Water Online the Magazine gives Water & Wastewater Engineers and end-users a venue to find project solutions and source valuable product information. We aim to educate the engineering and operations community on important issues and trends.
Issue link: http://wateronline.epubxp.com/i/303843
Tutorial wateronline.com ■ Water Online The Magazine 26 SCADA allows access to the database through programs developed in C++. This SCADA's feature allowed the programs developed for implementing the models to perform the reading and writing in tags configured for the model's input and output variables. Prediction Of The Settled Water Turbidity Figure 3 on page 25 shows the signs of turbidity and prediction in different real-time base, i.e., to each measurement the real current turbidity value is shown in the red curve and in accordance to the modeling; the blue curve shows the turbidity for the next 1.5 hours. Identification Of The Operator's Behavior The program developed to identify the operator's behavior uses resources already developed for the program in predict- ing settled water turbidity. The amendment made was basically in the class that implements the model being tested, in which we used the following matrices obtained from training for the calculation of the output, i.e., the value of dosage: Matrix P — matrix 3 x (4 +1) that represents the linear models; Array S — matrix 3 x 4 with the standard deviations for the Gaussian for each entry; Matrix Xm — matrix 3 x 4 with the average values for the Gaussian for each entry. This program reads the configuration file of the following data: • Tags of input variables — turbidity, color, flow, pH; • Tags of program output — color and dosage generated by the model; • Neuro-fuzzy model — averages and standard deviations of the membership functions, in addition to the linear models used for defuzzification, expressed by the matrix mentioned above. When the program is connected to the SCADA, it performs the following functions: • Reads variables that were collected manually, which are color and type of coagulant; • Reads and stores the input variables to each second, understanding that before storing the variable it undergoes a process of filtering; • Computes the dosage based on current parameters and model neuro-fuzzy; • Writes the dosage calculated and the color in the SCADA; • Shows onscreen all current process variables and makes a comparison of the program between the dosage of the model and the actual dosage performed by the operator. The results obtained in the field are presented in Figure 4. The graph shows the daily record history for the dosage values effectively practiced by the operator and those calculated by the model. Although we tried to use different topologies of ANN, the results of the field were not satisfactorily observing that variations in input variables had little significant weight in output variable. The model output was primarily influenced by its own delayed output. These facts raise several hypotheses, such as: • We may not have considered all the variables of entry necessary for the proper representation of the process. • The operator's action may have been the cause of low correlations obtained, making it difficult to obtain a representative model of the dynamics of the plant, based on the data collected. For future work we suggest the construction of a new database considering the system in open-loop control and segregating the operator action from the dynamics of the process. The modeling of the operator's behavior presented results considered satisfactory. This implemented model already has practical application and may be used to replace most of manual interventions. However, because it is a model in open- loop control, it still requires a feedback signal, in this case represented by the prediction of the settled water turbidity. Selma Parreira Capanema is an electrical engineer who develops studies in the field of computational intelligence with application to real-world problems in water supply systems. She currently works for Copasa, located in Brazil. Figure 4: Identification of the operator's behavior 2 4 _ V E R T _ 0 5 1 4 C l e a n w a t e r _ C o p a s a _ D G . i n d d 3 24_VERT_0514 Cleanwater_Copasa_DG.indd 3 4 / 2 1 / 2 0 1 4 3 : 0 5 : 5 6 P M 4/21/2014 3:05:56 PM