This article analyzes the relationship between artificial intelligence (AI) and photovoltaic (PV) systems. Solar energy is one of the most important renewable energies, and the investment of businesses and governments is increasing every year. AI is used to solve the most important problems found in PV systems, such as the tracking of the Max Power Point of the PV modules, the forecasting of the energy produced by the PV system, the estimation of the parameters of the equivalent model of PV modules or the detection of faults found in PV modules or cells. AI techniques perform better than classical approaches, even though they have some limitations such as the amount of data and the high computation times needed for performing the training. Research is still being conducted in order to solve these problems and find techniques with better performance. This article analyzes the most relevant scientific works that use artificial intelligence to deal with the key PV problems by searching terms related with artificial intelligence and photovoltaic systems in the most important academic research databases. The number of publications shows that this field is of great interest to researchers. The findings also show that these kinds of algorithms really have helped to solve these issues or to improve the previous solutions in terms of efficiency or accuracy.

Affordable and clean energy is one of the Sustainable Development Goals (SDG). SDG compliance and economic crises have boosted investment in solar energy as an important source of renewable generation. Nevertheless, the complex maintenance of solar plants is behind the increasing trend to use advanced artificial intelligence techniques, which critically depend on big amounts of data. In this work, a model based on Deep Convolutional Generative Adversarial Neural Networks (DCGANs) was trained in order to generate a synthetic dataset made of 10,000 electroluminescence images of photovoltaic cells, which extends a smaller dataset of experimentally acquired images. The energy output of the virtual cells associated with the synthetic dataset is predicted using a Random Forest regression model trained from real IV curves measured on real cells during the image acquisition process. The assessment of the resulting synthetic dataset gives an Inception Score of 2.3 and a Fréchet Inception Distance of 15.8 to the real original images, which ensures the excellent quality of the generated images. The final dataset can thus be later used to improve machine learning algorithms or to analyze patterns of solar cell defects.

In photovoltaic power plant inspections, techniques for module assessment play a crucial role as they enhance fault detection and module characterization. One valuable technique is luminescence. The present paper introduces a novel technique termed passive luminescence. It enhances both electroluminescence and photoluminescence imaging acquisition in photovoltaic power plants under normal operation in high irradiance conditions. This technique is based on the development of an electronic board, which allows the polarity of the module to be changed, enabling the current generated by the photovoltaic string to be injected into the module and producing electroluminescence effects. Additionally, the board can bypass the module and set an open circuit, inducing photoluminescence emission using sunlight as an excitation source. The proper coordination of the board and an InGaAs camera with a bandpass filter has allowed for the integration of a lock-in technique, which has produced electroluminescence and photoluminescence pictures that can be used for fault detection.

In this work, a method to predict the output power of Photovoltaic (PV) cells using their Electroluminescence (EL) images is presented. The data used includes Electroluminescence Images and the value of the Max Power Point computed from the Current–Voltage Curve of the cells. The method is used as follows: Firstly, the images are preprocessed to improve their quality. After that, a comparison between different Machine Learning methods from Traditional ones, such as Random Forest or Gradient Boosting, to Deep Learning methods, such as Recurrent Neural Networks or Convolutional Neural Networks is performed. Another significant contribution of this paper is that it analyzes the problem of unbalanced data, trying to solve it using Synthetic Images created by a Generative Adversarial Network. Our results show that the best model is the Gradient-Boosting based method using a pre-trained Resnet50 as a feature extraction method with a Mean Absolute Error (MAE) of 0.0341 and a Mean Squared Error (MSE) of 0.00211. The results also shows how the models trained with the unbalanced dataset are capable of obtaining results similar to the models trained with the balanced dataset.

An adaptive neuro-fuzzy inference scheme for defect detection and classification of solar PV cells