Non-dispersive infrared SF6 sensor with temperature compensation using ISSA_BP neural network

Abstract

Sulfur hexafluoride (SF6) is a widely used gas in the power industry, and its concentration monitoring is of great importance in power systems. In this study, a non-dispersive infrared (NDIR) gas sensor was developed for monitoring SF6 with a response time (t90) of 3–4 s. A dual-channel structure was designed to minimize the influence of other factors such as light sources. However, the sensor’s output signal is susceptible to temperature drift and non-linear characteristics. To address these challenges, a new compensation algorithm of improved sparrow search algorithm based on the backpropagation neural network (ISSA_BP) is presented to improve the temperature dependence and to correct the nonlinear errors of the sensors. The ISSA_BP algorithm extracts features from sensor-collected signals and predicts the concentration of SF6 gas by the BP neural network. The levy flight strategy and sine-cosine factor are employed within the sparrow search algorithm (SSA) to improve the efficiency and accuracy of the method. The experimental results show that the proposed model outperforms the state-of-the-art approaches, achieving an average relative error of 0.02 within the range of −10°C to 45°C and gas concentrations from 0 ppm to 5000 ppm, which is 3% lower than the industry standard. Compared with traditional approaches, the measurement accuracy is improved approximately by 0.02–0.03, confirming that the proposed algorithm has a high reference value for temperature compensation and nonlinear calibration, which is applicable in the production of infrared sensors for industrial settings.