Improving PV cleaning schedules without large datasets

Scientists in Algeria have developed a low-cost solution to optimize cleaning operations for all PV systems. The proposed approach works "effectively" without heavy data requirements, according to its creators.

Researchers at Algeria's Université de Ghardaia have developed a new method to enhance PV cleaning schedules that ensures independence from extensive datasets.

“Our method functions effectively without heavy data requirements,” the research's lead author, Charaf Abdelkarim Mosbah, told pv magazine. “It combines maximum power point tracking (MPPT) techniques, metaheuristic optimisation, and an intelligent cleaning score mechanism (ICSM).”

In the study “Smart and cost-effective optimisation of photovoltaic cleaning schedules,” published in Energy and Buildings, the research team explained that the method allows for “optimal” decision-making in both real-time and offline settings, thereby reducing unnecessary cleaning operations and improving overall energy yield. “What makes our approach stand out is its seamless integration with existing MPPT systems, enabling immediate deployment without further investment in hardware or data infrastructure,” Mosbah went on to say.

The scientists explained that cleaning operations require two fundamental steps – assessing cleanliness and making a decision – and noted that cleaning becomes “essential” only when the dust accumulates and persists on the module surfaces, with the analysis focusing on faults with constant or gradual characteristics.

The proposed method differentiates between two distinct MPPT operation modes: the exploration phase, when the algorithm actively seeks the MPP; and the exploitation phase, when the MPP has been reached. It also distinguishes between partial shading and uniform irradiance modes and defines the search space of each decision variable independently, which the scientists said is not the case with many other algorithms.

The new methodology uses ICSM to simplify dust accumulation detection. It initially evaluates if the PV system is operating normally or under partial shading, with the data being analyzed in real-time in conjunction with MPPT or evaluated later offline. “This feature allows the method to be applied during system downtime, enabling maintenance tasks, including cleaning schedule planning, to be efficiently managed without disrupting energy production,” the research group emphasized.

The novel approach also utilizes an intelligent counter to accumulate the results of the data analysis. “Our experimental results demonstrated a 98.4% accuracy in predicting optimal cleaning schedules, achieving superior performance compared to three other benchmark algorithms,” Mosbah stated.

The research team claims their new solution can be applied to all PV systems, including large-scale solar plants, with no additional costs.