MODELING AND SIMULATION OF A TANK SYSTEM USING PREDICTIVE CONTROLLERS

Resumo

The study and application of PID controllers -Proportional, Integral and Derivative- represents
a great advance in productivity to the automation, being this one the most used for control of dynamic
systems. However, despite the robustness and ease of use of PID, we have more efficient options for
nonlinear processes or with dominant delay time. Among them, it should be emphasized that studies
are developed on predictive controllers based on model or MPC-Model Predictive Control-. The present
work aims to simulate a tank system with the objective of studying and evaluating the applicability of
MPC implemented in PLC - programmable logic controller. The PLC was chosen as one of the main
industrial automation tools, showing robustness to applications in adverse environmental conditions and
is widespread in the sector. Allied to the PLC, a supervisory software was developed that is responsible
for receiving information via the MODBUS protocol and storing it in a database. Such automation tools
as PLC and supervisory software stand out, for their ease, simplicity and robustness. The information
obtained by the process sensors will be used to construct the dynamic model of the system. Among
the steps, it was initiated by the instrumentation of the system, construction of the sensor, calibration
and connection to the PLC. The logical controller, in addition to sending information for presentation
and storage, will be the tool used to interpret the process signals and then execute the predictive control
algorithms. Thus, MPC presents dependency on the model for its construction and performs iterative
optimizations based on constraints that can be imposed on the cost function. Therefore, it is observed
that the construction of its algorithm is more complex when compared to the PID. It then becomes an
advantage to develop process simulations and use for MPC controller tuning. The construction of the
simulated model will allow to adjust parameters of the predictive controller and the cost function of the
optimization algorithm and after we can check that MPC was able to run in PLC and with results based
on control performance index.