Nonlinear Predictive Control Of Coagulant Dosage

Coagulation process is an important part of potable water production, coagulation technique is applied to remove suspended particulates and harmful substances in raw water to achieve the purpose of purification. Process of treatment is a complicated physical-chemical process, therefore its control need to deal with nonlinear, time delay,random problems, etc.. Two projects are taken as background, which are Scientific and technological projects in Heilongjiang Province:"Research on Wastewater Treatment and Reuse Integration Technique of Small Towns in Northern Cold Areas"(Grant No. GB05C20202) and Cross-cutting Project:"coagulation process model Automatic Coagulant dosage System". Based on these two projects, forecasting quality of raw water, strategy for coagulation dosage control and coagulation dosage monitoring device are studied in this dissertation.Accurate mathematical models are significant foundations of coagulation dosage system analysis and design. However, it is difficult to build theoretical model since the process of coagulation is rather complex. In this dissertation, linear model and nonlinear model is established respectively during coagulation process by means of experimental methods. In the linear modeling, the coagulation process model is decomposed into two part, that is, the certain part and the random part, each of which is established separately. The model of the certain part is established by experimental methods, while the random part of coagulation model is fitted with ARIMA model. In the study of nonlinear modeling, Hammerstein model is adopted as coagulation process model and the parameter in this process is identified by Quantum-behaved Partical Swarm Optimization (QPSO).Raw water quality affects the subsequent process, and it also affects the quality of treated water. Implementing raw water quality prediction, and the results of forecast is provided to the control system. It not only improve dosing accuracy, but mastercontrol present situation and its trend of development, to analyze reasons and potential risks of accidents and to provide basic data and means for estimation in terms of water quality and water environment. Due to the characteristics of raw water turbidity such as nonlinear, non-stationary, several turbidity forecasting methods are proposed in this dissertation, which is based on linear and nonlinear methods. Firstly, the random series stabilized is forecasted by time series method. AR method on the basis of autocorrelation method are given to the water quality forecasting. Secondly, phase space reconstruction is applied in RBF neural network forecasting, so that the contents of data is expanded and the accuracy of forecasting is elevated. Thirdly, the intrinsic mode function (IMFs) are adaptively extracted via empirical mode decomposition (EMD) from a time series of turbidity according to the intrinsic characteristic time scales. Tendencies of these IMFs are forecasted with SVMs respectively, and then these forecasting results are compounded to turbidity series. The last step is to make a comparison between these results by way of simulation analysis.Predictive control algorithm is adapted for the characteristics object of nonlinear, variability, time-delay and uncertainty.The DMC algorithms based on non-parameter model and the GPC algorithms based on linear parameter model is successfully applied to control industrial process. In this dissertation, the predictive control algorithm of coagulant dosage process is studied. Meanwhile, the traditional DMC algorithms are improved: improved DMC algorithm combined with model simplified and predictive error correction algorithm is applied to decrease computational complexity and solve the model mismatch problem; An predictive control algorithm based on RBF neural network is researched, solve control rate by golden section method and the error correction algorithm is given. Reserch nonlinear predictive control Hammerstein model based, control rate solving algorithm are presented. At last the simulation results show the effectiveness of several algorithms.The last but not the least link is the study of engineering application. Scheme of coagulation monitoring system based on LonWorks techniques is given.The compound control method has been researched, control strategy which is composed of feedforward control and feedback control in control project. Through hardware and software design, and applied to simulative coagulation devices.