Soft Measurement Modeling Of Effluent Water Parameters Based On Hierarchically Neural Network For Wastewater Treatment

Effective methods must be proposed in key parameters measurement ofwastewater treatment process in order to avoid water pollution. As the wastewatertreatment is a complex mechanisms system which is strong coupling, time-delay andnonlinear, it is hard to measure the key parameters by using common instrument. Softmeasurement model must be established to solve the problem.Soft measurement model which is based on neural network is widely used forwastewater treatment process recently. The theory of neural network is very perfect,especially the feedforward neural network. The shortcomings of previous neuralnetworks are the results are not completely accurate and few key parameters aremeasured. So a hierarchically neural network (HNN) is proposed for the keyparameters measurement of wastewater treatment process. The new model canmeasure several parameters and has higher accuracy. At the same time BP learningalgorithm has low precision and easily falls into local minimum. So improved hybridParticle Swarm Optimization (IHPSO) is proposed which is for the new neuralnetwork. The main work of this paper is as follows:(1) The soft measurement model of wastewater treatment process is introduced.Establishing soft measurement model is the key of the soft measurement technique.Soft measurement model which is proposed in this paper has four parts, datacollection, data preprocessing, principal component analysis of data, neural networkmodel establishing. And the four parts are present in detail. Soft measurement modelcan be exactly established according to these four steps.(2) Hierarchically neural network (HNN) is established. Chemical oxygendemand (COD), biochemical oxygen demand (BOD), total nitrogen (TN) and totalphosphorus (TP) are four key parameters in wastewater treatment process. It isimportant to measure them. The hierarchically neural network is proposed based onthe relationship among the four key parameters. Then using the data of wastewatertreatment tests the performance of the new neural network. The results show that thenew neural network can be used in the measurement of wastewater treatment process.(3) An improved hybrid Particle Swarm Optimization (IHPSO) is proposed. Theresults of the hierarchically neural network (HNN) based on BP algorithm are notideal. So improved hybrid Particle Swarm Optimization (IHPSO) is proposed which is based on particle swarm optimization and combines crossover and mutation operators.It has quicker search speed and higher efficiency compared to BP algorithm. Thefunction performance tests show that the new algorithm can jump out of localoptimum. It enhances the global search capability of the particles. Meanwhile thesimulation results show that the soft measurement model which is based on improvedhierarchically neural network for wastewater treatment process has higher accuracy.