Fast detection of ammonia nitrogen concentration in biogas fluid using near-infrared spectroscopy combined with wavelength selection by coronavirus herd immunity optimizer

Abstract

Listen

Translate article

Ammonia nitrogen, as an essential intermediate product during the process of anaerobic fermentation, is an important indicator for analyzing the nitrogen nutrient level, the specific gravity of different nitrogen in the substrate, and the extent of organic matter breakdown in the fermentation system. To achieve rapid detection of ammonia nitrogen concentration in biogas fluid during anaerobic fermentation, a partial least squares regression correction model of ammonia nitrogen concentration was constructed by integrating the near-infrared transmission spectroscopy fusion data of different optical path cuvettes with the wavelength selection of the binary coronavirus herd immunity optimizer (BCHIO). The effectiveness of BCHIO wavelength selection was proved by comparing the best and average modeling performance with classic intelligent wavelength selection methods, such as genetic algorithm, simulated annealing algorithm, and binary particle swarm optimization algorithm. To obtain a smaller amount of modeling wavelength variables with high correlation, the strategy of taking repeatedly selected wavelength variables in the results of the multiple runs as feature wavelengths was proposed, and 74 feature wavelengths were selected to construct the ammonia nitrogen concentration correction model. For the validation set, the determination coefficient of the quantitative model was 0.9961, the root mean square error was 12.4901 mg/L, and the average relative error was 4.2210 %, which could satisfy the requirements of rapid detection of ammonia nitrogen concentration in biogas fluid during anaerobic fermentation. By wavelength selection of ammonia nitrogen selected by BCHIO, the wavelength variables involved in the process of modeling are significantly reduced, the variable dimension and model complexity are effectively reduced, and the detection accuracy and predictive capability of the regression model are improved, which offers theoretical backing for the online detection of ammonia nitrogen concentration in biogas fluid by near-infrared spectroscopy.