A Novel Hybrid Strategy for Detecting COD in Surface Water

Abstract

The prediction of chemical oxygen demand (COD) by ultraviolet–visible absorption spectrum is a common method. Many researchers use the absorbance at the characteristic wavelength to establish COD prediction models. However, selecting the characteristic wavelength is a problem. In this paper, the extreme values of absorption spectrum change rate, was proposed as a new characteristic parameter to determine the characteristic wavelengths. On this basis, a novel hybrid strategy for detecting COD in surface water was proposed. We first proposed to combine the first derivative method with the permutation entropy method (FDPE) to determine the characteristic wavelengths. Then we used partial least square (PLS) to establish a COD prediction model. Experimental results demonstrated the linear correlation coefficient (R2) of the FDPE_PLS was above 0.99 without turbidity interference. Secondly, a dual-wavelength method (DWM) was proposed to determine the turbidity values. The DWM used slopes of absorbance values at 400 nm and 600 nm to predict the turbidity values. Compared with the single-wavelength method, the DWM improves the measurement accuracy of turbidity. Finally, a new turbidity compensation method was proposed to compensate for the interference in the first derivative spectrum. After compensation, FDPE_PLS can predict COD concentrations accurately, whose R2 was 0.99.

Keywords

• COD
• first derivative
• permutation entropy
• partial least squares
• turbidity compensation