Identifiability And Optimal Estimation To The Parameters Of EBPR Model Of Activated Sludge Processe

The parameter estimation of ASMs is a prerequisite for their application to simulation of a wastewater treatment plant. This work should be made based on the evaluation of identifiability of model parameters. In this paper, a new method based on differential algebra was proposed to research the structural identifiability of ASMs. Moreover, the identifiability of parameters of enhanced biological phosphorus removal （EBPR） model was studied. The denitrifying phosphorus removal performance of A2/O system was enhanced by adjusting the ratios of return mixed liquor. Finally, based on optimal experimental design theory, the parameter estimation of EBPR model was evaluated in practice.（1） The differential algebra algorithm proposed in this paper could be successfully applied to research the structural identifiability of different activated sludge sub-models. The structural identifiability of the EBPR sub-model was evaluated by the algorithm for the first time. It provided a new general method to research the structural identifiability of ASMs.（2） If only one measured variable was considered, none of the parameters of anaerobic sub-model was uniquely identified. However, by combining data of two measured variables, S_A and S _PO4, the stoichiometric coefficient Y _PO4 and the half-saturation coefficient KA were structural identifiable. Similarly, YPHA and KPS were identifiable when two measured variables, electron accepter ( S_O or S_NO3) and S _PO4, were combined for phosphorous uptake processes. The structural identifiability of parameters in the EBPR sub-model could be improved by increasing the measured variables. The reduction factorηNO3 was identifiable when combined data of aerobic process and anoxic process were assumed.（3） Some parameters of EBPR model are correlation, such as KPP, X_PAO and qPHA in anaerobic phosphorus release process, or KPHA, X_PAO and qPP in phosphorus uptake process. To be uniquely identifiable, one of the three parameters of the anaerobic process, KPP、X_PAO and qPHA, is needed to be determined by other ways. Likewise, if prior information on one of the parameters, KPHA, X_PAO and qPP, is known, all parameters of the phosphorous uptake process are identifiable.（4） By controlling operational conditions, the denitrifying phosphorus removal performance of A2/O system could be enhanced. The average removal rates of COD and NH4+-N are 85% and 95.6% respectively. The average removal rate of P is 82.9% when system is steady-state. The ratio of anoxic phosphorous uptake is increased from 27.4% to 65.7%. The average ratio achieves 62.6%.（5） Results of sludge characteristic experiments showed that, the anoxic and aerobic specific phosphorus uptake rates were 5.79 mgP /（gMLSS.h） and 9.29 mgP /（gMLSS.h） respectively. The ratio was 62.3%. Thereby, the ratio of D_PAOs that could uptake phosphorous under anoxic condition and _PAOs was 62.3%.（6） By applying roulette wheel selection, uniform crossover and basic bit mutation operator, the genetic algorithm was constructed. Assuming the value of K PP is 0.01 mg（XPP）/mg(X_PAO), the parameters of anaerobic phosphorus release process, Y _PO4, K A, q PHA and X _PAO, were estimated by GA. Bsed on the results of batch experiment, the optimal values and the confidence intervals in 95% level were as follows, Y _PO40.36±0.014, K A3.27±1.286, q PHA3.41±0.219 and X _PAO1560±80.36. The errors expressed as percentage of parameter value were Y _PO43.88%, K A39.34%, q PHA6.44% and X _PAO5.15%, respectively（7） After changing the value of KPP by increasing 10%, the four parameters, Y _PO4, K A, q PHA and X _PAO, were estimated again. Results showed that the errors of Y _PO4 and X _PAO increase 0.56% and 0.84%, whereas the errors of KA and q PHA reduce 0.85% and 0.14%. Thereby, the bias problem would not be appeared when the value of KPP was altered.（8） Based on the modelling and D-criterion, a procedure of optimal experimental desigen for parameter estimation （OED/PE） of anaerobic phosphorus release process was proposed. It was.The optimal values and the confidence intervals in 95% level of parameters were as follows, Y _PO40.35±0.017, K A3.88±0.414, q PHA3.35±0.273 and X _PAO1500±71.87. The errors of parameters expressed as percentage of parameter value were Y _PO44.86%, K A10.67%, q PHA8.15% and X _PAO4.79%, respectively.（9） Compared to results deduced from another experimental condition, the results from the optimal experimental condition are better. The errors of Y _PO4 and q PHA increase 0.98% and 1.71%, respectively. However, the error of K A and X _PAO reduce 28.67% and 0.36%, respectively. The overall error of parameter vector was improved. All the errors are under 20% and could be accepted.