Soft Sensor Modeling For Effluent BOD In Wastewater Treatment Process Based On Dynamic Time-delay Estimation

Wastewater treatment is an important means to improve the efficiency of recycling of water resources.As the main indicator to measure the effect of wastewater treatment,the effluent biochemical oxygen demand(BOD_e)is the important basis for the optimization and control of wastewater treatment process.However,BOD cannot be obtained in real time,which makes the wastewater treatment process unable to be adjusted and controlled in time,resulting in increased costs and substandard discharge of wastewater.Therefore,this thesis focuses on the research on soft sensor modeling of BOD_e in the wastewater treatment process.The main research contents of this thesis are as follows:(1)In the wastewater treatment process,the source of the data is complex and the water quality parameters are coupled with each other seriously,which makes it difficult to eliminate redundant information among water quality parameters.Therefore,this paper proposes a forward variable selection method based on k-nearest neighbor mutual information.It selects correlated variable based on the criterion of maximizing the forward cumulative mutual information value of input variables and judges whether the added variable is redundant variables by calculating the redundant mutual information value between each new added variable and the subset of selected variables.Thus,it can eliminate redundant variables and obtain the optimal subset of auxiliary input variables.A public dataset case and a wastewater treatment process case show that the proposed method can effectively eliminate redundant information in parameters and achieve optimal selection of auxiliary input variables.(2)Aiming at dynamical characteristic of the time-delay among variables,a strategy based on static delimitation-dynamic update(SD-DU)is proposed in this paper to estimate the dynamic time-delay among variables.It obtains the static time-delay and determines the dynamic time-delay variation range based on the fuzzy curve analysis method.Then,the improved sliding time window method is applied to capture the dynamical time-delay characteristic among variables.Afterwards,an evaluation index based on the information entropy theory is proposed to verify the validity of the dynamic time-delay estimation results.Finally,the effectiveness of the proposed method is verified by a public dataset case,and it is applied to the wastewater treatment process,which effectively obtains the dynamic time-delays and realizes the time registration among water quality parameters.(3)Variables with different time-delays have different importance to the output variables.To fully extract the correlation between input variables and output variables,a novel soft sensor modeling method based on the weighted relevance vector machine is proposed in this paper.It assigns a proper weight to every time-delayed variable according to the proportion of the number of time windows,and then establishes the weighted relevance vector machine model with dynamic time-delay estimation(DTDE2-WRVM).Finally,the DTDE2-WRVM model is adopted for predicting the BOD_e.The results show that the proposed model can make full use of the information of different time-delayed variables and achieve a higher prediction accuracy.