considered the same as the units built in the milestone year, which means that we will not use the 2020 profile.

filter(row -> row.asset=="wind" && row.milestone_year == 2030, TIO.get_table(connection, "asset_both"))

Check the objective value and investments.

energy_problem.objective_value
filter(row -> row.asset=="wind", TIO.get_table(connection, "var_assets_investment"))

Compact method

Now try the compact method (vintage_method = "compact_profiles").

connection = DBInterface.connect(DuckDB.DB)
input_dir = joinpath(@__DIR__, "my-awesome-energy-system/tutorial-6-compact-method")
TIO.read_csv_folder(connection, input_dir)
TEM.populate_with_defaults!(connection)
energy_problem = TEM.run_scenario(connection)

EnergyProblem:
  - Model created!
    - Number of variables: 290
    - Number of constraints for variable bounds: 290
    - Number of structural constraints: 432
  - Model solved!
    - Termination status: OPTIMAL
    - Objective value: 8.619710327700023e6
    - Objective breakdown:
      - assets_fixed_cost_aggregated_vintage_method: 0.0
      - assets_fixed_cost_compact_vintage_method: 3.95735e6
      - assets_investment_cost: 192146.6391663987
      - flows_fixed_cost: 0.0
      - flows_investment_cost: 0.0
      - flows_operational_cost: 4.470213688533625e6
      - storage_assets_energy_fixed_cost: 0.0
      - storage_assets_energy_investment_cost: 0.0
      - units_on_operational_cost: 0.0
      - vintage_flows_operational_cost: 0.0

Check initial capacity - units built in different years are explicitly listed, meaning that their corresponding profiles are also considered.

filter(row -> row.asset=="wind" && row.milestone_year == 2030, TIO.get_table(connection, "asset_both"))

Check the objective value:

energy_problem.objective_value

8.619710327700023e6

And, check the investments:

filter(row -> row.asset=="wind", TIO.get_table(connection, "var_assets_investment"))

We use the worse but correct profile for wind built in 2020, leading to more required investments and hence higher costs.