2.3.4. Mutation

Mutation entails randomly selecting a child and replacing one of its genes through a single transformation (replace 0 with 1 or replace 1 with 0). The odds of mutation are considerably lower than those of crossover. Figure 5 illustrates the mutation process.

**Figure 5.** Mutation process of a child.

In summary, the genetic algorithm first executes (A) coding, followed by (B) fitness value calculation and reproduction, (C) crossover, and (D) mutation. A selection process is executed to choose a parent from newly reproduced child individuals, and the fitness value is recalculated to determine the suitability of the selected individual becoming a new parent. This evolution process is repeated until a termination condition is reached and an optimal solution is obtained. Examples of termination conditions are outlined as follows: (1) a limited number of generations is reached, and (2) the optimal solution is not updated for a certain number of consecutive iterations. The search process is considered to have reached convergence when either condition is met. Figure 6 shows the overall operating procedure of the genetic algorithm.

**Figure 6.** Genetic algorithm flowchart.

The parameters adjusted for the applied genetic algorithm are as follows: chilled water flow, cooling water flow, and chilled water output temperature. The calculated flow is converted to chilled water pump and cooling water pump frequencies by using similar principles. Finally, the chilled water temperature is set. Thus, the three parameters are controlled simultaneously.

## **3. Case Study**

This study executed the proposed HVAC system optimization control strategy involving FCU temperature control in a university in Taipei City, Taiwan. The study focused on energy conservation optimization control for an HVAC system in the university to meet the required levels of indoor thermal comfort. The HVAC system was installed in the teaching and administration building of the university. This HVAC system was determined to comprise a chilled water system, a chilled water pump, a cooling water pump, and two cooling towers. The teaching and administration building was observed to be four stories high, with the HVAC system being operated from 08:00 to 18:00 from Mondays to Fridays; 42 FCUs supply cold air. This study used a Gateway communication system to monitor the water system operation status of the HVAC in the building. In addition, supervisory control and data acquisition (SCADA) was used for energy conservation control in the centralized HVAC system.

The SCADA functions are outlined as follows:

1. FCU data collection

Data, such as *Ti*, *Ts*, and wind speed of the 42 FCUs in the teaching and administration building, were stored in a back-end server using the Modbus transmission method.

2. Chilled water system power consumption monitoring

Digital electricity meters were installed on the chilled water system, chilled water pump, cooling pump, and cooling towers of the HVAC system. Power consumption data were wirelessly transmitted to the SCADA database.

3. Chilled water system flow and temperature monitoring

Flow meters were installed on the piping system of the chilled water system and cooling machine, and thermometers were installed on the outlet and return pipes of the chilled water system. Therefore, data concerning the flow of chilled and cooling water and the temperatures of chilled and cooling water in the outlet and return pipes of the HVAC system were transmitted to the PLC controller. Finally, the entire sensor-driven data of the HVAC system were obtained through SCADA using the Modbus transmission method.

4. Chilled water system ON/OFF and frequency change control

Frequency converters were installed on the chilled water pump, cooling water pump, and cooling tower fans of the chilled water system. Such converters can adjust the frequencies of the chilled water pump, cooling water pump, and cooling tower fans in the event of changes in heat load. Moreover, they can turn the setups on and o ff through automatic scheduling.

5. FCU temperature control

The FCU temperature control point in the teaching and administration building enabled automatically setting the indoor temperature according to dynamic FCU temperature settings. Such control was achieved by transmitting temperature settings through SCADA.

The HVAC system in the aforementioned building was operated at full load (i.e., 60 Hz) in 2016. The annual mean power consumption of the chilled water system was 33.1 kWh. During the summer (i.e., June, July, and August) of that year—characterized by high power consumption—the monthly mean power consumption of the chilled water system was 62.7 kWh. Figure 7 shows statistics of the annual outdoor temperature distribution and power consumption of the HVAC system.

**Figure 7.** Outdoor temperature and HVAC system power consumption.

*Appl. Sci.* **2019**, *9*, 2391

In general, outdoor temperature is a crucial factor affecting the power consumption of an HVAC system in a building. Figure 8 illustrates a regression model of the outdoor temperature and power consumption of the chilled water system in 2016. The reliability of this regression model was determined to be 90.9%. As shown in this figure, the power consumption of the chilled water system increased as the outdoor temperature increased. The regression equation for power consumption is presented as follows:

$$k\text{Vlh} = 4.56\text{x} - 76.179\text{,} \tag{9}$$

where the regression coefficient *x* = *To.*

**Figure 8.** Energy consumption regression model of the chilled water system.

This study also applied the proposed HVAC system optimization control strategy involving FCU temperature control to a single chilled water system with variable frequency from June–August 2017. This strategy was used to calculate the heat demand of the building to correct the supply of chilled water flow. Compared with the HVAC system operated at full load, the HVAC system operated at optimized settings provided by the proposed strategy achieved maximum energy conservation in August, saving 54.9% of energy. The maximum efficiency (in kW/RT) of the chilled water system was 15.6%. Tables 2–4 present comparisons of the performance of the HVAC system operated at full load with that of the system operated at the optimized load in terms of power consumption, RT produced, and chilled water system efficiency in summer (i.e., from June to August).


**Table 3.** System optimization control strategy involving FCU temperature information.



**Table 4.** Energy consumption comparison.

Figure 9 shows a comparison of outdoor temperatures in 2016 with those in 2017. To understand the energy conservation engendered by the proposed strategy, the outdoor temperatures in 2017 were substituted into Equation (9)—the regression model for calculating the power consumption of a chilled water system—to determine the energy conservation e ffectiveness associated with various outdoor temperatures. The actual power consumption engendered by the proposed strategy was compared with the baseline power consumption calculated using the regression equation. The comparison revealed a maximum energy conservation of 65.6% (Figure 10); the energy conservation increased when outdoor temperature increased. Therefore, the proposed strategy can achieve energy conservation.

**Figure 9.** Outdoor temperature distribution.

**Figure 10.** Energy consumption simulation of HVAC system regression model.
