Hybrid System PSO-ANFIS for Optimization of The Water Level in The Tank

The speed and pressure of the water flow are determined by the height and volume of the water. The speed of the water flow in the actuator is determined by the use of this flow sensor system. A good tank-based water flow control model should be developed. At a certain point, the actuator stabilizes the rate of water production per minute. Therefore, it is necessary to develop an automatic and precise control technique. Many Artificial Intelligence (AI) methods are used in system optimization. Among them are Particle Swarm Optimization (PSO), Neural Network (NN), Fuzzy Inference System (FIS), and ANFIS. Adaptive Neuro Fuzzy Inference System (ANFIS) is a combination of NN and FIS. In this study, the PSO method was combined with ANFIS. This hybrid method produces better optimization compared to the previous method. The best water level control simulation results are PSO-ANFIS with an overshot of 0.572 pu, undershot of 0.563 pu, and flow output overshot of 0.008 pu, undershot of 0.009 pu.


Introduction
A linear model estimation approach was used to study the physical workings of the Water Level System (WLS).The purpose of WLS is to depict dynamic qualities near the equilibrium point [1].An identification method with measured input and output data is utilized to identify the dynamic nature of WLS.Technically, nonlinear time process control systems are frequently used in the development of water tank level instruments.WLS can be modeled as a global water system or partially as one [2].Fluid flow control systems in tanks are required for industrial processes and system improvement.The production system must provide a complete list of all procedures.To improve system performance, the water flow in the water tank volume is adjusted.
For the overall control system design, several models may result in various control strategies [3].A number of techniques for managing the water level have been investigated and developed, including employing an Firefly Algorithms (FA) and ANFIS controller [4] [5].The PSO-ANFIS Hybrid system had never been used in prior investigations.So, the optimum optimization technique is required.

Design Research
In this research there are 2 subsystem blocks in the water tank system, namely the water tank system and the valve system.The two systems act as joint interactions.Both systems act as a fluid flow to complete the entire interaction of the sorting section.The input system is technically influenced by a constant water flow rate, signal generator, and the maximum inflow of the tank [6].The water flow is channeled using a pump from the storage tank.The water flow rate is regulated using an actuator.Figure 1 shows a schematic of a wave tank system with 2 inflow tunnel systems  Figure 2 shows the designed water block system.This block informs that the Simulink block diagram is developed based on the combined subsystem.The uncontrolled simulink blocks interact in the main parts as presented in the valves and tank units.The Water Tank Sub System can be seen in Figure 3;

Conventional PID Model
This method sets the parameters Kp = 1, Ki =1, and Kd =1.So obtained after the system output is reached in a steady state [7][8] [9].The PID controller simulation can be seen in Figure 6; The PSO Algorithm Model is an algorithm that mimics the collective behavior of birds in search of food [10] [11].In this paper, optimization is carried out to find the value of the PID parameter so that the PID can produce the smallest overshot and undershot.The PSO parametersican be seen in Table 1[12] [13].Best (overall best position), velocity (speed) determines the direction of movement the position carried out in each iteration, inertial weights are used to control the impact of speed changes, acceleration coefficients (controlling the movement of one iteration can be determined independently [10].The best iteration results will be stored in the constant values of Kp_pso, Ki_pso, and Kd_pso then stored in the data workspace.

ANFIS
ANFIS is a modification and combination of the Fuzzy Inference System (FIS) mechanism with Newral Network (NN) [10][14].The fuzzy inference system used is the Tagaki-Sugeno-Kang (TSK) fuzzy model.This method provides convenience in computing.The input data for ANFIS is taken from PID-tuned PID training results [15].The ANFIS block diagram and ANFIS structure can be seen in the simulnk block Figure 7 and Figure 8.The steps for each layer of ANFIS are as follows; The first layer functions to convert crisp numbers into fuzzy but only involves each input; In the second layer, each input goes to the same layer to determine the firing strength; The third layer is the normalization calculation, by processing the re-weighting to get the total value; The fourth layer multiplies the input (x and y) to get the output in CRISP form; The fifth layer accumulates the results of the fourth layer.

RESULTS AND DISCUSSION
The simulation design is carried out with various controller models, namely PID controller, PID-Auto controller, PID-PSO controller, and hybrid PSO-ANFIS controller.In this study, all outputs from the design results of various Water Level Control Systems can be simulated using a Simulink diagram as shown in Figure 8 and Figure 9. Figure 9 demonstrates that the PSO-ANFIS model provided the smallest overshot and smallest undershot on the water level.This demonstrates that PSO-ANFIS is the ideal model for this study of water level.Figure 10 displays the zoomed figure.The PSO-ANFIS model provided the output flow's least overshot and undershot, as shown in Figure 10.This demonstrates that PSO-ANFIS is the ideal model for this study of water level.Figure 11 displays the zoomed figure.

Conclusion
From the results of the water level control simulation, the PSO-ANFIS model had the smallest overshot value of 0.572 pu, the smallest undershot value of 0.563 pu, and the output flow result had the smallest overshot value of 0.008 pu, the smallest undershot value of 0.009 pu.Thus, it may be said that PSO-ANFIS is the best controller model.

Figure. 1 .
Figure. 1. Surge tank system with 2 inflow tunnel systems Simulink Block uncontrolled diagram can be shown in Figure 2;

Figure. 3 .
Figure. 3. Design of Sub Sistem Water Tank Model Figure Block diagram of the sub-system on the valve system can be seen in Figure 4.

Figure. 4 .
Figure. 4. Design of The Valve Subsystem model Figure 3 depicts the Simulink block diagram for the water tank subsystem of the water tank system, and Figure 4 depicts the valve subsystem.The water entering and leaving these systems is controlled while it is flowing through a tunnel or pipe system.

Figure. 5 .
Figure. 5. Simulink the Block diagram PID Controller model The results of the training conducted by the PID in the case of this research the PID-PSO were entered in a simout which was then used as input (load data) to ANFIS.Particle Swarm Optimization (PSO) are three important components in PSO, namely Particles, cognitive components, and social components.There are two determinants of learning from particles, namely experience (cognitive learning) and combination learning (social learning).PSO algorithm development factor; swarm (number of particles in the population), Particles (individuals who have position and velocity), Personal best (is the current best position compared to the best solution proposed previously), Global

Figure. 9 .
Figure. 9. Results of Water Level (Zoom) Output Flow Results were simulated using a diagram as shown in Figure 10 and Figure 11.

3
The simulation results in Figures8 and 9display the overshot and undershot values from the results of controlling water levels using a variety of control strategies.The values of overshot and undershot from the results of the Output Flow are displayed in Figures10 and 11.Table2displays the overall outcomes.Tabel 2. Overshot and Undershot of Output Flow 7