Diseño e implementación de un estimador de estados híbrido como respaldo para un sistema Scada en una planta de tratamiento de agua residual

Abstract

In this project, hybrid state space models and finite automaton were used for the design and simulation of a hybrid state estimator, which has been proposed as support for the electronic instrumentation system designed by the company Hidromecánica Ingeniería for the Universidad Agustiniana at its headquarters. For the above, the treatment plant was modeled based on first order differential equations, which allowed a dynamic approach of the plant and a subsequent modeling of its general operation by means of finite automatons, which allowed observing the discrete behavior of the plant. The design of the solution was divided into two parts; in the first part, a system was implemented in LabVIEW software that allows the simulation of each stage with its respective finite automaton, in order to know the operation of the plant in some specific states, which are responsible for generating the outputs that control the process. In the second part, the design of the dynamic model was carried out, for this purpose the plant was divided into seven stages that correspond to the design phases proposed by the company, in each stage data such as the filling time of each tank, height and flow rates were investigated, to generate from these characteristics a nonlinear dynamic model that describes the system, this model contemplates other characteristics such as the opening of the valves, and the nonlinearity presented by the variation of the emptying speed of each tank. Although the dynamic behavior of the system obeys a nonlinear model, in this work linear models are used around operating points defined by the presence of HHL (Homogenizer High Level), HLL (Homogenizer Low Level) and HLL (Homogenizer Low Level) level switch type sensors. This is proposed because it is expected to obtain better results in the linear models around the operating points than through a single model that covers a continuous range of points; thus, the computational cost is reduced and the plant identification process is simplified. Based on the above, the nonlinear model and its approximation were implemented by means of linear models at the operating points to verify its dynamic operation. For validation purposes, the model was subjected to different perturbations, obtaining as a result a linear relationship between the error and the magnitude of the perturbation; this is interesting and of future work since it would allow predicting the possible measurements of the linear model if the output of the dynamic model and the magnitude of the perturbation are known. After the realization of the linearized dynamic model in Matlab, each stage was implemented with its respective equations in LabVIEW software, to observe the joint behavior of the discrete event model and the dynamic model; the results obtained from this implementation agree with previous real data measured with the PLC-based automation system, with which the real plant currently has, allowing to give validity to the model, the identification and implementation performed.