Correlation analysis and modeling of water quality indexes for swine breeding wastewater

According to the difference of treatment process about swine breeding sewage, the treatment methods are divided into ecological treatment, industrial treatment and centralized treatment. The components of sewage treated by industrial treatment are extremely complex, there will be a great risk of environmental pollution if the sewage is directly discharged into the natural water body. It's very important to monitor sewage quality. The monitoring methods commonly used in swine breeding sewage mainly include laboratory monitoring and automatic monitoring. The laboratory monitoring is traditional, which has the advantage of high data accuracy and the disadvantages of low efficiency and poor timeliness, the sewage indexes can be detected fast but costly using automatic monitoring method. To find a monitoring scheme that combined the advantages of laboratory monitoring method and automatic monitoring method, took the sewage from a large-scale pig farm as the research object, the change characteristics, correlation of seven main indexes of sewage quality and regression modeling of two main indexes were studied. The seven indeices were respectively ammonia nitrogen, total phosphorus, total nitrogen, chemical oxygen demand, the potential of hydrogen, dissolved oxygen and electrical conductivity. Through the detection and correlation analysis of 30 random samples from different seasons and climatic conditions, it was found that ammonia nitrogen, total nitrogen and electrical conductivity had similar variation trends and strong correlation each other, the correlation coefficient of ammonia nitrogen and total nitrogen was 0.772, and that of ammonia nitrogen and electrical conductivity was 0.775, the correlation coefficient of total nitrogen and electrical conductivity was 0.920. Based on the results of correlation analysis, many types of monadic regressive and multivariate regression models for ammonia nitrogen and total nitrogen were established respectively, the relatively optimal "polynomial regression model" (model I) for ammonia nitrogen and the "comprehensive model" (model V) for total nitrogen were determined by comparing the coefficient of determination, residual sum of squares and the mean square regression of each model. The verification results based on 10 sets of data showed that the estimated values of these two models were closest to the measured values, the coefficients of determination of model I and model V were 0.855 and 0.953 respectively. Therefore, these two models could be used to evaluate the concentration of ammonia nitrogen and total nitrogen in swine breeding sewage. The existing studies shown that the data obtained by laboratory monitoring and automatic monitoring had the same change law although the value was different, which meant that there was a good linear relationship between them, hence a linear regression model based on the automatic monitoring data could be established to achieve the monitoring of water quality indexes accurately and rapidly. Based on this conclusion and the above two models, the feasibility of an efficient and low-cost automatic monitoring scheme for swine breeding wastewater quality was analyzed in this study. The indexes involved in the solution included electrical conductivity, the potential of hydrogen, ammonia nitrogen, total phosphorus, total nitrogen, and chemical oxygen demand, the total nitrogen that was difficult and expensive to detect automatically does not require to detect directly, the concentration of which could be calculated by the value of ammonia nitrogen and electrical conductivity according to model V, the concentration of ammonia nitrogen with relatively low difficulty and cost could be obtained by the value of electrical conductivity according to model I, the detection of electrical conductivity and potential of hydrogen was more convenient and the cost was lower, the data of total phosphorus and chemical oxygen demand would be obtained by linear regression model based on automatic monitoring data. Compared with the existing monitoring methods, the number of indexes that needed to be detected directly in this scheme would be significantly reduced, which would make the overall difficulty and the cost of monitoring decreasing, and the monitoring efficiency improved. Consequently, these two models could provide an important theoretical basis for the establishment of an efficient and low-cost automatic monitoring scheme for swine breeding sewage.