Control systems are essential for the safety and navigation efficiency of ships. Various control methods have been applied. A variable-speed control device that adjusts the voltage power frequency according to the freshwater (FW) temperature output from the central cooling system is applied to the cooling seawater (SW) pump motor to save energy. A typical central cooling system is shown in Figure 1.

Figure 1: 
Diagram of the central cooling system

The application of variable speed pumps to save electrical energy in ship central cooling systems has been researched extensively [1]-[4]. The application of variable speed motors to SW pumps [5] and the importance of using an integrated approach to reliably evaluate central cooling system retrofit solutions for variable speed SW pumps to maximize the fuel efficiency of ships have also been studied [6].

A data-based proportional-integral-derivative (PID) process controller was designed, and a variable-speed SW pump and controller were applied to an actual oil tanker; it was confirmed that the fuel consumption of the diesel generator was lower, compared with that of a constant-speed SW pump [7]. In addition, a feed-forward control method was applied [8]. The feedback position of the PI controller was changed to control the speed of the SW pump, and energy savings were confirmed through a simulation [9].

However, prior researchers focused on saving energy by using two controllers for the SW pump and a three-way valve, which are parts of the central cooling system. Applying the two controllers to a central cooling system requires proper tuning to achieve a good control performance with energy savings.

Therefore, a linear quadratic (LQ) controller was proposed in this study. First, a mathematical model was derived from the actual central cooling system. Then, the proposed LQ controller, which integrated two controllers for the central cooling system as a multi-input single-output system, was developed. It was used to prove the effective regulation of the FW output temperature through efficient control of the SW pump rotational speed and effective control of the three-way valve opening against step disturbances through simulations.

A mathematical state–space model was required to develop the LQ controller. The modeling method was TY Jeon’s simulation model [9]. However, the developed mathematical state-space model applied two input variables: the electric voltage frequency of the SW pump and the opening of the three-way valve. This mathematical state–space model was linearized near the operating point. Table 1 lists the abbreviations used to derive the mathematical models.

If the optimal control vector for Equation (12) was equal to Equation (20),

a gain (K) that minimized the performance index(J) of Equation (21) was designed [18].

where Q is a positive-definite (or positive-semidefinite) Hermitian or real symmetric matrix, and R is a positive-definite Hermitian or real symmetric matrix. Matrices Q and R determine the relative importance of the error and expenditure of this energy, respectively.

In a central cooling system, the FW outlet temperatures of the three-way valve and HEX should be kept constant. Therefore, referring to Equation (11), the values of Matrix Q corresponding to T_ho and t_out were increased to 100 and 165, respectively, as shown in Equation (22). Matrix R was set as shown in Equation (23).

Equation (24) was obtained when Matrix K minimized the performance index J.

The control gain K was a 2 × 3 matrix. As shown in Figure 2, K₁ was the SW pump power frequency, and u₁ and K₂ were connected to the opening input u₂ of the three-way valve.

Figure 2: 
Diagram of the central cooling system

However, the developed mathematical model did not perfectly match the actual system. This implies that there is always an error. To eliminate this error, an integrator, which was 20% of K_{(2, 3)} was added, as shown in Figure 2.

The simulation test was conducted twice on a simulation model developed using MATLAB [9].

The first simulation was applied to the disturbance input. The FW and SW inlet temperatures were input as the step-up and step-down signals at 500s and 2000s, respectively, after the start of the simulation. The FW inlet temperature was input as 38°C, then 42°C, and finally 38°C, and the SW inlet temperature was input as 20°C, 25°C, and 20°C once again. After the step-up FW temperature was input, 25°C was maintained for 1500s.

For the second simulation, the disturbance input was applied in the same manner as in the first. However, the SW inlet temperature was step-down input from 24°C to 20°C, then step-up input to 24°C again.

Through the simulation test, variations in the FW outlet temperature, SW pump rotational speed, and opening degree of the three-way valve were monitored and recorded to confirm the control performance.

The simulation results for the step-up and down disturbances of the FW and SW inlet temperatures are graphically represented in Figure 3 (a). Figure 3 (b) shows the variation in the FW output temperature. The overshoot was less than ±0.02℃. The settling time within ±0.01 ℃ was approximately 38s and 90s. As shown in Figure 3 (c), the rotational speed of the SW pump reached 1420 rpm and then decreased to 1403 rpm after the input of the step-up disturbance. The three-way valve opened to a maximum of 0.95 after the step-up disturbance, and then approached 0.83, as shown in Figure 3 (d).

Figure 3: 
Simulation result with step-up and down disturbances

Figure 4 shows the simulation results against the step input as per (a). The difference was that a step-down input of the SW inlet temperature preceded the step-up input. The overshoot was within ±0.02℃ as per the graph (b). Figure 4 (c) shows that the SW rotational speed increased to 1147 rpm and then settled at 1135 rpm. The three-way valve opened up to 0.67 and was fixed at 0.63.

Figure 4: 
Simulation result with step-down and up disturbance of SW inlet temperature

The central cooling systemwas mathematically modeled to develop an LQ controller. The central cooling system modeled by the state–space equation was linearized near the operating point. Using a linearized model, a control gain K that minimized the performance index function J was obtained by assigning a state weight to the FW outlet temperatures of the HEX and three-way valve.

The developed LQ controller was applied to the simulation model to mitigate the step disturbance inputs of the FW and SW temperatures. The e of the LQ controller was validated through simulations. The FW outlet temperature was maintained at the desired value without significant overshoot, the SW pump rotational speed was efficiently controlled, and the three-way valve opening was effectively controlled. Thus, the good control performance of the integrated LQ controller developed in this study was established.