Physics-Based Method for Generating Fully Synthetic IV Curve Training Datasets for Machine Learning Classification of PV Failures

Abstract

Classification machine learning models require high-quality labeled datasets for training. Among the most useful datasets for photovoltaic array fault detection and diagnosis are module or string current-voltage (IV) curves. Unfortunately, such datasets are rarely collected due to the cost of high fidelity monitoring, and the data that is available is generally not ideal, often consisting of unbalanced classes, noisy data due to environmental conditions, and few samples. In this paper, we propose an alternate approach that utilizes physics-based simulations of string-level IV curves as a fully synthetic training corpus that is independent of the test dataset. In our example, the training corpus consists of baseline (no fault), partial soiling, and cell crack system modes. The training corpus is used to train a 1D convolutional neural network (CNN) for failure classification. The approach is validated by comparing the model’s ability to classify failures detected on a real, measured IV curve testing corpus obtained from laboratory and field experiments. Results obtained using a fully synthetic training dataset achieve identical accuracy to those obtained with use of a measured training dataset. When evaluating the measured data’s test split, a 100% accuracy was found both when using simulations or measured data as the training corpus. When evaluating all of the measured data, a 96% accuracy was found when using a fully synthetic training dataset. The use of physics-based modeling results as a training corpus for failure detection and classification has many advantages for implementation as each PV system is configured differently, and it would be nearly impossible to train using labeled measured data.

Artificial Intelligence in Photovoltaic Fault Identification and Diagnosis: A Systematic Review

“…However, after a rigorous filtering process adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) guidelines [6], 31 papers were retained for detailed examination. We found that many studies used I-V curves alongside ML for fault identification in PV arrays [7]. Techniques like Principal Component Analysis (PCA) [8] were commonly used for feature extraction, while complex PV systems adopted advanced methods, such as Recurrent Neural Networks (RNNs) [9] with satellite data and Convolutional Neural Networks (CNNs) [10], for analyzing voltage and current.…”

pvOps: a Python package for empirical analysis of photovoltaic field data

“…The package documentation for pvOps provides thorough examples exploring the various capabilities of each module. Additional details about the iv module capabilities are captured in (Hopwood et al, 2020;Hopwood, Stein, et al, 2022) while more information about the design and development of the text, timeseries, and text2time modules are captured in (Mendoza et al, 2021). Key package dependencies of pvOps include pandas (The pandas development team, 2020), sklearn (Pedregosa et al, 2011), nltk (Bird et al, 2009), and keras (Chollet & others, 2015) for analysis and matplotlib (Hunter, 2007), seaborn (Waskom, 2021), and plotly (Plotly Technologies Inc., 2015) for visualization.…”

Generating synthetic data for data-driven solutions via a digital twin for condition monitoring in machine tools

“…By their definition, faults or anomalous behaviour occurs much less frequently than normal operating conditions. 1 Because of this, data re-balancing efforts are usually required. However with faulty conditions there is also the issue that to collect data about faulty conditions, they need to occur, and this is far from ideal as faulty conditions usually lead to machine damage, or faulty components being manufactured.…”