A rolling-horizon cleaning recommendation system for dust removal of industrial PV panels

Chao Zhang, Yunfeng Ma, Zengqiang Mi, Fan Yang, Long Zhang

Research output: Contribution to journalArticlepeer-review


Periodic cleaning of photovoltaic (PV) panels, such as every three months, is a common industry practice. However, this fixed period may not be optimal for maximizing the profit of a PV power generation system, due to numerous time-variant influencing factors, such as weather, temperature, etc. To increase the overall profit of a solar farm, it is highly desirable to have a flexible cleaning schedule that considers time-variant influencing factors. For this requirement, a rolling-horizon cleaning recommendation system is presented in this paper. Within this cleaning recommendation system, a prediction model and profit model are proposed. The prediction model, called the ensemble long-term and nonlinear autoregressive, can provide a time-variant future horizon by analyzing and compressing the time-variant characteristics in historical information. The profit model based on mathematical constraints, can process time-variant future horizon output from prediction model to generate a flexible optimized recommendation for cleaning schedule. The effectiveness of the proposed system is validated in real farms and all data used in this paper is collected from real world. The two case studies in experiments show that the profit improvement can reach up to 6% and 30%, respectively.
Original languageEnglish
Article number122168
JournalApplied Energy
Issue numberB
Early online date31 Oct 2023
Publication statusPublished - 1 Jan 2024


  • Renewable energyDust removalPV panelRolling-horizonNonlinear autoregressive modelRecommendation system


Dive into the research topics of 'A rolling-horizon cleaning recommendation system for dust removal of industrial PV panels'. Together they form a unique fingerprint.

Cite this