Study of NH3-N prediction based on nonstationary time series in a sewage River

Using the deterministic analysis of non-stationary time series, this paper focuses on the prediction of NH₃-N in a sewage River. Considering environmental hazards of indiscriminate discharge of sewage, the experiment was performed on a sewage river of selected area. NH₃-N was measured at selected seven sampling locations from Aug 2007 to Aug. 2008. We conducted a total of nine samples. During the experiment, the sampling time is basically the same and time intervals of sampling are also consistent, therefore the sampled data (seven samples) can be looked on as a sequence of time. According to the requirements of time series analysis, manipulate the data and test the stationarity of the series. By time series analysis, an exponential smoothing of NH₃-N is developed. As long as we know the actual value and predicted value of NH₃-N concentration in last period, we can predict the next NH₃-N concentration. Significant test of model parameters and adaptive testing of the model show that the exponential smoothing of NH₃-N was significant. Using the data from water quality testing department of the city to test the model, which indicates that the veracity of the model was 82.9%, which meets the precision requirement of the mode1.