Proposing a neuro-fuzzy controller to compensate for the effect of disturbances on the MED-TVC desalination system

Abstract

Due to climate changes and disturbances in the steam production of power plants, multi-effect desalination (MED) plants are facing two possible disturbances: pressure change in motive steam (MS) and temperature change in seawater. Therefore, a controller is needed to compensate for these disturbances. However, the dynamics of these systems are highly nonlinear. Hence, a Takagi-Sugeno-Kang Fuzzy Logic Controller (TSKFLC) is designed for a MED-TVC1 on an industrial scale with four effects, and a condenser. This novel model-free controller is designed by training a neuro-fuzzy network using an input-output dataset extracted from an experimentally validated mathematical model that considers the ejector malfunctions. Hence, the dynamics of the system is fully considered in the controller because it is trained by its own input-output dataset. The controller performance in the first effect brine level regulation as the output of the closed-loop system is investigated in facing changes in the motive steam pressure (MSP) and inlet seawater temperature (SWT) as the main disturbances to prevent the phenomenon of flooding or emptying and possible damages. The behaviors of the open-loop and closed-loop systems in the presence of these disturbances are investigated. Simulation results show that the proposed multi-input-single-output (MISO) controller is so fast in facing disturbances. According to the results, by increasing MSP to 25 bar, it can be seen that the level of brine in the first effect decreased sharply, and the whole space is filled with vapor, which this phenomenon is called emptying. On the other hand, by decreasing the pressure, the level of brine reached 47.65 cm which is fourth times higher than the steady mode, and it brings a new phenomenon called flooding. Moreover, simulation results show that MSP has a higher impact compared to SWT. Also, according to the results, the system needs only 400 s to reach a stable situation showing the fast performance of the neuro-fuzzy controller facing disturbances.