Delta Module optimization#

The Delta Module is designed for sensitivity analysis, allowing users to explore the impact of parameter variations on optimization results. By creating multiple variations (called “deltas”), the module runs a series of optimizations to assess how different parameter values affect the outcome. This is particularly useful in understanding the robustness of a solution and identifying key drivers in the system’s behavior.

Warning

The Delta Module operation should be used with caution due to its potential volumetry. Many optimizations may be triggered, which can require significant time and computational resources. Misuse of this operation (e.g., launching too many optimizations or using excessively large contexts) can lead to server slowdown and severe issues.

Workflow#

Step 1: Import Reference Contexts into ACSG#

Import one or more reference contexts into a specific study in ACSG. These contexts represent the baseline scenarios or test cases for assets that will be modified later.

Step 2: Calibrate Reference Contexts#

Once imported, calibrate the reference contexts in ACSG to ensure they represent the desired model. This calibration involves:

Running the contexts multiple times to ensure accuracy.
Analysing KPIs
Testing different XPRS engine options to identify the fastest one. For more details, see the section Launch context optimization in the Operations documentation.
Estimating the RAM requirements needed to solve the context, allowing maximum parallelization of calculations.

Step 3: Launch Delta Optimization#

Initiate the Delta Optimization Module from ACSG. The system will prompt you to upload a ZIP file containing:

A folder with the Delta Plan (self-contained folder with references to the study contexts).
A folder containing the KPIs to calculate, along with user-defined filters.

Configure the following advanced options:

Memory Allocation Configuration: Set to Custom.
Custom Memory Value: Enter a memory value with a 10% margin based on the calibration results.
XPRS Options: Choose the engine option that was fastest during calibration.

Step 4: Analyze Calculated KPIs#

Once the delta optimization is complete, view the calculated scalar KPIs in ACSG. Delta indexes are added to the results to distinguish between different contexts.

Step 5: Export and Further Analysis#

Results can be exported for further manipulation. You can calculate additional statistical indicators or perform other analyses using ACSG’s tools.

Delta Module Operation#

Expected ZIP File Structure#

The Delta Module expects a ZIP file with the following structure:

The csv_files folder contains configuration files for deltas, paths, timeseries, and curves.
The kpis.csv file specifies the KPIs to calculate for each context-delta combination.
The filters folder contains CSV files defining filters for each KPI specified in kpis.csv.

You can download an example of the ZIP file here: Download the Delta.zip file. This Delta.zip works with this example context BaseCaseDemo_FRonly

Delta Definition: `deltas.csv`#

The deltas.csv file defines the changes (deltas) applied to assets and parameters. For each delta, a series of actions either replace or rescale an attribute of an asset.

Available Actions

Replace: Replace the current value with another value (double, timeseries, or curve).
Rescale: Multiply the current value by a given factor.

Example: deltas.csv

deltas.csv contains serialized Deltas information.

name	object	data	type	value
low_demand	Power demand_FR	DEMAND	rescale	0.9
high_demand	Power demand_FR	DEMAND	rescale	1.1
no_nuke	Nuclear energy_FR	PMAX	replace	0
low_nuke	Nuclear energy_FR	PMAX	rescale	0.5
low_nuke	Gas production_FR	PRODUCTION_COST_CURVE	replace	#CURVE_1
high_nuke	Nuclear energy_FR	PMAX	replace	#TS1
high_nuke	Gas production_FR	PRODUCTION_COST_CURVE	replace	#CURVE_2

Replacement timeseries are prefixed with the usual tag #TS and curve data is prefixed with the tag #.

name: Name of the delta. A delta can apply modifications to multiple assets and parameters.
object: Name of the asset the delta is applied to. Use a wildcard (*) to match multiple assets (e.g., Solar* applies to all assets starting with “Solar”).
data: Attribute to modify.
type: Either replace or rescale.
value: A double, timeseries, or curve. Timeseries are prefixed with #TS, and curves are prefixed with #.

Associating Deltas with Reference Contexts: `paths.csv`#

The paths.csv file maps deltas to specific contexts. The user can specify up to 10 deltas per context or test case.

Example: paths.csv

paths.csv contains the delta paths defining which deltas to apply to each context and test case.

context	test_case	delta1	delta2
BCDemo_FRonly	*	no_nuke	low_demand
BCDemo_FRonly	*	low_nuke	low_demand
BCDemo_FRonly	*	high_nuke	low_demand
BCDemo_FRonly	*		low_demand
BCDemo_FRonly	*	no_nuke	high_demand
BCDemo_FRonly	*	low_nuke	high_demand
BCDemo_FRonly	*	high_nuke	high_demand
BCDemo_FRonly	*		high_demand

Wildcards (*) allow applying a combination of deltas to all test cases of a given context. Non-contiguous delta entries are not necessary, but they allow for convenient delta ordering.

context: Name of a context from the ACSG study.
test_case: Name of the test case for the context. Use * to apply the deltas to all test cases.
delta1, delta2, …, deltaN: Names of deltas from deltas.csv.

Defining Timeseries and Curves: `timeseries.csv` and `curves.csv`#

The timeseries.csv file contains timeseries data required for delta modifications.

Date: Column for dates or timestamps.
Other Columns: Each additional column contains a timeseries. The header must match the name in deltas.csv.

Example: timeseries.csv

timeseries.csv contains replacement timeseries.

Date	1
01/01/2030 00:00	100
01/01/2030 01:00	100
01/01/2030 02:00	100
01/01/2030 03:00	100
…	…
31/12/2030 23:00	100

The curves.csv file contains curve data required for deltas. Each curve has its own column with coordinates listed in the same column.

Example: curves.csv

curves.csv contains replacement curve data.

Each column defines a curve, a series of couples (x_i,y_i) where x_i is a breaking point and y_i is a slope. These numbers are expected to be separated by a semicolon (;) and are converted to integers. Typically, the x_i are interpreted as a percentage of Pmax (in case of production costs, for instance).

CURVE_1	CURVE_2
“20;30”	“20;5”
“70;80”	“80;25”
“90;210”	“90;50”

KPI Selection: `kpis.csv`#

The kpis.csv file specifies the KPIs to calculate for each context-delta combination.

kpi: List of KPIs.
to_compute: Indicates whether the KPI should be computed for all context combinations specified in paths.csv.

Note: Temporal KPIs cannot be computed for delta optimizations as they generate too large data volumes.

Warning

KPIs already available in your study will be overwritten with the delta module optimization.

Defining KPI Filters: `filters/`#

The filters/ folder contains CSV files defining filters for calculating KPIs. Each file corresponds to a KPI listed in kpis.csv.

If no filter file exists for a KPI, it will be calculated for all combinations of contexts and deltas.
Filter columns must match index columns in the results dataframe.

Example: KPIs Filters

This table outlines the filters for KPIs, specifying the parameters to apply during calculations.

energy	asset	technology	node
electricity	Hybrid_heat_pump_FR	*	*
*	Hybrid_heat_pump_DE	*	*
*	*	TRANSMISSION	*
electricity	*	PIPELINE	*
*	*	PIPELINE	FR

In this example, wildcards (*) are used to indicate that there are no specific filters applied for that column.

Before launching the delta module, ensure that KPIs and filters are properly defined to avoid performance bottlenecks due to large data volumes.