Photovoltaic prognostics and heath management using learning algorithms

Abstract

A novel model-based prognostics and health management (PHM) system has been designed to monitor the health of a photovoltaic (PV) system, measure degradation, and indicate maintenance schedules. Current state-of-the-art PV monitoring systems require module and array topology details or extensive modeling of the PV system. We present a method using an artificial neural network (ANN) which eliminates the need for a priori information by teaching the algorithm "good" performance behavior based on the initial performance of the array. The PHM algorithm was tested on two PV systems under test at the Outdoor Test Facility (OTF) at the National Renewable Energy Laboratory (NREL). The PHM algorithm was trained using two months of AC power production. The model then predicted the output power of the system using irradiance, wind, and temperature data. Based on the deviation in measured AC power from the AC power predicted by the trained ANN model, system outages and other faults causing a reduction in power were detected. Had these been commercial installations, rather than research installations, an alert for maintenance could have been initiated. Further use of the PHM system may be able to indicate degradation, detect module or inverter failures, or detect excessive soiling.

Photovoltaic performance characterization: Optimization by regression basis with application to health management

“…Synopses are shown in the following three charts below-where the solid lines are the medians and the bands are given by the 1 st and 3 rd quartiles of the distributions. 1 Solutions derived from the associated covariance matrix demonstrate significant detrimental instability with small changes in the input set or inverting the matrix becomes unstable or even illposed. From these figures we see that the linear method has higher positive spring and summer bias as well as a larger spread in the daily bias distribution for most months.…”

“…The standard approaches (see [1] and references therein) to characterizing plant performance and subsequently perform prognostics and health management (PHM)-which are suitable for fleet-wide implementation-fall into two basic categories: physical or component models-such as the Sandia Photovoltaic Array Performance Model [2]- [3] where system-specific performance parameters such as the module thermal, system electrical, and array configuration parameters are used without the need for prior performance or meteorological data; and mathematical models-such as traditional regression techniques or neural networks, where combinations of predictor variables based on measured historical meteorological inputs are chosen to form a basis for subsequent prediction. The regression techniques and neural network approaches differ in that the basis functions for the latter are not determined a priori or in a consistent manner.…”

Insights into metastability of photovoltaic materials at the mesoscale through massiveI–Vanalytics

“…Riley and Venayagamoorthy 43 employed a recurrent neural network to model maximum power time series behavior for a systematic model of PV performance and also described the utility of this methodology for prognostics and health management of PV systems over time. 44 Other authors more recently have been utilizing machine learning algorithms to better predict the behavior of PV systems beginning within the constraints of the diode model, but these approaches have utility in a more supervised manner as well. For example, genetic algorithms 45 and differential evolution has been used to parametrize the I-V curve for sample PV cells.…”

Statistical and Domain Analytics Applied to PV Module Lifetime and Degradation Science

“…It is currently not designed to determine the degradation rate of PV arrays or to quantitatively evaluate the impact of climatic stresses on a PV module's performance. This model serves as a reference for correcting PV modules' performance to standard test condition with real-world climatic data [25]- [27].…”