**2. Research Background**

Most energy conservation methods implemented in cooling systems for a single chilled water system apply frequency down-conversion control to the chilled water system according to the change in outdoor climate; alternatively, such methods adjust the parameter settings of the chilled water system on the basis of the minimum RT demand. However, few methods apply energy conservation control according to indoor heat demands. To meet the indoor thermal comfort demand, users adjust the operating temperature (*Top*) of an HVAC system through an FCU temperature control panel. This demand defines the required RT that the chilled water system must produce.

The present study focuses on an HVAC system optimization control strategy involving FCU temperature control. In this strategy, indoor air information is included in the analysis, and a user's heat demand volume for a chilled water system is determined. The determined heat demand volume is then converted to RT demand for the chilled water system. This strategy can be used to change the chilled water flow and cooling water flow to meet indoor heat demands while ensuring minimum power consumption. The main techniques involved in this strategy are as follows: algorithm for dynamic FCU temperature setting, method for indoor heat demand conversion, and optimization control strategy for energy consumption of single cooling machine of the chilled water system.

#### *2.1. Dynamic FCU Temperature Settings*

An FCU distributes cold air produced by a chilled water system to an indoor area. In such an area, the demand for cold air can be determined by observing a user's adjustments of FCU temperature settings. The Daily Discomfort Score (DDS) is developed to assess demand response (DR) and thermal comfort in a building. Baseline and preconditioning scenarios are used to demonstrate the e ffectiveness of the Daily Discomfort Score in evaluation thermal comfort and demand response HVAC system set-point control [23]. The ASHRAE Standard 55-1992 states that the comfort zone for summer conditions, air temperature to be between 23 ◦C and 26 ◦C and relative humidity between 20% and 60%. Surveys of human thermal response in South East Asia produce the Auliciems' equation for estimating thermal neutrality on the basis of the mean monthly dry bulb temperature [24].

In the proposed method, a regression model that includes outdoor temperature and indoor sensing temperature as variables is used to calculate the indoor HVAC temperature:

$$\text{Ta} = \infty \times (\text{Ti}) + \text{β} \times (\text{To}) + \text{c} \tag{1}$$

where *Ta* is the indoor HVAC temperature, *Ti* is the indoor sensing temperature, *To* is the outdoor temperature, α and β are regression coe fficients, and *c* is a constant.

In Equation (1), the range of the indoor HVAC temperature is not constrained; hence, the temperature may be excessively high when the indoor sensing temperature or outdoor temperature is markedly high. To maintain indoor temperature quality, the indoor HVAC temperature can be corrected using tan–1 and limited to the indoor comfortable range of 23–26 ◦C. The indoor comfort temperature (*Tc*) can be expressed as follows:

$$\text{Tc} = \text{[tax}^{-1}(\text{Ta} - \text{24.5})] + \text{24.5} \tag{2}$$

An experiment conducted in this study revealed that when *Top* was adjusted, the indoor environmental temperature normally required 10–20 min to approach *Top*. To maintain indoor temperature quality and meet the cooling air request, the proposed algorithm based on dynamic FCU temperature settings automatically sends an HVAC setting temperature (*Ts*) that is within the indoor comfort range every 30 min and calculates the temperature di fference between *Tc* and *Top* to determine the indoor thermal comfort (*ITC*), as presented in Equation (3):

$$ITC = Tc - Top\tag{3}$$

The derived *ITC* is then used to correct *Ts*to achieve indoor comfort. The conditions for determining *ITC'* are as follows:

$$\begin{array}{c} ITC' = \begin{cases} \ +1 \sim -1, \; \mathrm{Ts} = \; \mathrm{Ts} + [\! \Lambda T \times (\! \mathrm{C}^{\vee})\!] \\ \end{array} \\ \begin{array}{c} \mathrm{Others}, \; \mathrm{Ts} = \; \mathrm{Top} \end{array} \end{array} \tag{4}$$

where Δ *T* = *Tc* − *Ti* and *C*γ is the power law conversion coe fficient.

The cooling e fficiency of an HVAC system is subject to numerous factors. In general, *Top* is not equal to *Ti*; in particular, *Ti* is typically higher than *Top* in summer. To maintain indoor comfort, the proposed algorithm determines that users are satisfied with the current environmental temperature when the derived *ITC* value is within ±1 ◦C. Additionally, a gamma transformation function is adjusted through power law conversion to adjust *Ts*; thus, *Ti* approaches *Tc*. The purpose of power law conversion is to ensure a high output *Ts* when the input Δ *T* is high, which enables rapid indoor temperature response. When the input Δ *T* is low, a low *Ts* output is obtained to prevent excessive energy waste.

An *ITC* value that is not within ±1 ◦C signifies that users are not satisfied with the current environmental temperature. Accordingly, *Top* is adjusted using the FCU temperature control panel. In this situation, the FCU dynamic temperature setting algorithm uses *Top* as *Ts* for the HVAC system to meet the user-demanded temperature. Subsequently, the algorithm recalculates *Tc* and re-executes the *ITC* determination conditions and temperature control in 30 min.

#### *2.2. Indoor Heat Demand Conversion*

A conventional chilled water system estimates the required RT by relying solely on the di fference in temperature between chilled water entering and that exiting the system. However, such a system cannot determine whether the estimated RT can meet the heat demand of an area. Therefore, a measurement index is required to enable the chilled water system to adjust RT supplies. A conventional FCU provides only *Ts* and *Ti* information. By contrast, this study proposes a technique that integrates the difference between *Ts* and *Ti* with the FCU wind speed to obtain a comfort weight (W) for chilled water system control. Figure 1 illustrates the system framework.

**Figure 1.** Framework of system for deriving comfort weight.

This system is based on a multiple-input single-output (MISO) fuzzy controller, where the output is W. The MISO controller is described as follows:

1. The change in temperature settings (F1) is used to understand the required RT for a user, as presented in Equation (5). A negative F1 value (e.g., a change in temperature setting from 26 ◦C to 24 ◦C) indicates that the user feels hot, whereas a positive F1 value indicates that the user feels cold:

$$F\_1 = \text{Ts }(t) - \text{Ts }(t-1) \tag{5}$$

In Equation (5), t is the sampling time.

2. The target temperature error (F2) is the difference between Ti and Ts, and it represents indoor cooling level, as presented in Equation (6). A negative F2 value (e.g., an actual indoor temperature of 24 ◦C and set temperature of 26 ◦C) signifies overcooling, whereas a positive F2 value signifies undercooling. Additionally, wind speed is used to determine whether the FCU is turned on or off:

$$\mathbf{F}\_2 = \mathbf{T}\mathbf{i} - \mathbf{T}\mathbf{s} \tag{6}$$

In Equation (6), N represents a negative value and P represents a positive value.

Figure 2 shows the input and output fuzzy sets of the fuzzy controller. Table 1 presents the fuzzy rules applied by the system for weight derivation. A *W* value of 2 means insufficient RT from the chilled water system, indicating a warm indoor environment. By contrast, a *W* value of −2 means excessive RT from the chilled water system, indicating a cold indoor environment.

**Table 1.** Fuzzy rule.


In the proposed method, the minimum inference engine and center of gravity defuzzification are used to calculate the comfort weight per time unit *W(t)*, which is presented as follows:

$$\mathcal{W}(t) = \frac{\int\_{\mathcal{T}} B't dt}{\int\_{\mathcal{T}} B'(t) dt} \tag{7}$$

where *T* denotes the coverage range of sampling time t and B' denotes the fuzzy set processed by the minimum inference engine.

#### *2.3. Optimization Control Strategy for Single Cooling Machine of Chilled Water System*

Among heuristic optimization methods, genetic algorithms are most commonly used. Such algorithms are based on the concept of gene combination in evolution theory. The mechanisms of evolution of life, namely crossover, mutation, and reproduction, are used in genetic algorithms to obtain an optimal solution to a problem. The operating procedures of a genetic algorithm are described below.
