As the world moves towards achieving net-zero emission targets, it is becoming increasingly important to harness renewable sources of energy such as solar power. While the installation costs for solar energy systems are rapidly declining, the costs associated with cleaning from soiling, which is the accumulated dirt on solar panels that can degrade performance and energy yield, remains a major concern for farm operators.

This project aims to address this issue by exploring smart detection approaches for calculating soiling rates through the analysis of weather and panel output datasets gathered from a research facility in Australia. The project will compare physical and artificial intelligent approaches used for time-series analysis and change-point detection for soiling loss.

The study will introduce concepts for data preparation, data analytics, machine learning, renewable energy systems modelling, and energy meteorology using open-source Python packages and publicly available datasets. By normalizing observed panel outputs by modelled power under clear and clean conditions, the study will accurately calculate soiling loss rates and provide insights into ways to reduce cleaning costs while maintaining optimal panel performance.

Overall, this project will contribute to the development of efficient and cost-effective renewable energy systems, which are essential for achieving net-zero emissions targets.


Photovoltaic and Renewable Energy Engineering

Research Area

Machine learning | Renewable energy systems modelling | Energy meteorology

High-Performance Computing Environment with applicants having a basic understanding of time series analysis and renewable energy systems modelling. A student with coding experience in python is desirable.

  • A broad survey of artificial intelligent and physical-based approaches for calculating soiling rates.
  • Michael G. Deceglie, Leonardo Micheli and Matthew Muller, "Quantifying Soiling Loss Directly From PV Yield," in IEEE Journal of Photovoltaics, 8(2), pp. 547-551, 2018 DOI: 10.1109/JPHOTOV.2017.2784682