| With the rapid development of the economy,the demand for water quality of drinking water is also increasing.How to guarantee good water quality and reduce the cost of water making has become a hot topic in water treatment industry.As a key link in water treatment,coagulation has the characteristics of nonlinear,large time delay and multi disturbance.Especially the micro-vortex coagulation process has more complex hydraulic conditions than the traditional coagulation process.In engineering applications,the problem of pesticide waste and poor water quality is often caused by inaccurate dosage of coagulants.In this paper,the mathematical control model of micro-vortex coagulant dosage process is constructed by function fitting.Moreover,the bionic algorithm of BP neural network and genetic algorithm is introduced into the process of micro-vortex coagulation.Whereafter,the micro-vortex coagulant control model based on BP neural network and the micro-vortex coagulant control model based on BP neural optimized by GA are constructed.The results and conclusions are as follows.(1)Determination of turbidity control targets and acquisition of modeling data1)The experiment of micro-vortex coagulation water purification is carried out.When the flow is 6.00 m3/h,8.00 m3/h and 10.00 m3/h,and the dosage range is 16.00?40.00 mg/L,the effluent turbidity decreases first and then increases with the increase of dosage.And when the dosage is 30.00 mg/L,the effluent turbidity reaches the lowest.At this point,it is the best dosage.In addition,in the range of 24.00?36.00 mg/L of coagulant dosage,the effluent turbidity is relatively stable,basically under 1.00 NTU,and this turbidity is determined as the control target of the dosage control model.2)In the premise of the effluent turbidity is less than or equal to 1.00 NTU,the dosage of coagulant is within the range of 24.00?36.00 mg/L,and the range of influent flow is 5.50?10.00 m3/h,including the conditions of 5.50,6.00,6.50,7.00,7.50,8.00,8.50,9.00,9.50,10.00 m3/h to obtain the flow,pH value,influent turbidity and effluent turbidity under the dosage of the corresponding coagulant.Thereby,the experiment provides the modeling parameters and data base for the research on the control model of micro-vortex coagulant.(2)Study on the mathematical model of micro-vortex coagulant dosage processThe MATLAB function fitting toolbox was used to fit the five indexes of raw water flow,pH value,influent turbidity,effluent turbidity and the dosage of coagulant.In the condition of effluent turbidity is less than or equal to 1.00 NTU,the two order linear function fitting showed that there was an obvious mathematical relationship among the five indicators and coagulant dosage,and the mathematical control model of micro-vortex coagulation dosing was obtained through non-linear function fitting.y=-23.3469+4.8436 x1+3.7803 x2-0.7377 x3+4.4275 x4-0.0878 x1x2-0.04716 x2x3+0.03687 x2x4-0.1784 x21-0.0486 x22+0.098 x23-5.8705 x24 where y is coagulant dosage,mg/L;x1 is raw flow,m3/h;x2 is influent turbidity,NTU;x3 is pH value;x4 is effluent turbidity,NTU.After analysis,it is found that the prediction error of coagulant dosage of the mathematical model is ±2.00 mg/L,and the error of individual point is abrupt to 3.00 mg/L,which illustrating that the error is large.(3)Study on micro-vortex coagulant dosage process control model based on BP neural networkOn the basis of the five indexes of the original water flow,the pH value,the influent turbidity,the effluent turbidity and the corresponding coagulant dosage,the number of nodes in the single hidden layer of the BP neural network is determined by using the empirical formula,and through MATLAB,the micro-vortex coagulant dosage process control system based on BP neural network is constructed.And the error analysis shows that the prediction error fluctuate within the range of ± 1.00 mg/L,indicating that the prediction error of BP neural network is smaller than the constructed mathematical model.(4)Study on micro-vortex coagulant dosage process control model based on BP neural network optimized by genetic algorithmThe genetic algorithm is applied to optimize the weights and thresholds of BP neural network.And based on genetic algorithm and BP neural network,the control system of micro-vortex coagulant dosage process is established.The simulation shows that the prediction error of the optimized network is in the range of ±0.50 mg/L,and there is no point where the error varies greatly.Thus it can be seen that after adding genetic algorithms,not only the prediction accuracy of the BP neural network is greatly improved,but the stability of the network prediction results is also improved.Therefore,the optimized BP neural network has better performance than the unoptimized BP neural network.(5)Test verificationThe mathematical control model,BP neural network control model and BP neural network control model optimized by GA is applied to practical application to verify their effect.In the 30 samples obtained from the experiment,7 effluent turbidity samples of the mathematical model exceeded the control target of effluent turbidity(<1.00 NTU),accounting for the overall 23.3%;only 2 effluent turbidity samples of the BP control model exceeded the control target of effluent turbidity,accounting for the overall 6.7%;all the samples of the BP control model optimized by GA is below the control target of effluent turbidity,it shows that there are no samples beyond control.Therefore,the control effect of BP neural network control model optimized by GA is better than the former two,which has a good control effect on the process of micro-vortex coagulation. |
PDF Full Text Request