energy_methods.pv_reliability#

Power-versus-age models and variation metrics from Kaaya et al. (2021).

These are deterministic extrapolations, not survival probabilities or warranties. The stretched-exponential coefficient has units year**(-beta); coefficients from different curve families are not interchangeable annual loss percentages.

energy_methods.pv_reliability.normalized_rmse(measured, predicted)[source]#

Equation 19, percent: RMSE/mean, or RMSE/range for exactly zero mean.

The paper concerns positive quantities. Negative means are rejected rather than producing negative errors. A constant zero reference is defined only for an exact match (zero error); otherwise its normalization is undefined.

Parameters:
  • measured (Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str])

  • predicted (Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str])

Return type:

float

energy_methods.pv_reliability.relative_difference(measured, predicted)[source]#

Equation 20: 100*abs(p-m)/max(abs(p),abs(m)); (0,0) is defined as 0.

This symmetric difference is not percentage error relative to the reference. Equal shapes are required; no implicit broadcasting or missing-value removal.

Parameters:
  • measured (Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str])

  • predicted (Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str])

Return type:

ndarray[tuple[Any, …], dtype[float64]]

class energy_methods.pv_reliability.ReliabilityCurve(kind, k, beta=1.0, gamma=1.0, mu=1.0)[source]#

Bases: object

Table 2 curve with time in years and power normalized to initial power.

k is fractional, never a numerical percent. beta belongs to the stretched exponential (Pan); gamma and mu to wear-out (Kaaya). Gamma and k are inseparable through their ratio and cannot both be fitted.

Parameters:
  • kind (Literal['linear', 'exponential', 'stretched_exponential', 'wear_out'])

  • k (float)

  • beta (float)

  • gamma (float)

  • mu (float)

kind: Literal['linear', 'exponential', 'stretched_exponential', 'wear_out']#
k: float#
beta: float#
gamma: float#
mu: float#
property k_units: str#

Dimensional coefficient units, with gamma treated as dimensionless.

power(age_years)[source]#

Evaluate Table 2; linear predictions are not clipped below zero.

Parameters:

age_years (Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str])

Return type:

ndarray[tuple[Any, …], dtype[float64]]

failure_time(retained_power=0.8)[source]#

Analytic crossing time, years, for a retained-power threshold in (0,1).

Parameters:

retained_power (float)

Return type:

float

class energy_methods.pv_reliability.ReliabilityFit(curve, nrmse_percent, observations, calibration_end_year, jacobian_condition)[source]#

Bases: object

Calibration result; residual error is not a prediction confidence interval.

Parameters:
  • curve (ReliabilityCurve)

  • nrmse_percent (float)

  • observations (int)

  • calibration_end_year (float)

  • jacobian_condition (float)

curve: ReliabilityCurve#
nrmse_percent: float#
observations: int#
calibration_end_year: float#
jacobian_condition: float#
energy_methods.pv_reliability.fit_reliability(age_years, normalized_power, *, kind, gamma=1.0)[source]#

Fit Table 2 with bounded, deterministic multistart least squares.

Fit k and (for nonlinear families) beta or mu. Wear-out gamma is FIXED to the caller’s value to avoid the exact k/gamma degeneracy. Ages must be distinct, increasing, measured from a declared origin. The normalization must be supplied; no hidden intercept, temperature correction or filtering. Shape bounds [0.05, 10] and k bounds [1e-10, 10] are numerical choices. Report the family and shape alongside k; do not rank lifetimes by fit alone.

Parameters:
  • age_years (Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str])

  • normalized_power (Buffer | _SupportsArray[dtype[Any]] | _NestedSequence[_SupportsArray[dtype[Any]]] | complex | bytes | str | _NestedSequence[complex | bytes | str])

  • kind (Literal['linear', 'exponential', 'stretched_exponential', 'wear_out'])

  • gamma (float)

Return type:

ReliabilityFit