Correlation-Based LHS of PV Inverters Electrothermal Model for Computationally Efficient Reliability Assessment

Reliability assessment of power electronic converters (PECs), especially photovoltaic (PV) inverters, requires detailed simulations to predict component aging and lifetime. This process uses electrothermal models driven by irradiance and temperature profiles to evaluate the thermal stress on their components. Surrogate models like Lookup Tables are commonly employed to reduce the high computational cost of this assessment, though they remain computationally demanding. This paper introduces an optimized sample generation approach using Latin Hypercube Sampling (LHS) with correlation-based space-filling criteria to reduce computational costs in PEC reliability assessment. By analyzing the pairwise correlation of samples, the proposed LHS method minimizes sample size without sacrificing accuracy. The study compares traditional one-factor-at-a-time (OFAT) sampling with optimized LHS, demonstrating that LHS achieves comparable accuracy to OFAT sampling, reducing computational demands to one-third, This approach offers a scalable and topology-independent method for reliability assessment, significantly improving the flexibility and efficiency in the design of PV inverters.