Fig. 1. Decomposition of OH Raman stretching vibration peaks corresponding to different local hydrogen bonds of water molecules

Fig. 2. Optical layout for Raman spectral measurement of water with controllable temperature

Fig. 3. Flow chart of optimization process of spectral peak fitting using L-M algorithm

Fig. 4. Decomposed sub-peaks of 2800-3800 cm^-1 and residual between fitted and original spectra

Fig. 5. Comparison of Raman peaks of OH at different temperatures and electrical conductivities after spectral pretreatment.(a) Different temperatures; (b) different electrical conductivities

Fig. 6. Two-dimensional distribution diagram of spectral intensity at characteristic peak position as a function of temperature and electrical conductivity. (a) 3200 cm^-1; (b) 3490 cm^-1

Fig. 7. Normalized intensity of OH Raman subpeaks varying with temperature and electrical conductivity. (a) Varying with temperature; (b) varying with electrical conductivity

Fig. 8. Temperature and electric conductivity prediction results and related evaluation indicators of different quantitative algorithms on test set. (a) PLSR; (b) LASSO; (c) SVR; (d) LSTM+AM

Fig. 9. Structure diagram of Stacking ensemble learning model

Fig. 10. Prediction results and evaluation indicators of temperature and electric conductivity in test set using Stacking model