[BurhanAbdullah]

Summary

This PR adds a common comparison layer for validating time-domain simulation results across power-system engines.

Different simulators can produce results using different time steps and event timestamps. Comparing their raw outputs directly can therefore be misleading. This implementation aligns the trajectories around the same physical event and compares equivalent channels over their common time interval.

The comparison reports:

• Mean absolute error and root mean square error
• Maximum absolute error and integrated absolute error
• Peak values from both simulations
• Missing channels, unit mismatches and insufficient overlap
• Whether each channel meets the configured tolerance

Implementation

The new powermcp.dynamic_validation module provides three main data structures: ChannelSeries, DynamicSnapshot and DynamicComparisonReport.

It uses linear interpolation within the observed time range and does not extrapolate. Channel identifiers and units must match explicitly; the implementation does not silently convert units or assume that differently named signals are equivalent.

The comparison also supports channel-specific tolerances and places a limit on the number of alignment points.

Tests

Added focused tests covering different solver time steps, event-time alignment, partial overlap, missing channels, unit mismatches, invalid inputs, tolerance handling and trajectory differences.

Scope

This PR establishes the comparison core. It does not yet include PSS/E or ANDES extraction adapters, nor does it claim to establish stability equivalence from waveform similarity alone. Metrics such as damping ratio, critical clearing time and rotor-angle stability can be added once the engine-specific outputs and study definitions are validated.

The aim is to provide a reliable foundation for cross-engine dynamic validation without making assumptions about simulator outputs.