Study On Soft Sensor Model Of Wastewater Measuring Based On SVM

To solve the problem that some parameters are difficult to be measured on-line in the process of wastewater disposal, this paper makes research on the soft measurement model, which can make it possible to get parameters ,on-line and make wastewater disposal automatically. This paper introduces the basic of the soft measurement model and data processing, and analyzes various soft measurement models, especially SVM(Support Vector Machines). Based on SVM, this paper presents a soft measurement model to measure the parameters which are measurable but costliness or unmeasured on-line, such as TN, and gives a method to preprocess input data of the model by PCA (Principal Component Analysis), RS (Rough Sets) and RS-PCA. It is more effective compared to the model based on he BPNN(Baek Propagation Neural Net), according to results of the simulation by Matlab.The paper mainly contains:1. The paper introduces soft sensor and wastewater treament ,studies some kinds of methods of founding soft measurement model, taking their scope of application and the peculiarity of the wastewater disposal plant into account. The paper analyzes ANN and SVM and summarizes the advantages and disadvantages of both methods emphatically.2. Because the forecast precision of the soft measurement model will be affected by the error and correlation of the input data. The paper devises the method to preprocess the input data using PCA, RS and RS-PCA. With PCA, it is able to reduce effect of error, eliminate linear correlation of data and reduce dimension of the input data. With RS, it is able to eliminate linear correlation of variables and reduce dimension of the input data. They can both simplify the model structure and improve the forecast effect. Furthermore, RS can reduce the number of input variables,and can not eliminate datas' linear correlation contrast with PCA. Using RS-PCA can combine the advantages of PCA and RS.3. The effectiveness of the model is verified with the field-measured data of a certain municipal wastewater disposal plant. Different methods (SVM, PCA-SVM, RS-SVM, RS-PCA-SVM, BP, PCA-BP, RS-BP, RS-PCA-BP) are presented to set up corresponding models for some variables that are difficultly and expensively measured, such as TN, TP and SVI. In the end, the best model is chosen as the final model that has the best effectiveness.4. Design the soft measurement system of wastewater disposal based on SunyPcc500E controller, using the soft measurement model of wastewater disposal as mentioned above.