Skip to content

Commit

Permalink
Implemented major restructuring to allow for specifying:
Browse files Browse the repository at this point in the history 
- Node (R11, R12, or R20)
- Base region (by default NAM for each)
- Base year
- SSP scenario version (updated/review or previous)
- SSP scenario
- Projection method (learning, GDP adjusted, or convergence with splines)
- Convergence year
  • Loading branch information
@measrainsey
measrainsey committed Aug 7, 2023
1 parent 76deb4a commit faf6306
Show file tree
Hide file tree
Showing 5 changed files with 573 additions and 590 deletions.
106 changes: 77 additions & 29 deletions message_ix_models/tools/costs/gdp.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -103,9 +103,7 @@ def get_gdp_data() -> pd.DataFrame:


# Function to read in (under-review) SSP data
def process_raw_ssp_data(
sel_node: str = "r12", reference_region: str = "R12_NAM"
) -> pd.DataFrame:
def process_raw_ssp_data(input_node: str, input_ref_region: str) -> pd.DataFrame:
"""Read in raw SSP data and process it
This function takes in the raw SSP data (in IAMC format), aggregates \
Expand All @@ -132,12 +130,23 @@ def process_raw_ssp_data(
- gdp_ppp_per_capita: GDP per capita (in units of billion US$2005/yr / million)
"""
# Change node selection to upper case
node_up = sel_node.upper()
node_up = input_node.upper()

# Check if node selection is valid
if node_up not in ["R11", "R12", "R20"]:
print("Please select a valid region: R11, R12, or R20")

# Set default reference region
if input_ref_region is None:
if input_node.upper() == "R11":
input_ref_region = "R11_NAM"
if input_node.upper() == "R12":
input_ref_region = "R12_NAM"
if input_node.upper() == "R20":
input_ref_region = "R20_NAM"
else:
input_ref_region = input_ref_region

# Set data path for node file
node_file = package_data_path("node", node_up + ".yaml")

Expand Down Expand Up @@ -236,15 +245,16 @@ def process_raw_ssp_data(
)

# If reference region is not in the list of regions, print error message
if reference_region.upper() not in df.region.unique():
reference_region = input_ref_region.upper()
if reference_region not in df.region.unique():
print("Please select a valid reference region: " + str(df.region.unique()))
# If reference region is in the list of regions, calculate GDP ratios
else:
df = (
df.pipe(
lambda df_: pd.merge(
df_,
df_.loc[df_.region == reference_region.upper()][
df_.loc[df_.region == reference_region][
["scenario_version", "scenario", "year", "gdp_ppp_per_capita"]
]
.rename(columns={"gdp_ppp_per_capita": "gdp_per_capita_reference"})
Expand Down Expand Up @@ -273,16 +283,18 @@ def process_raw_ssp_data(


def linearly_regress_tech_cost_vs_gdp_ratios(
gdp_ratios_df: pd.DataFrame, tech_cost_ratios_df: pd.DataFrame
gdp_df: pd.DataFrame,
cost_ratios_df: pd.DataFrame,
input_base_year: int,
) -> pd.DataFrame:
"""Compute linear regressions of technology cost ratios to GDP ratios
Parameters
----------
gdp_ratios_df : pandas.DataFrame
Dataframe output from :func:`.get_gdp_data`
tech_cost_ratios_df : str -> tuple of (str, str)
Dataframe output from :func:`.calculate_region_cost_ratios`
Dataframe output from :func:`.process_raw_ssp_data`
region_diff_df : str -> tuple of (str, str)
Dataframe output from :func:`.get_weo_region_differentiated_costs`
Returns
-------
Expand All @@ -298,19 +310,19 @@ def linearly_regress_tech_cost_vs_gdp_ratios(
- stderr: standard error of the linear regression
"""

gdp_2020 = gdp_ratios_df.query("year == 2020").reindex(
gdp_base_year = gdp_df.query("year == @input_base_year").reindex(
["scenario_version", "scenario", "region", "gdp_ratio_reg_to_reference"], axis=1
)
cost_capital_2021 = tech_cost_ratios_df.reindex(
["weo_technology", "region", "cost_type", "cost_ratio"], axis=1
inv_cost_base_year = cost_ratios_df.reindex(
["message_technology", "region", "reg_cost_ratio"], axis=1
)

df_gdp_cost = (
pd.merge(gdp_2020, cost_capital_2021, on=["region"])
.groupby(["cost_type", "scenario_version", "scenario", "weo_technology"])
pd.merge(gdp_base_year, inv_cost_base_year, on=["region"])
.groupby(["scenario_version", "scenario", "message_technology"])
.apply(
lambda x: pd.Series(
linregress(x["gdp_ratio_reg_to_reference"], x["cost_ratio"])
linregress(x["gdp_ratio_reg_to_reference"], x["reg_cost_ratio"])
)
)
.rename(
Expand All @@ -331,11 +343,9 @@ def linearly_regress_tech_cost_vs_gdp_ratios(

# Function to calculate adjusted region-differentiated cost ratios
# using the results from the GDP linear regressions
def calculate_adjusted_region_cost_ratios(
gdp_df,
linear_regression_df,
reference_region: str = "R12_NAM",
):
def calculate_gdp_adjusted_region_cost_ratios(
region_diff_df, input_node, input_ref_region, input_base_year
) -> pd.DataFrame:
"""Calculate adjusted region-differentiated cost ratios
This function calculates the adjusted region-differentiated cost ratios \
Expand All @@ -361,13 +371,46 @@ def calculate_adjusted_region_cost_ratios(
- cost_ratio_adj: the adjusted region-differentiated cost ratio
"""

if reference_region.upper() not in gdp_df.region.unique():
print("Please select a valid reference region: " + str(gdp_df.region.unique()))
df_gdp = process_raw_ssp_data(
input_node=input_node, input_ref_region=input_ref_region
).query("year >= 2020")
df_cost_ratios = region_diff_df.copy()

# If base year does not exist in GDP data, then use earliest year in GDP data
# and give warning
base_year = int(input_base_year)
if int(base_year) not in df_gdp.year.unique():
base_year = int(min(df_gdp.year.unique()))
print(
f"Base year {input_base_year} not found in GDP data. \
Using {base_year} for GDP data instead."
)

# Set default values for input arguments
# If specified node is R11, then use R11_NAM as the reference region
# If specified node is R12, then use R12_NAM as the reference region
# If specified node is R20, then use R20_NAM as the reference region
# However, if a reference region is specified, then use that instead
if input_ref_region is None:
if input_node.upper() == "R11":
reference_region = "R11_NAM"
if input_node.upper() == "R12":
reference_region = "R12_NAM"
if input_node.upper() == "R20":
reference_region = "R20_NAM"
else:
reference_region = input_ref_region

# Linearly regress technology cost ratios to GDP ratios
df_linear_reg = linearly_regress_tech_cost_vs_gdp_ratios(
df_gdp, df_cost_ratios, input_base_year=base_year
)

if reference_region.upper() not in df_gdp.region.unique():
print("Please select a valid reference region: " + str(df_gdp.region.unique()))
else:
df = (
linear_regression_df.loc[linear_regression_df.cost_type == "inv_cost"]
.drop(columns=["cost_type"])
.merge(gdp_df, on=["scenario_version", "scenario"])
df_linear_reg.merge(df_gdp, on=["scenario_version", "scenario"])
.drop(
columns=[
"gdp_ppp_per_capita",
Expand All @@ -377,21 +420,26 @@ def calculate_adjusted_region_cost_ratios(
]
)
.assign(
cost_ratio_adj=lambda x: np.where(
reg_cost_ratio_adj=lambda x: np.where(
x.region == reference_region,
1,
x.slope * x.gdp_ratio_reg_to_reference + x.intercept,
),
year=lambda x: x.year.astype(int),
scenario_version=lambda x: np.where(
x.scenario_version.str.contains("2013"),
"Previous (2013)",
"Review (2023)",
),
)
.reindex(
[
"scenario_version",
"scenario",
"weo_technology",
"message_technology",
"region",
"year",
"cost_ratio_adj",
"reg_cost_ratio_adj",
],
axis=1,
)
Expand Down
28 changes: 15 additions & 13 deletions message_ix_models/tools/costs/learning.py
View file
Original file line number Diff line number Diff line change
@@ -1,14 +1,14 @@
import numpy as np
import pandas as pd

from message_ix_models.tools.costs.config import (
BASE_YEAR,
FIRST_MODEL_YEAR,
LAST_MODEL_YEAR,
PRE_LAST_YEAR_RATE,
)
from message_ix_models.util import package_data_path

# Global variables of model years
BASE_YEAR = 2021
FIRST_MODEL_YEAR = 2020
LAST_MODEL_YEAR = 2100
PRE_LAST_YEAR_RATE = 0.01


# Function to get GEA based cost reduction data
def get_cost_reduction_data() -> pd.DataFrame:
Expand Down Expand Up @@ -139,13 +139,13 @@ def project_ref_region_inv_costs_using_learning_rates(
# Set default reference region
if input_ref_region is None:
if input_node.upper() == "R11":
input_ref_region = "R11_NAM"
reference_region = "R11_NAM"
if input_node.upper() == "R12":
input_ref_region = "R12_NAM"
reference_region = "R12_NAM"
if input_node.upper() == "R20":
input_ref_region = "R20_NAM"
reference_region = "R20_NAM"
else:
input_ref_region = input_ref_region
reference_region = input_ref_region

# Get cost reduction data
df_cost_reduction = get_cost_reduction_data()
Expand All @@ -161,18 +161,19 @@ def project_ref_region_inv_costs_using_learning_rates(
# Filter for reference region, then merge with learning scenarios and discount rates
# Calculate cost in reference region in 2100
df_ref = (
regional_diff_df.query("region == @input_ref_region")
regional_diff_df.query("region == @reference_region")
.merge(df_learning_reduction, on="message_technology")
.assign(
cost_region_2100=lambda x: x.reg_cost_base_year
- (x.reg_cost_base_year * x.cost_reduction),
b=lambda x: (1 - PRE_LAST_YEAR_RATE) * x.cost_region_2100,
r=lambda x: (1 / (LAST_MODEL_YEAR - input_base_year))
* np.log((x.cost_region_2100 - x.b) / (x.reg_cost_base_year - x.b)),
reference_region=reference_region,
)
)

seq_years = list(range(FIRST_MODEL_YEAR, LAST_MODEL_YEAR + 10, 10))
seq_years = list(range(FIRST_MODEL_YEAR, LAST_MODEL_YEAR + 5, 5))

for y in seq_years:
df_ref = df_ref.assign(
Expand All @@ -194,7 +195,6 @@ def project_ref_region_inv_costs_using_learning_rates(
"reg_cost_ratio",
"reg_cost_base_year",
"fix_to_inv_cost_ratio",
"first_technology_year",
"learning_rate",
"technology_type",
"cost_reduction",
Expand All @@ -205,6 +205,8 @@ def project_ref_region_inv_costs_using_learning_rates(
id_vars=[
"message_technology",
"scenario",
"reference_region",
"first_technology_year",
],
var_name="year",
value_name="inv_cost_ref_region_learning",
Expand Down
Loading

0 comments on commit faf6306

Please sign in to comment.