Browsing by Author "Daniel-Durandt, Francisca Muriel"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Advancements in spectral corrections for enhanced photovoltaic performance monitoring and modelling.
    (Stellenbosch : Stellenbosch University, 2024-03) Daniel-Durandt, Francisca Muriel; Rix, Arnold J. ; Stellenbosch University. Faculty of Engineering. Dept. of Electrical and Electronic Engineering.
    ENGLISH ABSTRACT:Innovative advancements in spectral corrections for improved photovoltaic (PV) performance monitoring and modelling are presented, contributing to the sustainability of PV technology in the fight against climate change and reduced reliance on fossil fuels. Traditional performance metrics like the performance ratio (PR) have been improved for irradiance and temperature but not for the spectrum, a known PV performance influence. This study also emphasises the importance of accurate data quality standards for reliable PR calculations. The irradiance used to estimate PR requires decomposition and transposition models. Errors in these measurements translate to over or underestimations of PR. The quantification of the spectrum also remains an obstacle, as spectral variables in developing countries are rare. An assessment of the spectral effects on large PV plants indicated that Normal and weather-corrected PR decrease when air mass and precipitable water are higher. The wavelengths where the most significant changes in these spectral variables occur are when the PV module’s spectral response is higher; thus, a more pronounced spectral effect on performance and power output. Quality control checks are imperative for large irradiance datasets, and as the dataset size grows, these checks become more time-consuming. The study addressed this challenge by proposing an automated quality control procedure, initially evaluated in two case studies. After the automated procedure showed promising results, it underwent further validation across 24 stations within the SAURAN database. A review of the SAURAN database recommended each site’s data quality and identified irradiance patterns to group the stations in clusters. The study developed, tested, and validated a novel decomposition model for hourly direct normal irradiance (DNI) estimations in Southern Africa. The model development methodology utilised he SAURAN database and its quality-controlled stations. The models underwent validation for clearness index levels ranging from 0.175 to 0.875. The substantial reduction in errors in the estimation of DNI for the localised models demonstrated the validity of the model development. Further, using the SAURAN review’s clustering method, clustered models for different areas in Southern Africa were developed. The clustered decomposition models are especially useful when a potential PV plant is within the area. However, if no data is available, the data can be substituted using the SAURAN database. A Southern African model encompassing multiple climates, altitudes and geographical profiles showed enhanced DNI estimations, which has a universal application for the project assessed the sensitivity of spectral variables on the spectral mismatch to address the data scarcity problem in developing countries. The analysis showed how to substitute spectral data when none is available. Aerosol optical depth at 500nm showed a more significant effect on spectral irradiance than ozone, translating to spectral mismatch calculations. The project validated estimating ozone and a constant aerosol optical depth value for spectral mismatch calculation. Before developing the spectral correction methodology, the dataset must exclude potential biases not associated with the spectrum. A study of inverter-level performance analysis of the long-term degradation caused by wind and temperature effects on a utility-scale plant in a semi-arid region indicated that different plant areas degrade at different rates. Observations of higher degradation rates were associated with the areas with less wind and higher temperatures. The wind has a cooling effect, which enhances the PV performance. Another potential bias was the decomposition and transposition models’ inherent temporal bias on different metrics. Hourly-based decomposition models do not translate well to sub-hourly models. These can lead to over or underestimations for the irradiance, affecting the PR estimations. As real-time monitoring trends increase with data-driven techniques, eliminating the temporal bias becomes increasingly important. Using the newly developed decomposition models with anisotropic transposition models showed lower instances of over and underestimating the PR. Wrong estimations offer a distorted view of the plant’s current and predicted performance. A novel spectral correction method, when using short-term measurements, for correcting the PR to align better with the annual PR is presented in this work. The results indicated that, by correcting the spectrum, short-term measurements can predict the annual PR with superior performance over the Normal and weather-corrected PR. The model includes a geographical location-based variable, air mass, and a technology-based variable, a newly defined spectral correction factor with universal application. The novel approach is the first documented effort to address the spectrum’s influence on PR for utility-scale PV plants. The model demonstrated satisfactory temporal transferability from 5- minute to hourly to monthly aggregated data. The project produced significant advancements in PV performance monitoring and modelling by addressing various aspects of spectral corrections. The universal application has widespread applications in the PV industry, performance modelling and monitoring, and forecasting.