| Activated sludge wastewater treatment technology is considered as an important environ-mentally sustainable method for the degration of organic matter in wastewater. For the characteristics of the complex mechanisms, strong coupling, intrinsic nonlinearities and time varying in biological wastewater treatment process, the traditional control schemes hardly guarantee good control performances of the wastewater treatment system. In this thesis,several improved intelligent algorithms and control strategies have been developed, and applied in the key variable prediction, optimization and the control designs of biological wastewater treatment process.The main results of the thesis are summarized as follows:1. For the monitoring problem of key parameters with the unreliable measurement in the biological wastewater treatment system, a predictor for the effluent concentration based on an improved extreme learning machine has been developed. Firstly, combining the memetic evolutionary mechanisms of shuffled frog leaping algorithm(SFLA), an improved hybrid intelligent algorithm named DEPSO has been developed. The performance of DEPSO is veri?ed by simulations. Then, DEPSO is used to optimize the network hidden node parameters and an improved learning algorithm named DEPSO-ELM has been developed.The improving performance of DEPSO-ELM has been shown by simulations on regression applications of several real world datasets. Finally, DEPSO-ELM is verified effective in the prediction of ef?uent from biological wastewater treatment plant.2. For the setpoint determination problem of controlled variables in the wastewater treatment process, an optimization and control method based on an improved differential evolution algorithm has been studied. Firstly, combining with the evolutionary mechanisms of SFLA and enforcing local learning strategy, an improved DE algorithm has been developed.And its performance has been verified by simulations. Then, the operating cost and the effluent quality are combined into a single cost function for evaluating the plant performance.And on the simulation platform of BSM1 provided by the international water association, the setpoints of controlled variables are optimized by the proposed algorithm. Finally, the simulation results show that the optimized control system results in a decrease of operational costs, a drastic decrease of the time above discharge time and a signi?ant improvement of the effluent quality.3. Multi-criteria selection of optimum wastewater treatment plant control setpoints based on an improved DE algorithm has been studied. Firstly, an improved DE algorithm with self-adaptive control parameters and optimum differential vector enforcing learningmechanism has been developed. Simulations on benchmark functions have been carried to verify the proposed algorithm. Then, on the BSM1 simulation platform, the optimization and control have been studied, with the multi-criteria functions set as system operating cost, the effluent quality and the constraint set as microbiological risk. Simulation results show that the optimized wastewater treatment processes operate more efficiently than the compared scenarios. Multi-criteria method can overcome the weight selection randomness problem. The results with a Pareto distribution provide different optimal scenarios considering the three criteria for the operating personnel.4. For the control problem of dissolved oxygen(DO) concentration in biological wastewater treatment process, two controllers are developed. The first one is the neuron self adaptive PID controller with the supervised Hebb learning regulation, and the other is RBF neural net-work controller. The relationships between RBF NN control performances and the number of training data have been tested by simulations. And the tracking performance of the proposed controllers was verified by simulations on the simplified BSM1 platform.5. For the control problem of DO concentration, a neural nonlinear adaptive control de-sign technique has been developed. Firstly, based on a simplified DO concentration model,a desired nonlinear feedback controller has been developed. And radial basis function neural networks are used to approximate the uncertain desired controller. Then, it is rigorously proved that semiglobal uniform ultimate boundedness of all the closed-loop system signals is guaranteed by the Lyapunov method. Finally, simulation studies are performed to demonstrate the effectiveness of the proposed adaptive controller without an off-line training for NNs.6. Considering the external disturbance in biological wastewater treatment processes, a neural adaptive control design technique using a disturbance observer is developed. Firstly,based on a simplified DO concentration model, RBF NNs are used to approximate the un-certain dynamics of the wastewater treatment process. The effect of the unknown external disturbance and the NN approximation error are combined into a compounded disturbance,and estimated by a nonlinear disturbance observer. Then, rigorously proved by Lyapunov method, the adaptive NN control based on the disturbance observer can guarantee semiglobal uniform boundedness of the closed-loop system signals and the disturbance estimate error.Finally, simulation studies are performed to demonstrate the tracking performance and the robustness of the proposed controller. |
PDF Full Text Request