PROPOSAL FOR A SOLAR IRRADIANCE MONITORING SYSTEM WITH LDR

Abstract

The study proposes the use of Light Dependent Resistors (LDRs) as a cost-effective and efficient alternative to pyranometers. It presents proposals for a model capable of converting values measured by LDR into solar irradiance data. The research describes the development of an Internet of Things-based monitoring system using the ThingSpeak platform. In this system, analog voltage signals from the LDR, which varies its resistivity according to luminosity, are captured and sent to the server for comparison with values measured by the pyranometer (W/m²). Data from both devices were collected over three weeks, limited to the daily operating interval of the pyranometer. In an initial analysis, the correlation between readings revealed a coefficient of determination (R²) of 0.87, demonstrating the existing correlation between LDR and pyranometer measurements. However, it is noteworthy that the LDR is not the most suitable for solar irradiance measurements but was employed here to evaluate the performance of machine learning methods. Thus, 12,565 measurements were obtained, with 80% used for training and 20% for testing the applied models, including Polynomial Regression (PR), Support Vector Machine (SVM), and Artificial Neural Network (ANN). Results indicate that SVM and ANN achieved an R² of 0.92 in testing, surpassing PR, which reached 0.9 in predicting solar irradiance based on LDR. SVM stood out due to its lower complexity and faster training time. However, upon analyzing the mean error betweenpyranometer measurements and those predicted by the models, it was observed that PR had an error of 22.48%, SVM registered 18.48%, and ANN showed an error of 20.43%, confirming that this sensor is not the most suitable for this purpose.