This paper explains the fusion of the physics-based material degradation mechanism with the statistics-based data modeling approach for predicting the degradation rate of photovoltaic (PV) modules. The degradation of PV module is mainly associated with the module construction type and climatic condition at its use location. The aim of this paper is to quantify the effect of dynamic environmental stresses (dynamic covariates) on the power degradation of the module over its lifetime. There are various physics-based models, such as Arrhenius model, for understanding the physical or chemical reaction-related root causes of PV degradation. But, to estimate the underlying material properties, such as activation energy (Ea), statistical modeling plays a key role. In addition, instead of being continuously monitored, the performance characteristics of PV modules are often measured only at intervals like quarterly or annually, which makes it difficult to model the complete degradation path of the module. On the other hand, the information on dynamic covariates is recorded more frequently with the development of sophisticated sensors and data acquisition systems. This information can be integrated through physics-based models to study the effects of environmental variables in degradation processes. Hence, in this paper, a cumulative exposure model is used to link the module degradation path and the environmental variables, including module temperature (both static and cyclic), ultraviolet radiation, and relative humidity, which are recorded as multivariate time series.

Encapsulant discoloration is a well-known field degradation mode of crystalline-silicon photovoltaic modules, particularly in the hot climate zones. The discoloration rate is influenced by several weathering factors, such as UV light, module temperature, and humidity, as well as the permeability of oxygen into the module. In this article, a rate dependence model employing the modified Arrhenius equations to predict the degradation rate for encapsulant discoloration in different climates is presented. Two modeling approaches are introduced, which utilize the field and accelerated UV testing degradation data in conjunction with the field meteorological data to determine the acceleration factor for encapsulant browning. A novel method of accelerated UV stress testing at three simultaneous module temperatures in a single environmental chamber test run is implemented to estimate the activation energy for browning. The test was performed on three field-retrieved modules to capture the wear-out failure mechanism. The degradation in short-circuit current Isc rather than maximum power is used as a decisive parameter for the discoloration analysis. Furthermore, the developed model has been used to predict the Isc degradation rate for the Arizona field characterized by a hot and dry climate and is validated against the field-measured value. It has also been applied to other climate types, e.g., the cold and dry climate of New York.

This paper presents a novel single loop approach to design the components of the load sharing systems by optimally allocating the failure probabilities to each component, thereby satisfying the overall system reliability requirement. The Reliability–Based Design Optimization (RBDO) of load sharing systems is computationally intensive due to the dynamic nature of component failure probabilities, since the failure of one component will vary the failure probabilities of other working components. Many RBDO methods have been successfully utilized to design individual components, however using these methods for handling system level reliability constraints is still a challenging task. This is because of a drop in accuracy and computational efficiency, especially when considering a load sharing system, where there is dependency in failure probabilities of components. The key idea is to integrate Stress–Strength Interference (SSI) theory with discrete (or) continuous time-discrete state Markov model for the reliability assessment of system, with the states being the condition of components (working/failed). This method takes advantage of the state transition probability matrix to represent the dynamic nature of the system performance. A numerical example of a simple load sharing system with two I-Beams is presented to illustrate and evaluate the performance of the proposed methodology.