| Work Detail |
A scientist in Sweden has developed a new hybrid local features-based method using thermographs to identify faulty solar panels.
A researcher from Sweden’s Jönköping University has proposed a machine learning-based health monitoring approach for PV systems using infrared thermography.
The method is based on a hybrid local features-based approach to monitor panels, designed to resist scaling, noise, rotation, and haze. It achieved 98% training accuracy and 96.8% testing accuracy.
“Existing image processing-based machine learning approaches for health and fault diagnosis are often limited to specific datasets and suffer from issues such as sensitivity to rotation, scaling, noise, blurring, and haze,” said Dr. Waqas Ahmed in his paper. “These approaches also face trade-offs between memory usage and accuracy. Deep learning-based approaches, while powerful, have high computational complexity, memory, and processing requirements, and are prone to underfitting and overfitting without strong datasets and careful hyperparameter tuning.”
The novel method begins by capturing infrared thermographs with an infrared camera. In the preprocessing step, it assesses the quality of the thermographs, improving them with a dehazing algorithm and contrast adjustment in the grayscale channel if too much noise is present. Each thermograph is then divided into 5×5 pixel sub-thermographs.
Local features are extracted from each sub-thermograph using Gaussian and nonlinear methods. The most significant 80% of these features are retained, while irrelevant and redundant values are removed. A k-means unsupervised clustering algorithm then reduces the feature vector to 300 elements per thermograph for optimal memory use.
“Shallow classifiers, such as support vector machine (SVM), train the model on the feature vectors. A 5-fold cross-validation approach ensures proper model training,” said Ahmed. “A test vector from unseen thermographs is utilized to test the model’s accuracy in classifying the PV panels into three health-based classes: healthy, hotspot, and faulty.”
The novel method tested on a 44.24 kW crystalline silicon (c-Si) rooftop PV system in Lahore, Pakistan, consists of eight strings, each with 22 PV modules in series, totaling 5.28 kW. The system installs 376 PV modules, each rated at 240 W. Infrared thermographs were captured under ambient temperatures ranging from 32 C to 40 C, wind speeds of 6.9 m/s, and an irradiance level of 700 W/m². The thermographs were randomly divided, with 80% used for training and 20% for validation.
“The results of this study are particularly striking, with the method achieving a remarkable 98% training accuracy and 96.8% testing accuracy with 5-fold cross-validation,” said Ahmed. “Additionally, the model’s precision values of 92%, 100%, and 100%; recall values of 100%, 100%, and 90%; and F1 scores of 0.958, 1.0, and 0.947 for the faulty, healthy, and hotspot classes, respectively, indicate a high level of performance across these metrics.”
Compared to other artificial intelligence (AI) approaches in the literature, only the RB scale-invariant feature transform outperformed the proposed method, with a score of 98.66%. The strongest SURF scored 97.6%, deep neural features (pre-trained network) and shallow classifier scored 97%, RGB, LBP, mean HOG, and texture scored 96.8%, isolated CNN scored 96%, and Texture, HOG, and PCA scored 94.1%.
Ahmed presented the results in “Enhancing solar PV reliability with hybrid local features and infrared thermography,” which was recently published in Energy Reports. |
