Metasurface-Based Solar Absorption Prediction System Using Artificial Intelligence

Table 5

Comparison of the proposed method with recent literature.

Refs.	Materials	Methods	Limit	% of total mean absorption	Incidence angle (degree)	Resonance thickness (μm)	Accuracy (%)	Execution time (sec)	R^2
[21]	Graphene	LSTM	0.2–1.4 μm	92	0 to 70	0.1–0.8	85	32	0.98
[22]	SiO2	Regression models	0.2–1.0 μm	95	20 to 60	0.8	91	35	0.96
[23]	Titanium and gallium arsenide metal		500–4000 nm	99.69	0 to 50	—	86	26	0.97
[24]	Graphene	CNN-regression	0.2–0.4 μm	91	0 to 80	0.2–1	90	20	0.999
[25]	Gold multipattern swastika (DLMP) based SiO2	General regression neural network	0.1–3 μm	95	0 to 60	0.5–1	91	30	0.92
[26]		CNN	5.7 μm	97	—	—	—	17	0.86
[27]	MgF2 substrate	Polynomial regression models	0.2–1.2 μm	95.5	10 to 80	0.6–1.8	92.2	21	0.97
[28]	Ge2Sb2Te5 (GST) substrate	KNN	0.2–1.5 μm	92	—	—	90.2	38	0.9
This work	WTa-SiO2	PC-AE and GE-deep AlexNet	0.1–2.5 μm	98	0 to 20	0.6, 0.8, and 1.0	99.8	15	0.999