Modeling Of Soft Measurement For Anaerobic Digestion Based On LS-SVM

Anaerobic biological technology can recovery energy in the removal of organic pollutants, which provides a practical approach to address energy and environmental issues. However, anaerobic digestion is a complex,nonlinear, biochemical process. In particular, methanogenic bacteria are very sensitive to changing surroundings, so reasonable monitoring and control of anaerobic digestion process are needed in order to establish stable operation on a high level of organic loading. In practice,both the monitoring and control of anaerobic digestion have proved to be quite difficult tasks. One reason for that is the lack of reliable on-line measurement sensors, capable of monitoring some of its important variables and variation of some parameters of the process through time. Another reason is the nonlinear nature of anaerobic digestion, which makes it difficult to construct a model that completely describes the process, therefore it is difficult to build models that are able to predict the effects of short-timescale events on anaerobic digester and thereby control the process. Traditionally, we can develop more advanced equipment to solve the first problem and construct classical mathematical model, when severe simplifications of the process representation were assumed, to solve the second problem.Recently, soft-sensing technology based on inferential control provides new idea to solve problems mentioned above. Soft sensor method support vector machine(SVM) has more rigorous theoretical and mathematical foundation, good generalization performance and no local minimum problem which makes it a perfect method to solve small sample, non-linear and high dimension problems. In this paper, soft-sensing models based on least square support vector machine were constructed to predict effluent VFA and COD of anaerobic digestion after discussing the basic theory of soft-sensing technology. And the main conclusions are as follows:Granular sludge can accelerate the startup of anaerobic reactor, but environmental changes can cause the changes in types of anaerobic fermentation. Shock loads on anaerobic fermentation system for a short time will affect the stability of the system, and the hydraulic shock caused the greatest influence on the system, followed by concentration shock and bicarbonate buffer absent shock. All of the shock loads may make the p H value and methane density degrade, on the contrary they will increase the ORP value, carbon dioxide density, the effluent COD and VFA of the system, but vary in their response speed and variation law. Principal component analysis can be used to analyze the relationship between multiple variables and reduce the dimension of input variables effectively, thus will reduce the complexity of modeling significantly. The results demonstrate that a good forecasting performance was achieved through the steady LS-SVM model under steady state for predicting effluent VFA, the maximum relative error, mean absolute percentage error(MAPE),root mean square error(RMSE) and correlation coefficient(R) value of which were 4.72%,1.61%, 1.08 and 0.9996, respectively.Compared the steady LSSVM model, the non steady LS-SVM model had bigger RMSE(15.83%) and MAPE(15.45), but there was a good predicting performance, and R value was 0.9984.When training, the performance of the steady model based on LS-SVM for predicting effluent COD is acceptable, and the maximum relative error, mean absolute percentage error(MAPE) and root mean square error(RMSE) of which were 11.45%,0.79% and 3.08, respectively. However, the performance drops when predicting. Even though, correlation coefficient(R) value of 0.9752 was achieved, which means this model can predict the variation of effluent COD in general.The results of dynamic LS-SVM models for predicting effluent VFA and COD under concentration, hydraulic and bicarbonate buffer absent shocks showed that they have a good forecasting performance, all of whose correlation coefficient(R) value are greater than 0.99.Among these three shocks, the dynamic LS-SVM model under bicarbonate buffer absent shock achieved the optimal performance, and followed by dynamic model under hydraulic shock.Some exploratory and pioneering research work on the application of soft measurement technology in anaerobic fermentation were done in this paper, which may give reference to improve the monitoring level of anaerobic fermentation...