known-issues.md

Known Issues

Missing Output Files

It has been observed that some generated files are missing.

Current Issue

For unknown reasons, certain files are absent from the generated output. This file lists missing files.

We are currently investigating the cause of this issue and exploring potential solutions.

---

Changes in Element IDs or Names

Current Issue

At times, certain elements within the electrical network experience changes in their names or may be replaced.

Our policy regarding these changes is as follows:

• If an element's name or ID changes, we aim to replace the new identifier with the old one to avoid impacting neural network training.
• However, if an element is replaced due to damage or malfunction and the new element does not have identical characteristics (e.g., maximum intensity or voltage acceptance), it is treated as a distinct element.

In practice, when element IDs change, the detection of the new name is not always performed correctly. This often leads to inconsistencies.

Solution Approach

The detection mechanism is being improved to consider additional fields for better tracking of changes.

1. For each element, all relevant identifying fields are concatenated using a "__" (double underscore) separator to form a unique persistent_id. This persistent_id remains consistent for a given element across different files.
2. When comparing two files, we:
   • Identify persistent_id values that were present in the first file but missing in the second.
   • Identify persistent_id values that appear in the second file but were absent in the first.
3. To match elements from the old file to the new file, we assume that changes in ID, name, and persistent_id should be relatively minor. We use Levenshtein distance calculations to find the best possible match.
4. Steps 2 and 3 are repeated for all files in the dataset.
5. Once the matrix mapping changing elements is established, all new IDs are replaced with their corresponding old IDs.

---

Example Scenario

We illustrate this approach with a simple example.

First File:

<?xml version="1.0" encoding="UTF-8"?>
<iidm:network xmlns:iidm="http://www.powsybl.org/schema/iidm/1_12" id="snapshot_2021-01-02-1000" caseDate="2021-01-02T10:00:00.000+01:00">
...
    <iidm:voltageLevel id="ARGIAP6" nominalV="225.0">
        <iidm:switch id="ARGIAP6_ARGIA   6TR631    SA.1" name="ARGIA   6TR631    SA.1" />
        <iidm:switch id="ARGIAP6_ARGIA   6AT761    SA.1" name="ARGIA   6AT761    SA.1" />
    </iidm:voltageLevel>    
...
</iidm:network>

Second File:

<?xml version="1.0" encoding="UTF-8"?>
<iidm:network xmlns:iidm="http://www.powsybl.org/schema/iidm/1_12" id="snapshot_2021-01-02-1100" caseDate="2021-01-02T11:00:00.000+01:00">
...
    <iidm:voltageLevel id="ARGIAP6" nominalV="225.0">
        <iidm:switch id="ARGIAP6_ARGIA   6AT762    SA.1" name="ARGIA   6AT761    SA.1" />
        <iidm:switch id="ARGIAP6_ARGIA   6TR631    SA.1" name="ARGIA   6TR632    SA.1" />
    </iidm:voltageLevel>    
...
</iidm:network>

Observations:

• The second file is more recent (see the caseDate attribute).
• The first switch has a different ID in the second file.
• The second switch has a different name in the second file.
• The order of the two switches has changed, which can occur in real-world datasets.

Persistent ID Computation

For switches, the persistent_id is generated using:

• Voltage level id
• Switch id
• Switch name
• Switch nodes (node1 and node2)

Generated persistent_id values:

First File:

• ARGIAP6__ARGIAP6_ARGIA 6TR631 SA.1__ARGIA 6TR631 SA.1__0__18 → switch_1_a
• ARGIAP6__ARGIAP6_ARGIA 6AT761 SA.1__ARGIA 6AT761 SA.1__0__6 → switch_2_a

Second File:

• ARGIAP6__ARGIAP6_ARGIA 6AT762 SA.1__ARGIA 6AT761 SA.1__0__6 → switch_1_b
• ARGIAP6__ARGIAP6_ARGIA 6TR631 SA.1__ARGIA 6TR632 SA.1__0__18 → switch_2_b

Levenshtein Distance Calculation:

	Distance to switch_1_a	Distance to switch_1_b
switch_2_a	8	2
switch_2_b	1	11

Based on these distances, the algorithm correctly identifies:

• switch_1_a corresponds to switch_2_b
• switch_1_b corresponds to switch_2_a

This aligns with the observed inversion of switch order in the second file.

Final Adjustments

After resolving ID mismatches, the second file is updated:

<?xml version="1.0" encoding="UTF-8"?>
<iidm:network xmlns:iidm="http://www.powsybl.org/schema/iidm/1_12" id="snapshot_2021-01-02-1100" caseDate="2021-01-02T11:00:00.000+01:00">
...
    <iidm:voltageLevel id="ARGIAP6" nominalV="225.0">
        <iidm:switch id="ARGIAP6_ARGIA   6AT761    SA.1" name="ARGIA   6AT761    SA.1" />
        <iidm:switch id="ARGIAP6_ARGIA   6TR631    SA.1" name="ARGIA   6TR631    SA.1" />
    </iidm:voltageLevel>    
...
</iidm:network>

Conclusion

While the switch order remains different, the correct identifiers have been restored, ensuring consistency across files.