Research Of Predictive Control In Water Treatment Plant Based On Neuron Network

In the process of surface water treatment, coagulation is an indispensable course. Under certain conditions, the amount of coagulant dosage directly determines the coagulation effect, adding coagulant accurately can effectively reduce the burden of filtering, disinfection equipment. However, due to raw water turbidity, PH, water flow, and other factors, coagulation dosage is the most difficult course. Because it is facing a complex physical and chemical process, is a highly nonlinear system, thus it is hard to establish accurate mathematical model. At the same time, for water treatment process from adding dosage, after sedimentation, filtration, more than one hour is needed to achieve stability.At present, there are two main methods of dosage control system used in domestic water plants; they are one-closed-loop control and forward compensation control. For coagulant administration which is a nonlinear and large time delay system, both the traditional control method are difficult to response immediately, when the water quality parameters make sudden changes, that is why their effectiveness has limitations. By analyzing the status, a new method of predictive control based on RBF network with forward compensation and feedback structure was proposed in this paper.Firstly, two one-step predicting models were established by BP and RBF network respectively, and then the multi-step models on RBF network were also constructed and studied. By comparison, the one-step model based on RBF was selected to be the internal model. Then, a closed-loop control system was established based on predictive deviation control algorithms, meanwhile the Golden Section algorithm is used to solve control rate. Next, the proof of system stability and zeros steady-state error characteristic was given in this paper. Finally, system performance was simulated by Matlab/Simulink; the results proved superiority and feasibility of this control method.Because the spread of this paper is to establish automatic system of waterworks, LonWorks technology is selected to design the dosage node. Products developed by LonWorks have the characteristics of openness and interoperability, and the products which are acceptable for remote monitoring are digital and net-based.According to the demand of management and control of coagulant dosage, this paper designed the functions of hardware and determined the final scheme. According to module design rules, this paper gave LonWorks control module and data collection module (including analog input and output circuit) respectively. At the same time, this paper provided analog input and output program, control rate solving program and debugger program using Neuron C. Finally, the actual operational data showed that, compared with traditional artificial control system, the out-let water turbidity controlled by this method remained stable, the necessary coagulant consumption is reduced, as a result, the cost of production is cut down to some extent, and the control effect is obvious.