Study On The Methods Of Water Quality Real-time Monitoring And Intelligentized Warning In Water Source

In recent years, the problem of the pollution of the water environment has a seriousimpact on the normal human life, and the formation of the water blooms is a typicalproblem in the pollution of the water environment. The water blooms formed, involvedin a lot of uncertain factors, so it is difficult to model and forecast the formation of thewater bloom with the mathematical model. In the study of the water bloom process inrecent years, many domestic and foreign scholars studied the methods of the water bloomsfrom the mechanism modeling and intelligent prediction, and achieved a certain amount ofresearch, makes the water blooms forecast has become one of the priorities of the waterresearch.For the water blooms prediction problem, this dissertation proposes a water bloomshort-term forecasting model respectively based on the Elman artificial neural network andLibsvm support vector machine. The new intelligent research methods are designed toprovide the comprehensive forecast of the water blooms and form relatively completewater bloom prediction system. Compared the accuracy of the two prediction models,applied the better fitting model algorithm to the Suzhou water source water bloomprediction system. Using the software of hybrid programming method to create waterbloom prediction platform and providing a more comprehensive reference in water qualitymonitoring and preventive treatment.First, according to the water quality data source provided from Suzhou water sources,determine the index system of the water bloom prediction. Based on the water bloommechanism and energy accumulation characteristics, define the indexes pH, oxygenconsumption, temperature, light intensity, turbidity, ammonia nitrogen, total phosphorus,total nitrogen, dissolved oxygen and Chl_a as a predictor of water blooms system to set thefoundation of the forecasting methods.Secondly, proposed the Elman feedback neural network water bloom short-termforecasting model, and establish initially the water bloom prediction model through themethod of trial and error in the hidden nodes of the model. Depending on the timeinterval for model training and prediction simulation, obtained the Elman network has agood performance in short-term prediction of the water bloom. Based on the time seriescharacteristics of the water blooms, propose the Libsvm support vector machine regressionalgorithm model as well. Select the same time interval prediction, draw the Libsvm model has a higher accuracy compared with the Elman water bloom prediction model. Atthe same time, use the fuzzy information granulation support vector machine regressionalgorithm to analyze the future trends and change intervals of ten days chlorophyll insample sets and achieved good results.Finally, apply the Libsvm short-term forecasting model to the water quality real-timemonitoring and intelligent warning system in Suzhou water source. Use the mixedprogramming technology of Matlab2011and Visual Studio2005to build the Suzhou watersource water bloom prediction system application platform and get a good applicationeffect.