A Reliability-Optimized Maximum Power Point Tracking Algorithm Utilizing Neural Networks for Long-Term Lifetime Prediction for Photovoltaic Power Converters

Shahbazi, Mahmoud; Smith, Niall Andrew; Marzband, Mousa; Habib, Habib Ur Rahman

Abstract

The reliability of power converters in photovoltaic systems is critical to the overall system reliability. This paper proposes a novel active thermal-controlled algorithm that aims to reduce the rate of junction temperature increase, therefore, increasing the reliability of the device. The algorithm works alongside a normal perturb and observe maximum power point tracking algorithm, taking control when certain temperature criteria are met. In conjunction with a neural network, the algorithm is applied to long-term real mission profile data. This would grant a better understanding of the real-world trade-offs between energy generated and lifetime improvement when using the proposed algorithm, as well as shortening study cycle times. The neural network, when applied to 365 days of data, was 28 times faster than using standard electrothermal modeling, and the lifetime consumption was predicted with greater than 96.5% accuracy. Energy generated was predicted with greater than 99.5% accuracy. The proposed algorithm resulted in a 3.3% reduction in lifetime consumption with a 1.0% reduction in the total energy generated. There is a demonstrated trade-off between lifetime consumption reduction and energy-generated reduction. The results are also split by environmental conditions. Under very variable conditions, the algorithm resulted in a 4.4% reduction in lifetime consumption with a 1.4% reduction in the total energy generated.

Citation

Shahbazi, M., Smith, N. A., Marzband, M., & Habib, H. U. R. (2023). A Reliability-Optimized Maximum Power Point Tracking Algorithm Utilizing Neural Networks for Long-Term Lifetime Prediction for Photovoltaic Power Converters. Energies, 16(16), Article 6071. https://doi.org/10.3390/en16166071

• Journal Article Type: Article
• Acceptance Date: Aug 15, 2023
• Online Publication Date: Aug 19, 2023
• Publication Date: Aug 2, 2023
• Deposit Date: Aug 21, 2023
• Publicly Available Date: Aug 21, 2023
• Journal: Energies
• Electronic ISSN: 1996-1073
• Publisher: MDPI
• Peer Reviewed: Peer Reviewed
• Volume: 16
• Issue: 16
• Article Number: 6071
• DOI: https://doi.org/10.3390/en16166071
• Public URL: https://durham-repository.worktribe.com/output/1722437