다변량 회귀분석에 의한 일차침전지 유출수 수질 예측

Abstract

다변량 회귀분석에 의한 일차침전지 유출수 수질예측 시뮬레이션에 대한 결론은 다음과 같다. 1. 표면반응 분석을 이용한 다변량 회귀분석으로 유입수의 유량, BOD, SS, TN, TP, 1차슬러지량, 하루전 SS를 이용한 일차침전지 유출수수질을 재현할 수 있는 회귀 모델식을 만들었다. 2. 유출수 BOD, COD, SS, TN, TP의 RMSE(root mean square error)는 각각 3.81, 2.74, 4.7, 4, 0.2 이었다. 3. 2006년 자료로 추정된 회귀모델은 2007년 자료로 검정한 결과 실측값을 잘 재현할 수 있었다.

The release of polluted water by an inadequate sewage treatment cause a serious environmental problems and disease in animals and plants. Recently, strict environmental controls and Water Quality Standards have been adopted and consequently ICA(Instrumentation, Control and Automation) techniques have been developed rapidly in order to manage and control the sewage treatment process. As a result, various predictive models have been proposed by using simulation and mathematical models. Many models which are developed for the sewage treatment process are focused on a predicting technique which is to detect an irregular situation of the process and prevent a problem from occurring in the process. Recently, various Neural Networks-based models which might be efficient in pattern recognition and prediction problems have been used in previous researches. However, there some unsolved issues in neural network models. The input vectors in neural networks should be normalized between 0 and 1 if the activation function used is the standard sigmoid. Another problem is the noise of input vectors could effect to the predictive accuracy. Because of these problems, it would not be appropriate to use previous popular models for the process design and the process control optimization of the sewage treatment. In this research, the Primary Sedimentation Tank Effluent Estimation model, which can be substitute for the ASMs proposed by IWA task group and adaptable to an integrated model of the sewage treatment, has been proposed in order to overcome the problems of previous studies. The main results in this study ara as follows; 1. A multiple regression model in which independent variables are inflow, BOD, SS, TN, TP, and first sludge has been developed to predict the primary sedimentation tank effluent. 2. The RMSE(root mean square error) of BOD, COD, SS, TN and TP are 3.81, 2.74, 4.7, 4, 0.2 respectively. 3. The proposed regression model, which has been developed with data of year 2006, has been tested on the data of year 2007 and produced a good performance in terms of predictive accuracy.