Wastewater Treatment System Predicting And Optimization Control Based On Immune Algorithm

Biological wastewater treatment process is a non-linear strongly coupling time-varying system with complex mechanism. The introduction of intelligent algorithms is a strong impetus for development on wastewater treatment research. Immune algorithm is one of intelligent algorithms that solve various complex problems based on biological immune mechanism. The immune algorithm has been widely applied in fault diagnosis, optimization, and intelligent control, et al. More in-depth theory and application studies of immune algorithm should be carried out. The applications of immune algorithm in wastewater treatment system are far from well-enough. Based on previous work, we further study and improve the immune algorithm. The improved immune algorithm in the thesis is applied in anomaly data detection, effluent water quality prediction, model parameters estimation, decoupling control and optimal control of wastewater treatment system.The main contents of the thesis are outlined as follows.1. In order to obtain wastewater treatment data with high quality, an improved negative selection algorithm (INSA) is proposed to detect and eliminate the anomaly data. Variable-sized detectors are employed to improve the performance of INSA. At the same time, the minimum radius of detectors is limited. The effectiveness of INSA is verified through anomaly detection in time series data. Employing the INSA to detect the wastewater treatment data, simulation results show that the INSA can improve the detection precision of wastewater anomaly data.2. In order to improve effluent water quality prediction precision of wastewater treatment system, the main factors which have influence on the effluent quality are analyzed according to the process flow and the characteristics of influent water quantity of a wastewater treatment plant. Wastewater treatment system is a multi-input multi-output system. But the traditional support vector regression machine (SVRM) algorithms are only used for single-output systems. If several SVRM models are constructed for multi-input multi-output systems, it will increase the complexity of the algorithm and the precision is poor for the correlation of output variables. In order to solve prediction problem of multi-output system, a method of multi-output least squares support vector regression machine (LS-SVRM) based on immune optimization is proposed. The multi-output LS-SVRM is used to predict effluent quality, using the immune algorithm to optimize the parameters of multi-output LS-SVRM. Simulation results show that the proposed method has a better prediction precision for wastewater treatment system.3. Aiming at the parameters uncertainty problem of activated sludge model, a parameters estimation method based on improved immune algorithm is proposed. The model parameters used in different environments can not take the same parameter values. A periodically varying mutation operator is designed based on immune mechanism and periodical evolution of organism to improve the search ability of the improved immune algorithm. If only the affinity is taken as an immune selection evaluation criterion, low affinity antibodies will be overly inhibited. An improved immune selection operator is designed by introducing the antibody concentration to the affinity as the evaluating index. The convergence of the improved immune algorithm is analyzed based on the Markov chain. In order to test the effectiveness of the algorithm, it is applied to solve the function optimization problems. The improved immune algorithm is applied in parameters estimation of activated sludge model. Experimental results indicate that the improved immune algorithm is of high estimation precision.4. Aiming at the characteristic of the strong coupling between ammonia nitrogen and nitrate nitrogen, and the conventional PID control method is difficult to achieve satisfactory control performance, a PID neural network (PIDNN) approach is employed to achieve decoupling control for ammonia nitrogen and nitrate nitrogen, where dissolved oxygen concentration and internal circulation flow are regarded as the control inputs. PIDNN connection weights are easy to fall into local optimum. In order to solve this problem, the immune algorithm is proposed to optimize the connection weights of PIDNN. The stability condition of the control system is analyzed. In order to improve the algorithm's ability to find optimal parameters of connection weights, the immune algorithm adopts the improved immune selection operator which introducing the antibody concentration to the affinity as the evaluating index. The stoichiometric coefficients and kinetic parameters adopt activated sludge model parameter estimation values. Simulation results show that the proposed method has a better decoupling capabilities and control quality for biological wastewater treatment system.5. Aiming at the optimal control problem of wastewater treatment process operation cost, a novel immune algorithm is proposed to calculate the optimal value of operation cost, which takes the two most important control parameters, sludge wastage and dissolved oxygen as control variables, regards total substrate discharge and effluent water quality as restriction factors and operation cost of residual sludge treatment, sludge return and aeration as performance index. In order to improve the search efficiency, a novel scale-variable hybrid mutation operator is designed, which introduces the mutative scale method in the Gauss mutation and Cauchy mutation. The convergence, stability and time complexity of the algorithm are then analyzed, and an optimal control of operation cost is performed with the algorithm for wastewater treatment. Compared with the basic immune algorithm and genetic algorithm, experimental results indicate that the novel immune algorithm is of high search efficiency and low mean and variance of wastewater treatment operation cost.