Korean scientists have developed a groundbreaking deep-learning alternative model to optimise aluminium frames in glass-glass bifacial photovoltaic panels (PVs).
{alcircleadd}
Addressing the escalating challenges posed by the growing size and weight of PV modules, the researchers highlighted a critical issue: the heightened susceptibility of aluminium frames to deflection. Deflection refers to the lateral deformation experienced by a structural element under load, a concern amplified by the increasing dimensions of modern PV panels.
Aluminium frames play a crucial role in providing structural support to solar panels. They ensure they maintain their shape and integrity by preventing bending, warping, or sagging caused by their weight or external factors.
The innovative deep-learning hybrid model promises to revolutionise the design and performance of aluminium frames, offering a robust solution to mitigate deflection risks. This breakthrough underscores South Korea’s leadership in renewable energy technology and signals a significant stride towards more efficient and durable solar power systems worldwide.
The scientists clarified, “If the module undergoes deformation due to the load, it may lead to cracking or delamination of ribbons inside the module, increasing dead cells.”
“Bifacial modules are relatively heavier compared to monofacial modules. Therefore, when subject to increased surface area, there is a higher likelihood of frame deformation due to self-weight, leading to potential module deflection.”
Bifacial modules generate solar power from both the front and back sides of the panel, unlike traditional monofacial panels, which utilise only the front side with an opaque backsheet.
The academics began by outlining five key design factors for the frame, encompassing features such as ridges, grooves, and a hollow section. They defined three specific levels in millimetres for each factor. Subsequently, they conducted 243 experiments to assess deflection using finite element analysis (FEA), a computational technique pivotal in solving engineering and mathematical physics problems.
It was created as a surrogate module due to the significant computational demands of FEA.
The scientists detailed, “Deep neural network (DNN)-based surrogate modelling involves finding an approximate function that minimises the loss function for the given data. We applied Bayesian optimisation, deriving hyperparameters that minimise the loss function.”
Bayesian optimisation is a systematic approach to optimising global functions without assuming specific functional forms. It excels when sampling is costly, and the objective function is obscure but can be accessed through sampling.
After conducting 243 experiments, researchers employed innovative methods to enhance their results. They then trained and tested a novel model using a 9:1 ratio. This model, a Finite Element (FE) surrogate, proved impressively accurate compared to real Finite Element Analysis (FEA). The training set showed an average mean absolute percentage error (MAPE) of 0.0017 and a coefficient of determination (R2) of 0.9972, while the test set exhibited values of 0.0020 and 0.9962, respectively.
Encouraged by these findings, the researchers optimised the frame of a large 585 W bifacial PV module containing 78 M10 cells. The commercially available aluminium frame weighed 3.2 kg. This frame exhibited a deflection of approximately 12.3 mm with a displacement of around 2.8 mm in the opposite direction.
The goal of optimising the novel model is to design a frame that minimises deflection while keeping weight as low as possible. Increasing the thickness of the frame can reduce deflection, but it also tends to raise production costs.
The scientists said, “The factor values that minimise deflection and weight were a = 1.5176 mm, b = 13.7105 mm, c = 1.5012 mm, d = 2.9898mm, e = 4.3123. At this point, the deflection was 11.1 mm, and the weight was 3.6 kg.”
Generating the 1 million datasets took 0.957 seconds (s), obtaining the predicted deflection and weight values for these datasets took approximately 72.014 s, and finding the optimal values from these predictions took about 0.264 s. In contrast, using traditional FEA to obtain deflection and weight for a single case took about 4,800 s.”
Despite increasing the frame’s weight by approximately 12.8 per cent, their innovative method resulted in a 9.6 per cent reduction in deflection.
Note: Most of the information gathered from PV Magazine
Solar power applications are one of the fast-growing end-user applications of aluminium extrusions. Over the past decade, solar power has gone from an emerging, niche technology to becoming one of the mainstream electricity-generating sectors worldwide. Aluminium extrusions find application in frames and mounting systems in the solar power industry. To know more about aluminium’s usage in the solar sector, get access to AL Circle’s industry-focused report, “The World of ALuminium Extrusions – Industry Forecast to 2030”.
This post was originally published on 3rd party site mentioned in the title of this site
