Machine Learning Algorithms Show Promise in Predicting PV-Powered Street Lighting Performance

Strategic Insights -

Researchers at the Department of Computer Engineering have conducted a comparative analysis of five machine learning algorithms to predict the operational conditions of PV-powered streetlights based on luminance levels and energy consumption. The study, published in IEEE Access, found that Deep Neural Networks (DNNs) and Deep Belief Networks (DBNs) achieved the lowest error rate (2.5%) and highest accuracy (97%) with high-quality data. However, the computational expense of these algorithms requires significant resources and lengthy training periods. In contrast, Linear Regression (LR) is faster and better suited for real-time applications, but shows the lowest performance in recall, F1 scores, and ROC-AUC.

Key Takeaways:

  • The study investigates the operational performance of PV-powered street lighting systems, highlighting the need for accurate data on energy consumption values corresponding to varying lighting intensities and environmental factors.
  • A comprehensive database was created and used to train five machine learning algorithms: Linear Regression (LR), Expectation Maximization (EM), Random Forest (RF), Deep Belief Networks (DBNs), and Deep Neural Networks (DNNs).
  • DNNs and DBNs achieved the lowest error rate (2.5%) and highest accuracy (97%) with high-quality data, while LR is distinguished by its rapid response time, enabling predictions with minimal training on large datasets.
  • The computational expense of DNNs and DBNs requires significant resources and lengthy training periods, whereas RF and LR are faster and better suited for real-time applications.
  • DNNs demonstrate superior performance in recall, F1 scores, and ROC-AUC, while LR shows the lowest performance.
  • The study highlights the importance of selecting the appropriate machine learning algorithm for real-world applications, considering factors such as computational expense, accuracy, and training time.

Statistics:

  • 2.5% error rate achieved by DNNs and DBNs
  • 97% accuracy achieved by DNNs and DBNs
  • 13.5 hours training period required for DNNs
  • 2.5 hours training period required for DBNs
  • 100% of luminance levels were accurately predicted by DNNs

Sources:

  • Operational Performance Assessment of PV-Powered Street Lighting: A Comparative Study of Different Machine Learning Prediction Models. IEEE Access, 2025, 13():135232-135253.
  • IEEE Access, publisher. ()
  • DOI: 10.1109/ACCESS.2025.3594171
  • DOI: 10.1109/ACCESS.2025.3594171
  • NewsRx. Study Data from Department of Computer Engineering Provide New Insights into Machine Learning (Operational Performance Assessment of PV-Powered Street Lighting: A Comparative Study of Different Machine Learning Prediction Models). Journal of Engineering. August 18, 2025; p 3069.