Prediction Of Wastewater Treatment Based On Genetic Algorithm Optimization BP Artiifcial Neural Network

In the wastewater treatment process， the activated sludge method is a more advancedand mature process. Because of the complex mechanism， time-variant and delay of thewhole process， the wastewater treatment process is hard to develop accurate mathematicalmodel. It is a kind of typical of the complex industrial processes. Because a part of key waterquality parameters can’t be monitoring online， it is hard to guide the operation of thewastewater treatment system timely and effectively. Aiming at the above， the wastewaterquality monitoring technology has become urgent a problem to be solved in thisindustry.ANN has the advantages of nonlinear approximate characteristics、parallel distributearchitecture、better error acceptance、 adaptive learning and inductive ability， it is aneffective method for multi-factors， inaccurate，and fuzzy information processing. Geneticalgorithm overcome the shortcomings of ANN algorithm. IN this paper， on the basis ofthorough analyzing past achievement，the author studied a wastewater parameters softsensor method on the basis of genetic-ANN algorithm. The main content of this paper is asfollows:At first， know the wastewater process parameters influenced by degradable organics，Nitrification and denitrification and phosphorus removal process response， and analyse thewastewater process control parameters and the water parameters. And establish the activatedsludge wastewater quality parameters prediction model based on BP-ANN.Secondly， Because BP has slow convergence speed and get into the minimum in aeasily， randomly choice of initial weights and threshold，this particle propose a BP-ANNsoft sensor model based on genetic algorithm to optimize，combining with advantages ofthe genetic algorithm and BP-ANN， which apply to optimize the weights threshold of BPneural network and a better search space instead of the randomly choices of general initialweights and thresholds， then trains and learns the network to convergent in this space，search out the optimal solutions or the approximate optimal solutions.Thirdly， With the measured data of a wastewater treatment plant for the trainingsamples， establish two water quality prediction model applied to aeration tanks COD whichcan’t be on-line measured in the wastewater treatment process: one based on BP-ANN andsecond on GA-BP.， then compares and analysis the measurement results of the two methods.Comparing with BP neural network the results of the simulation show that the optimized algorithm has the advantage of fast convergence speed and higher forecastaccuracy The wastewater parameters soft-sensing method based GA-BP not only contributeto real-time control in the wastewater processing，but also has a positive effect in the othercomplex process control system.