Publications

The timely detection of photovoltaic (PV) system failures is important for maintaining optimal performance and lifetime reliability. A main challenge remains the lack of a unified health-state architecture for the uninterrupted monitoring and predictive performance of PV systems. To this end, existing failure detection models are strongly dependent on the availability and quality of site-specific historic data. The scope of this work is to address these fundamental challenges by presenting a health-state architecture for advanced PV system monitoring. The proposed architecture comprises of a machine learning model for PV performance modeling and accurate failure diagnosis. The predictive model is optimally trained on low amounts of on-site data using minimal features and coupled to functional routines for data quality verification, whereas the classifier is trained under an enhanced supervised learning regime. The results demonstrated high accuracies for the implemented predictive model, exhibiting normalized root mean square errors lower than 3.40% even when trained with low data shares. The classification results provided evidence that fault conditions can be detected with a sensitivity of 83.91% for synthetic power-loss events (power reduction of 5%) and of 97.99% for field-emulated failures in the test-bench PV system. Finally, this work provides insights on how to construct an accurate PV system with predictive and classification models for the timely detection of faults and uninterrupted monitoring of PV systems, regardless of historic data availability and quality. Such guidelines and insights on the development of accurate health-state architectures for PV plants can have positive implications in operation and maintenance and monitoring strategies, thus improving the system’s performance.

Abstract not provided.

The precise estimation of performance loss rate (PLR) of photovoltaic (PV) systems is vital for reducing investment risks and increasing the bankability of the technology. Until recently, the PLR of fielded PV systems was mainly estimated through the extraction of a linear trend from a time series of performance indicators. However, operating PV systems exhibit failures and performance losses that cause variability in the performance and may bias the PLR results obtained from linear trend techniques. Change-point (CP) methods were thus introduced to identify nonlinear trend changes and behaviour. The aim of this work is to perform a comparative analysis among different CP techniques for estimating the annual PLR of eleven grid-connected PV systems installed in Cyprus. Outdoor field measurements over an 8-year period (June 2006-June 2014) were used for the analysis. The obtained results when applying different CP algorithms to the performance ratio time series (aggregated into monthly blocks) demonstrated that the extracted trend may not always be linear but sometimes can exhibit nonlinearities. The application of different CP methods resulted to PLR values that differ by up to 0.85% per year (for the same number of CPs/segments).

Geographic analysis of photovoltaic (PV) performance factors across large regions can help relevant stakeholders make informed, and reduced risk decisions. High temporal and spatial resolution meteorological data from the National Solar Radiation Database are used to investigate performance and cost as an effect of varying system characteristics such as the module temperature coefficients, mounting configurations and coatings. The results demonstrated the strong climatic dependence that these characteristics have on annual energy yield whereas the revenues were dominated by the electricity price.

Accurate quantification of photovoltaic (PV) system degradation rate (RD) is essential for lifetime yield predictions. Although RD is a critical parameter, its estimation lacks a standardized methodology that can be applied on outdoor field data. The purpose of this paper is to investigate the impact of time period duration and missing data on RD by analyzing the performance of different techniques applied to synthetic PV system data at different linear RD patterns and known noise conditions. The analysis includes the application of different techniques to a 10-year synthetic dataset of a crystalline Silicon PV system, with emulated degradation levels and imputed missing data. The analysis demonstrated that the accuracy of ordinary least squares (OLS), year-on-year (YOY), autoregressive integrated moving average (ARIMA) and robust principal component analysis (RPCA) techniques is affected by the evaluation duration with all techniques converging to lower RD deviations over the 10-year evaluation, apart from RPCA at high degradation levels. Moreover, the estimated RD is strongly affected by the amount of missing data. Filtering out the corrupted data yielded more accurate RD results for all techniques. It is proven that the application of a change-point detection stage is necessary and guidelines for accurate RD estimation are provided.

Operation and maintenance (OM) and monitoring strategies are important for safeguarding optimum photovoltaic (PV) performance while also minimizing downtimes due to faults. An OM decision support system (DSS) was developed in this work for providing recommendations of actionable decisions to resolve fault and performance loss events. The proposed DSS operates entirely on raw field measurements and incorporates technical asset and financial management features. Historical measurements from a large-scale PV system installed in Greece were used for the benchmarking procedure. The results demonstrated the financial benefits of performing mitigation actions in case of near zero power production incidents. Stochastic simulations that consider component malfunctions and failures exhibited a net economic gain of approximately 4.17 €/kW/year when performing OM actions. For an electricity price of 59.98 €/MWh, a minimum of 8.4% energy loss per year is required for offsetting the annualized OM cost value of 7.45 €/kW/year calculated by the SunSpec/National Renewable Energy Laboratory (NREL) PV OM Cost Model.

Many Electric Power Systems (EPS) already include geographically dispersed photovoltaic (PV) systems. These PV systems may not be co-located with highest-priority loads and, thus, easily integrated into a microgrid; rather PV systems and priority loads may be far away from one another. Furthermore, because of the existing EPS configuration, non-critical loads between the distant PV and critical load(s) cannot be selectively disconnected. To achieve this, the proposed approach finds ideal switch locations by first defining the path between the critical load and a large PV system, then identifies all potential new switch locations along this path, and finally discovers switch locations for a particular budget by finding the ones the produce the lowest Loss of Load Probability (LOLP), which is when load exceed generation. Discovery of the switches with the lowest LOLP involves a Particle Swarm Optimization (PSO) implementation. The objective of the PSO is to minimize the microgird's LOLP. The approach assumes dynamic microgrid operations, where both the critical and non-critical loads are powered during the day and only the critical load at night. To evaluate the approach, this paper includes a case study that uses the topology and Advanced Metering Infrastructure (AMI) data from an actual EPS. For this example, the assessment found new switch locations that reduced the LOLP by up to 50% for two distant PV location scenarios.

A linear performance drop is generally assumed during the photovoltaic (PV) lifetime. However, operational data demonstrate that the PV module degradation rate (Rd) is often nonlinear, which, if neglected, may increase the financial uncertainty. Although nonlinear behavior has been the subject of numerous publications, it was only recently that statistical models able to detect change-points and extract multiple Rd values from PV performance time-series were introduced. A comparative analysis of six open-source libraries, which can detect change-points and calculate nonlinear Rd, is presented in this article. Since the real Rd and change-point locations are unknown in field data, 960 synthetic datasets from six locations and two PV module technologies have been generated using different aggregation and normalization decisions and nonlinear degradation rate patterns. The results demonstrated that coarser temporal aggregation (i.e., monthly vs. weekly), temperature correction, and both PV module technologies and climates with lower seasonality can benefit the change-point detection and Rd extraction. This also raises a concern that statistical models typically deployed for Rd analysis may be highly climatic-and technology-dependent. The comparative analysis of the six approaches demonstrated median mean absolute errors (MAE) ranging from 0.06 to 0.26%/year, given a maximum absolute Rd of 2.9%/year. The median MAE in change-point position detection varied from 3.5 months to 6 years.

Abstract not provided.

Corrective maintenance strategies are important for safeguarding optimum photovoltaic (PV) performance while also minimizing downtimes due to failures. In this work, a complete operation and maintenance (OM) decision support system (DSS) was developed for corrective maintenance. The DSS operates entirely on field measurements and incorporates technical asset and financial management features. It was validated experimentally on a large-scale PV system installed in Greece and the results demonstrated the financial benefits of performing corrective actions in case of failures and reversible loss mechanisms. Reduced response and resolution times of corrective actions could improve the PV power production of the test PV plant by up to 2.41%. Even for 1% energy yield improvement by performing corrective actions, a DSS is recommended for large-scale PV plants (with a peak capacity of at least 250 kWp).

The IEA PVPS Task 13 group, experts who focus on photovoltaic performance, operation, and reliability from several leading R&D centers, universities, and industrial companies, is developing a framework for the calculation of performance loss rates of a large number of commercial and research photovoltaic (PV) power plants and their related weather data coming across various climatic zones. The general steps to calculate the performance loss rate are (i) input data cleaning and grading; (ii) data filtering; (iii) performance metric selection, corrections, and aggregation; and finally, (iv) application of a statistical modeling method to determine the performance loss rate value. In this study, several high-quality power and irradiance datasets have been shared, and the participants of the study were asked to calculate the performance loss rate of each individual system using their preferred methodologies. The data are used for benchmarking activities and to define capabilities and uncertainties of all the various methods. The combination of data filtering, metrics (performance ratio or power based), and statistical modeling methods are benchmarked in terms of (i) their deviation from the average value and (ii) their uncertainty, standard error, and confidence intervals. It was observed that careful data filtering is an essential foundation for reliable performance loss rate calculations. Furthermore, the selection of the calculation steps filter/metric/statistical method is highly dependent on one another, and the steps should not be assessed individually.

Photovoltaic (PV) soiling profiles exhibit a sawtooth shape, where cleaning events and soiling deposition periods alternate. Generally, the rate at which soiling accumulates is assumed to be constant within each deposition period. In reality, changes in rates can occur because of sudden variations in climatic conditions, e.g., dust storms or prolonged periods of rain. The existing models used to extract the soiling profile from the PV performance data might fail to capture the change points and occasionally estimate incorrect soiling profiles. This work analyzes how the introduction of change points can be beneficial for soiling extraction. Data from nine soiling stations and a 1-MW site were analyzed by using piecewise regression and three change point detection algorithms. The results showed that accounting for change points can provide significant benefits to the modeling of soiling even if not all the change point algorithms return the same improvements. Considering change points in historical trends is found to be particularly important for studies aiming to optimize cleaning schedules.

The IEC 61853 photovoltaic (PV) module energy rating standard requires measuring module power (and hence, efficiency) over a matrix of irradiance and temperature conditions. These matrix points represent nearly the full range of operating conditions encountered in the field in all but the most extreme locations and create an opportunity to develop alternative approaches for calculating system performance. In this article, a new PV module efficiency model is presented and compared with five published models using matrix data collected from four different PV module types. The results of the comparative analysis demonstrated that the new model improves on the existing ones exhibiting root-mean-square errors in normalized efficiency well below 0.01 for all cases and PV modules. The analysis also highlighted its ability to interpolate and extrapolate performance between and beyond measured matrix points of irradiance and temperature, establishing it as a robust yet relatively simple model for several applications that are detailed throughout this article.

A main challenge for integrating the intermittent photovoltaic (PV) power generation remains the accuracy of day-ahead forecasts and the establishment of robust performing methods. The purpose of this work is to address these technological challenges by evaluating the day-ahead PV production forecasting performance of different machine learning models under different supervised learning regimes and minimal input features. Specifically, the day-ahead forecasting capability of Bayesian neural network (BNN), support vector regression (SVR), and regression tree (RT) models was investigated by employing the same dataset for training and performance verification, thus enabling a valid comparison. The training regime analysis demonstrated that the performance of the investigated models was strongly dependent on the timeframe of the train set, training data sequence, and application of irradiance condition filters. Furthermore, accurate results were obtained utilizing only the measured power output and other calculated parameters for training. Consequently, useful information is provided for establishing a robust day-ahead forecasting methodology that utilizes calculated input parameters and an optimal supervised learning approach. Finally, the obtained results demonstrated that the optimally constructed BNN outperformed all other machine learning models achieving forecasting accuracies lower than 5%.

Data integrity is crucial for the performance and reliability analysis of photovoltaic (PV) systems, since actual in-field measurements commonly exhibit invalid data caused by outages and component failures. The scope of this paper is to present a complete methodology for PV data processing and quality verification in order to ensure improved PV performance and reliability analyses. Data quality routines (DQRs) were developed to ensure data fidelity by detecting and reconstructing invalid data through a sequence of filtering stages and inference techniques. The obtained results verified that PV performance and reliability analyses are sensitive to the fidelity of data and, therefore, time series reconstruction should be handled appropriately. To mitigate the bias effects of 10% or less invalid data, the listwise deletion technique provided accurate results for performance analytics (exhibited a maximum absolute percentage error of 0.92%). When missing data rates exceed 10%, data inference techniques yield more accurate results. The evaluation of missing power measurements demonstrated that time series reconstruction by applying the Sandia PV Array Performance Model yielded the lowest error among the investigated data inference techniques for PV performance analysis, with an absolute percentage error less than 0.71%, even at 40% missing data rate levels. The verification of the routines was performed on historical datasets from two different locations (desert and steppe climates). The proposed methodology provides a set of standardized analytical procedures to ensure the validity of performance and reliability evaluations that are performed over the lifetime of PV systems.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Although common practice for estimating photovoltaic (PV) degradation rate (RD) assumes a linear behavior, field data have shown that degradation rates are frequently nonlinear. This article presents a new methodology to detect and calculate nonlinear RD based on PV performance time-series from nine different systems over an eight-year period. Prior to performing the analysis and in order to adjust model parameters to reflect actual PV operation, synthetic datasets were utilized for calibration purposes. A change-point analysis is then applied to detect changes in the slopes of PV trends, which are extracted from constructed performance ratio (PR) time-series. Once the number and location of change points is found, the ordinary least squares method is applied to the different segments to compute the corresponding rates. The obtained results verified that the extracted trends from the PR time-series may not always be linear and therefore, 'nonconventional' models need to be applied. All thin-film technologies demonstrated nonlinear behavior whereas nonlinearity detected in the crystalline silicon systems is thought to be due to a maintenance event. A comparative analysis between the new methodology and other conventional methods demonstrated levelized cost of energy differences of up to 6.14%, highlighting the importance of considering nonlinear degradation behavior.

A main challenge towards ensuring large-scale and seamless integration of photovoltaic systems is to improve the accuracy of energy yield forecasts, especially in grid areas of high photovoltaic shares. The scope of this paper is to address this issue by presenting a unified methodology for hourly-averaged day-ahead photovoltaic power forecasts with improved accuracy, based on data-driven machine learning techniques and statistical post-processing. More specifically, the proposed forecasting methodology framework comprised of a data quality stage, data-driven power output machine learning model development (artificial neural networks), weather clustering assessment (K-means clustering), post-processing output optimisation (linear regressive correction method) and the final performance accuracy evaluation. The results showed that the application of linear regression coefficients to the forecasted outputs of the developed day-ahead photovoltaic power production neural network improved the performance accuracy by further correcting solar irradiance forecasting biases. The resulting optimised model provided a mean absolute percentage error of 4.7% when applied to historical system datasets. Finally, the model was validated both, at a hot as well as a cold semi-arid climatic location, and the obtained results demonstrated close agreement by yielding forecasting accuracies of mean absolute percentage error of 4.7% and 6.3%, respectively. The validation analysis provides evidence that the proposed model exhibits high performance in both forecasting accuracy and stability.

Optimum and reliable photovoltaic (PV) plant performance requires accurate diagnostics of system losses and failures. Data-driven approaches can classify such losses however, the appropriate PV data features required for accurate classification remains unclear. To avoid misclassification, this study reviews the potential issues associated with inabilities to separate fault conditions that overlap using certain data features. Feature selection techniques that define each feature's importance and identify the set of features necessary for producing the most accurate results are also explored. The experiment quantified the amount of overlap using both maximum power point (MPP) and current and voltage (I-V) curve data sets. The I -V data provided an overall increase in classification accuracy of 8% points above the case where only MPP was available.

Optimum and reliable photovoltaic (PV) plant performance requires accurate diagnostics of system losses and failures. Data-driven approaches can classify such losses however, the appropriate PV data features required for accurate classification remains unclear. To avoid misclassification, this study reviews the potential issues associated with inabilities to separate fault conditions that overlap using certain data features. Feature selection techniques that define each feature's importance and identify the set of features necessary for producing the most accurate results are also explored. The experiment quantified the amount of overlap using both maximum power point (MPP) and current and voltage (I-V) curve data sets. The I -V data provided an overall increase in classification accuracy of 8% points above the case where only MPP was available.

It is a common approach to assume a constant performance drop during the photovoltaic (PV) lifetime. However, operational data demonstrated that PV degradation rate (R_{D}) may exhibit nonlinear behavior, which neglecting it may increase financial risks. This study presents and compares three approaches, based on open-source libraries, which are able to detect and calculate nonlinear R_{D}. Two of these approaches include trend extraction and change-point detection methods, which are frequently used statistical tools. Initially, the processed monthly PV performance ratio (PR) time-series are decomposed in order to extract the trend and change-point analysis techniques are applied to detect changes in the slopes. Once the number of change-points is optimized by each model, the ordinary least squares (OLS) method is applied on the different segments to compute the corresponding rates. The third methodology is a regression analysis method based on simultaneous segmentation and slope extraction. Since the 'real' R_{D} value is an unknown parameter, this investigation was based on synthetic datasets with emulated two-step degradation rates. As such, the performance of the three approaches was compared exhibiting mean absolute errors ranging from 0 to 0.46%/year whereas the change-point position detection differed from 0 to 10 months.

Abstract not provided.

Dust accumulation significantly affects the performance of photovoltaic modules and its impact can be mitigated by various cleaning methods. Optimizing the cleaning frequency is essential to minimize the soiling losses and, at the same time, the costs. However, the effectiveness of cleaning lowers with time because of the reduced energy yield due to degradation. Additionally, economic factors such as the escalation in electricity price and inflation can compound or counterbalance the effect of degradation on the soiling mitigation profits. The present study analyzes the impact of degradation, escalation in electricity price and inflation on the revenues and costs of cleanings and proposes a methodology to maximize the profits of soiling mitigation of any system. The energy performance and soiling losses of a 1 MW system installed in southern Spain were analyzed and integrated with theoretical linear and nonlinear degradation rate patterns. The Levelized Cost of Energy and Net Present Value were used as criteria to identify the optimum cleaning strategies. The results showed that the two metrics convey distinct cleaning recommendations, as they are influenced by different factors. For the given site, despite the degradation effects, the optimum cleaning frequency is found to increase with time of operation.

Abstract not provided.