Forecasting For Sludge Volume Of Biological Wastewater Treatment Index Process

Sludge Volume Index?SVI?is an important index widely used to analyze the sedimentation characteristics of sludge.The manual detection process of this index has problems such as tedious steps,time-consuming,high cost,and unable to detect in real time.Real-time reliable and accurate prediction of SVI have practical application value for process monitoring,fault diagnosis and operation of wastewater treatment processes.In this paper,the morphological characteristics of sludge microorganisms and process parameters such as oxygen and p H values in the aeration tank were used as features to conduct SVI modeling and application research respectively.On the one hand,image segmentation and morphological feature extraction are used to establish an SVI prediction model for the morphological characteristics of floc in sludge.On the other hand,the distributed SVI forecasting model is used to solve the sewage data of multiple water quality monitoring stations on multiple sewage treatment lines.Aiming at the diversity of multi-waterline operating conditions,this method constructs a model suitable for the operating conditions of each waterline,and the model is more robust than the single-waterline model.The main research work of this thesis includes the following points:1.Image processing and analysis of activated sludge microscopic images is an effective tool for predicting sludge settling capacity and early detection of filamentous bulking.It is an effective way to build a soft-sense model of sludge volume index based on the morphological characteristics of microorganisms.The segmentation performance of flocs and filamentous bacteria in the microscopic images directly affects the effects of image processing and analysis.Segmentation of the phase contrast microscopic images is a challenging problem because of the weak greyscale distinction between the flocs and filaments,as well as the artifacts of halos and shadows.In this work,we proposed an automatic flocs and filaments segmentation method for the phase contrast microscopic image using a U-Net deep learning structure with data augmentation.A loss function combining binary cross entropy function and dice coefficient is proposed to improve segmentation accuracy and sensitivity with unbalanced foreground and background samples.The performance of the segmentation algorithm is evaluated using accuracy,precision,recall,F-measure,Intersection over Union?Io U?metrics.Experiments on the lab-scale activated sludge process have been carried out to verify the proposed image segmentation method.Our proposed u-net models with the combined loss function give better results compared to the u-net model with Io U of 0.74,F-measure of 0.82 and precision of 0.81.2.The improved U-Net segmentation model was used to segment the flocs in phase contrast microscopic images collected from laboratory-scale activated sludge.Based on the segmentation results,the morphological characteristics of flocs in activated sludge were further studied Extraction method and selection of morphological features to generate data samples.Use data samples to build a network model that predicts SVI from morphological features.In order to enhance the performance of the SVI prediction model based on the stochastic configuration algorithm,improve the model's resistance to overfitting and speed up the training of the model,look for suitable nodes in the classic stochastic configuration network algorithm,add nodes and calculate the output weight model residuals to determine whether the There are three parts of the stopping conditions,and three improved measures of the algorithm are given,including L2 regularized stochastic configuration network?L2SCN?,early stopping stochastic configuration network?Val SCN?,and fast training convergence stochastic configuration network?Fast SCN?.Simulation experiments results show the effectiveness of the proposed algorithms.3.Aiming at the distributed SVI modeling and forecasting of the sewage data dispersedly stored in the multi-line monitoring workstation,a construction method of distributed regularization stochastic configuration network?ADMM-SCN-Enet?was proposed.The elastic network penalty term is added to the distributed algorithm to prevent overfitting.Specifically,the distributed regularized stochastic configuration network builds on a set of local SCNs model to solve large global optimization problems with regularization penalties in a collaborative manner via the alternating direction multiplier method?ADMM?.The local SCN models are built under the supervision mechanism with inequality constraint and assumed that the input weights of hidden layer for all local models are consistent.ADMM optimization problem with ???₁-norm?the lasso?,???₂-norm?ridge regression?and elastic-net penalties are employed to calculate alternately output weights and the Lagrange multipliers of the distributed model through the decomposition-coordination procedure.A comprehensive study on five benchmark datasets and the ball mill experimental data have been carried out to verify the proposed method.The experiment results show that the proposed distributed regularized stochastic configuration network has relative advantages in terms of accuracy and stability compared with the distributed random vector functional link network.Therefore,the distributed modeling and measurement of the activated sludge volume index of multiple activated sludge treatment lines in the region are realized,and the generalization ability and robustness of the activated sludge SVI measurement model are enhanced.Experimental results show the effectiveness,generalization and robustness of the distributed algorithm.4.Designed and developed the activated sludge volume index prediction software,and implemented the method proposed in this paper.The software's main features include sludge microbial phase contrast microscopic image segmentation,morphological feature extraction and data set generation,improved random configuration network algorithm based on morphological features of SVI modeling and forecasting,and the use of multiple wastewater treatment line monitoring workstations.The data builds a SVI regularized distributed random configuration network and performs SVI forecasting.The software has the characteristics of multithreading,program robustness,modular design and so on.The test results of the software show the effectiveness of the developed SVI forecasting system.