SAMSE 2018 IOP Conf. Series: Materials Science and Engineering 490 (2019) 062027
IOP Publishing doi:10.1088/1757-899X/490/6/062027
sample individual of the test sample set containing 25 sample individuals, and the predicted value of the model was different from the predicted value of the PCA-PSO-LSSVM model without the OCS strategy integrated, which was generally reflected as the deviation tends to be small, thus achieving dynamic adjustment and optimization of the model. Tab. 1 Model performances comparison on the testing data set Methods
MAXE /mg·L -1
MAXRE /%
MAE /mg·L -1
MRE /%
RMSE /mg·L -1
STD /mg·L -1
SVM
54.39 51.09 31.09
7.82 7.53 6.26
18.62 17.37 12.31
2.58 2.40 1.80
22.96 21.57 15.23
13.71 13.05
PCA-PSO-LSSVM
OCS-PCA-PSO-LSSVM
9.15
To visually compare the prediction performance of the above three model methods, the experimental values and predicted values of COD of 25 sample individuals in the test sample set are plotted in Fig.5. Through observation of the figure, it can be seen that compared with the PCA-PSO-LSSVM and SVM model methods, the COD results on each sample individual predicted with the OCS-PCA-PSO-LSSVM model method are more closely to their respective experimental values, thereby indicating that the OCS- The PCA-PSO-LSSVM model method has better generalization prediction ability and stronger dynamic stability.
1000 1100 1200 1300
Analysis Value SVM PCA-PSO-LSSVM OCS-PCA-PSO-LSSVM
400 500 600 700 800 900
0
5
10
15
20
25
Observation Number
Fig 5. Prediction results of COD on the generalization data set
4. Conclusions Made in China 2025 clearly pointed out "taking the deep integration of informatization and industrialization as the main line." Based on the safe and healthy management and high-efficiency production requirements in the anaerobic treatment process of papermaking wastewater, this paper focuses on the study of the soft-sensor prediction and dynamic optimization of the model based on the data-driving effluent COD as the water quality indicator, to promote the transformation of the paper industry from extensive development to sustainable development, from the end treatment to the resource utilization, promoting the intelligent management and control of the production process, the main conclusions are as follows: 1) In order to adapt to the structure of anaerobic reactor and the multivariable, nonlinear, time-varying features of the parameters, as well as the special complexity of papermaking wastewater process and the uncertainty of production behavior, the soft-sensor method integrating the modern data analysis technology and intelligent regression model have been developed and designed, which not
8
Made with FlippingBook Digital Publishing Software