add reactive_power_sigma attribute

[mgovers]

Related to #237
Related to #369

Changes

• Add new RealValue attributes p_sigma and q_sigma to power sensor input and update. These are optional attributes.
  • If p_sigma and q_sigma are both NaN, the power_sigma is the apparent power sigma and is required (current behaviour) for state estimation.
  • If p_sigma and q_sigma are both not NaN, p_sigma is the active power sigma and q_sigma is the reactive power sigma.
  • It is not supported to provide only one of p_sigma and q_sigma, just like how it is not supported to only provide 1 of three phases of an asymmetric RealValue. The behavior is undefined in the core. In the validation function in Python side we raise an error.
  • Update data definition
  • Update documentation
• Create a new struct in calculation_parameters.hpp with PowerSensorCalcParam, VoltageSensorCalcParam, as the calculation parameter output of relevant sensor components. In VoltageSensorCalcParam is the same as old SensorCalcParam. PowerSensorCalcParam should have variance_p and variance_q, these are also RealValue. So it is three phase for asymmetric calculation.
• Change the sensor component to store power_sigma, p_sigma and q_sigma as member data. Store them as per unit value as it is now. Note: the base power for sym_power_sensor and asym_power_sensor is different. Change the way to generate PowerSensorCalcParam:
  • If p_sigma is NaN: Set variance_p = variance_q = power_sigma**2 / 2
    • For asymmetric calculation (for both sym_power_sensor and asym_power_sensor), variance_p = variance_q = power_sigma**2 / 2 * [1, 1, 1]
  • If p_sigma is not NaN: Set variance_p = p_sigma**2 and variance_q = q_sigma**2
    • For asymmetric calculation of sym_power_sensor, variance_p = p_sigma**2 * [1, 1, 1], same as Q.
    • For symmetric calculation of asym_power_sensor, variance_p = mean(p_sigma**2), same as Q.
  • If either p_sigma is provided and q_sigma is not, or vice versa. The behaviour is undefined. That's why we only check p_sigma in the above two cases.
  • If either p_sigma or q_sigma is inf, set both variance_p and variance_q to inf.
  • If p_sigma is not provided and power_sigma is inf, set both variance_p and variance_q to inf.
• Change the MeasuredValue logic, to split voltage and power measurement from one single main_value. For power measurement, aggregate and deaggregate P/Q separately based on variance_p and variance_q.
• For the current iterative linear solver, for a power measurement, the weighting factor is 1/(variance_p + variance_q). For asymmetric calculation, this is a diagonal matrix.

[mgovers]

@TonyXiang8787 this is ready for review. Remaining sonar cloud code smells are not caused by this PR

[mgovers]

remaining code smells either already on main or related to equivalent on main (see e.g. https://sonarcloud.io/project/issues?resolved=false&rules=cpp%3AS1709&id=PowerGridModel_power-grid-model&open=AYm75yP7TgVATTPnBGJv )

[mgovers]

@TonyXiang8787 i think i resolved all comments and it's ready for review again

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
9 Code Smells

98.7% Coverage
0.0% Duplication