Adaptive temperature compensation for MoS₂ humidity sensor in complex environments using ISSA-BP neural network

High-precision humidity detection in complex environments is essential across various fields. In this study, a high-performance MoS₂ humidity sensor with a dynamic response time of less than 3 s was developed using molten salt-assisted chemical vapor deposition. To address the challenges posed by dynamic ambient temperature changes on sensor accuracy, an improved sparrow search algorithm-back propagation (ISSA-BP) neural network was constructed to mitigate temperature drift and correct nonlinear errors in the sensor output. The ISSA-BP neural network utilizes a global optimization strategy with adaptive learning, significantly enhancing accuracy and efficiency by optimizing the initialization and iterative update processes of traditional algorithms. Experimental results indicate that the proposed ISSA-BP achieves an average relative error of just 0.75% in humidity sensors across various environmental conditions, representing a 5.8-fold improvement in accuracy compared to traditional methods. Additionally, the algorithm demonstrated high robustness and accuracy across different environments, sensors, and datasets, confirming its applicability in complex and variable scenarios.