Temperature Compensation Method for Gas Sensors Based on Lightweight Mixture of Experts (MoEs)

To address the nonlinear drift challenge of gas sensors under wide temperature ranges in the Industrial Internet of Things (IoT), this article proposes a Chaotic Dynamic Harris Hawk Optimization-based Mixture of Experts (CDHHO-MoEs) temperature compensation model. First, gas sensors are fabricated, and an MoE architecture is designed by incorporating their temperature-dependent characteristics, where a temperature-aware attention mechanism is introduced to enhance thermal sensitivity. A multistrategy hybrid chaotic Harris Hawk optimization algorithm is further developed to improve the parameter optimization capability of the MoE model. Finally, benchmark tests on an STM32H743 microcontroller demonstrate that the proposed model achieves ultra-fast inference speed and high compensation accuracy, showing significant potential for IoT applications.