Multiobjective Model-predictive Control Based On Self-organizing Fuzzy Neural Network For Wastewater Treatment Process

Wastewater treatment processes(WWTPs) are typical complex industrial processes with the characteristics of highly nonlinear, interrelated, uncertainty and time-varying. The passively accepted flow and composition of influent,the complex of biological and physical reactor, the difficulty in parametric obtaining and the unclear of the object model, make it difficulty in controlling sewage treatment processes and reaching water quality standards. Moreover, the WWTPs are serious energy consuming processes. Therefore, the researches focus on the key issues of smooth operation, water quality, energy-saving, and exploration of intelligent control techniques. It is great significance to achieve effective control of sewage treatment processes.Model-predictive control(MPC) teconiques are applied to complex industrial processes. However, due to the highy nonlinear of the WWTPs, it is difficult to establish an accurate model of the object. Model-predictive control is generally difficult to satisfy the needs of the control performance in WWTPs due to the shortcoming of dissatisfied adaptive tracking a dynamic object.To solve the difficulty in establishing the model for WWTPs, this paper proposes a nonlinear model-predicitve control technique(NMPC), consisting of self-organizing fuzzy neural network prediction(SOFNN) and nonlinear controller. With the NMPC, it achieves a precise control of dissolved oxygen in WWTPs. It has the advantage of precise control and strong adaptive capability.For the smooth operation and lower energy consumption, a novel nonlinear multiobjective model-predictive control method(NMMPC) is proposed to achieve the stable control of dissolved oxygen, nitrate and other important control variables. The purposes of smooth operation, water quality standards, energy-saving are ensured.The main research and innovation are as follows:1. The research of prediction model in WWTPs. A prediction model based on SOFNN is proposed to establish accurate model of WWTPs with accurate predictive value and reliable future tracking trajectories. The SOA-SOFNN model can determine the network size and its parameters simultaneously in two stages. Firstly, all parameters of SOFNN are adjusted by an efficient second order algorithm(SOA) which incorporates an adaptive learning rate strategy into the learning process to take the advantages of fast convergence and powerful search ability. Secondly, the structure of SOFNN can be self-designed using the relative importance index of each rule during the learning process. The fuzzy rules of SOA-SOFNN are generated or pruned systematically to reduce the computational complexity and improve the generalization. All are designed to increase the capability in identifying the characteristic of WWTPs. Also, the convergence of the proposed SOA-SOFNN has been analyzed to confirm the new method is much computationally efficient. It provide reliable prediction values for sewage treatment process.2. The research of nonlinear model-predictive control in WWTPs. A nonlinear model-predictive control based on SOFNN(SOA-SOFNN-MPC) is proposed due to the necessity for dissolved oxygen control, lack of control means and lack of adaptive capability. The models have excellent adaptive capability to track the trajectory, maintain the prediction accuracy. Moreover, the stability of SOA-SOFNN-MPC is analyzed. The SOA-SOFNN-MPC is applied to the Benchmark Simulation Model(BSM1) to maintain the dissolved oxygen. It not only improves the control performance to avoid the abnormal in actived sludge processs, but also achieves the goals of improving water quality and energy-saving.3. The research of the nonlinear multiobjective model-predictive control in WWTPs. Due to the multivariable, multiparametric, interrelated characteristics, and a plurality of control loops, the nonlinear multiobjective model-predictive control scheme, consisting of SOA-SOFNNs is proposed to avoid the control problems caused by coupling. Thus the better water quality and energy-saving will be reached. The control scheme also uses SOA-SOFNNs to identify objects. The WWTPs are effectively decoupled by a plurality of models and well controlled by the multiple objectives controller via multiparametric multiobjective linear programing(mp-moLP). Similarly, the stability of the scheme is proved to keep the reliable of the control performance. The NMMPC is also applied to the BSM1, and the results reveal that the control scheme can effectively track the dissolved oxygen and nitrate. The performance of WWTPs is notablely improved.