In 2014, building accounts for 41% of the total U.S. Energy consumption (U.S. Energy Information Administration, 2015). Within this, 43% goes to the HVAC. The design of thermal comfort directly relates to the energy performance of the HVAC system. Moreover, thermal comfort has great variation between occupants, as is shown in the study of Takashi et al (Takashi, Shin-ichi, Takashi, & Masato, 2010). PMV is the most commonly adopted model for thermal comfort, but it does not take personal and seasonal variation into consideration. On the contrary, fuzzy logic expert system imbedded the complexity and uncertainty (fuzziness) definition (Dounis & Manolakis, 2001) of thermal comfort and integrates personalized preferences into the HVAC control system (Farrokh, Ali, Burcin, Tatiana, & Michael, 2014) and is promising in providing a higher level of thermal comfort a lower energy consumption. In the following section, we will have a brief review of the approaches of fuzzy control logic in building thermal environment control.

Fuzzy control is based on fuzzy logic, which maps continuous analogue input to propositions regarding membership of “fuzzy sets” (Wikipedia). Fuzzy set is an ordered pair of a set U and the membership function m: U -> [0, 1]. The term “fuzzy” expresses the “partial truth” or the degree of membership of fuzzy sets.

Fuzzy control is based on fuzzy logic, which maps continuous analogue input to propositions regarding membership of “fuzzy sets” (Wikipedia). Fuzzy set is an ordered pair of a set U and the membership function m: U -> [0, 1]. The term “fuzzy” expresses the “partial truth” or the degree of membership of fuzzy sets.

Comparing with other control approaches which also involves learning, the advantages of the fuzzy logic approach are: 1) fuzzy rules are suitable for both descriptive and predictive applications; 2) the results of the fuzzy control system is more interpretable, which is easier to provide guidance to the system design; 3) the form of membership functions can be adjusted easily, increasing the adaptability of the model to different control requirements; 4) the model also produces an importance-ranking of input variables (Wang, 2003).

The general process of a fuzzy control includes: 1) “fuzzification”: mapping inputs to fuzzy sets with membership functions, whose basic forms include “triangular, trapezoidal and Gaussian” distributions; 2) “Processing”: invoking a set of rules that decide the output 3) “deffuzification”: combining and mapping outputs to corresponding fuzzy sets and finally provide a “crisp” value, which can be either a specific value such as temperature set point, relative humidity or a binary ON/OFF decision (Jin, Sung, Youngchul, & Seung-Hoon, 2011).

Dounis and Manolakis designed a PMV-based thermal environment control system using fuzzy logic. There are two variables, PMV, ambient temperature, involved in the antecedent and three variables (actuators), “auxiliary heater, auxiliary cooler and window open angle (for natural ventilation), in the consequent. For the fuzzification stage, they chose the triangular shaped membership function. At the inference stage, rules are generated and eliminated based on context-based guidelines, for example, heating and cooling actuators should not operate at the same time. At the defuzzification state, a combined method of “mean of maxima” and “center of area” are used. The system is tested against a simulator consisting of mathematical models of outdoor temperature and humidity, window, room, indoor RH and PMV. The result showed that the PMV is effectively regulated with the fuzzy control system in both winter and summer.

Another PMV-based system that involves user input is designed by Maher et al. (Hamdi & Lachiver, 1998). The input to the fuzzy model are the traditional six variables involved in PMV calculation and the user input of thermal comfort level measured in a categorical metric of Cold / OK / Hot. The output of the model are User Thermal Comfort Level (UTCL), which is the fuzzy version of PMV adjusted with user input, temperature set point and air velocity. The performance of the fuzzy control system is compared with two other systems: system-1 with constant set point, system-2 with night setback thermal stat. The fuzzy control system has achieved more hours of thermal comfort and consumed 20% less energy than system-1 and 8% less energy than system-2.

Another approach, proposed by Farrokh et al. (Farrokh, Ali, Burcin, Tatiana, & Michael, 2014) , is to directly acquire the thermal preference of occupants and operate on this preference without “translating” it into PMV. They adopted a user input scale, referred to as Thermal Perception Index TPI) with a discrete integral sensational scale from -50 to 50. The physical thermal conditions are accessed in each room with a sensor network in order to investigate the correlation between user perception and actual environment parameters. A typical data point is an ordered pair of (TPI, temperature). The TPI values are divided into M fuzzy sets and “WM fuzzy pattern recognition approach” is used in deciding the corresponding temperature range for each fuzzy set. The controlled parameter is the temperature set point for each zone. Based on the learned comfort temperature, the optimal temperature set point is determined by minimizing the average deviation from comfort temperature. The system is tested in an office building with an interview in assessing the user satisfaction and a measurement of the air flow rate of the VAV system in determine the energy performance. They discovered an improvement satisfaction of thermal conditions and a 39% reduction in average air flow rate.

However, one of a slight drawback of the classical fuzzy control system is that the membership functions and rules selected without solid optimization, limiting the performance of the fuzzy system. Upon this reflection, an integrated system of fuzzy control and neural network is introduced in the study of Jin et al (Jin, Sung, Youngchul, & Seung-Hoon, 2011). This approach tunes the system performance and optimizes the parameters of the membership function. The performance between the two systems are accessed with “percentage of comfort period” and “deviation from set point”. In terms of the former criterion, the two systems perform equally well. The standard deviation of the adaptive approach has better performance with a 0.13 standard deviation comparing to 0.22 of the fuzzy-only approach.

In this summary, we have briefly reviewed some approaches in building thermal environment control systems using fuzzy control logic. The fuzzy control approach is suitable for this application as a result of the fuzzy nature of the thermal environment itself. Some systems are PMV-based, others operates directly on user preferences. There are some drawback of the PMV-based system: 1) the two human determined variables, clothing and metabolic rate, are usually not measured for actual occupants but are assumptions. 2) It requires a complicated sensor network for the measurement of the other environment related input variables of PMV. Thus some recent studies tends to prefer a direct user preference metric. Integrations between fuzzy control system and other AI-based method are proposed as a way of increasing the accuracy of the fuzzy control method.

References Dounis, A., & Manolakis, M. (2001). Design of a Fuzzy System for Living Space Thermal-comfort Regulation.

Farrokh, J., Ali, G., Burcin, B.-G., Tatiana, K., & Michael, O. (2014). User-led decentralized thermal comfort driven HVAC operations forimproved efficiency in office buildings.

Hamdi, M., & Lachiver, G. (1998). A Fuzzy Control System Based on the Human Sensation of Thermal Comfort.

Jin, W. M., Sung, K. J., Youngchul, K. b., & Seung-Hoon, H. (2011). Comparative study of artificial intelligence-based building thermal control methods - Application of fuzzy, adaptive neuro-fuzzy inference system,and artificial neural network.

Takashi, A., Shin-ichi, T., Takashi, Y., & Masato, S. (2010). Thermal comfort and productivity - Evaluation of workplace environment in a task conditioned office.

U.S. Energy Information Administration. (2015, 4 17).

Wang, L.-X. (2003). The WM Method Completed: A Flexible Fuzzy System Approach to Data Mining.

The general process of a fuzzy control includes: 1) “fuzzification”: mapping inputs to fuzzy sets with membership functions, whose basic forms include “triangular, trapezoidal and Gaussian” distributions; 2) “Processing”: invoking a set of rules that decide the output 3) “deffuzification”: combining and mapping outputs to corresponding fuzzy sets and finally provide a “crisp” value, which can be either a specific value such as temperature set point, relative humidity or a binary ON/OFF decision (Jin, Sung, Youngchul, & Seung-Hoon, 2011).

Dounis and Manolakis designed a PMV-based thermal environment control system using fuzzy logic. There are two variables, PMV, ambient temperature, involved in the antecedent and three variables (actuators), “auxiliary heater, auxiliary cooler and window open angle (for natural ventilation), in the consequent. For the fuzzification stage, they chose the triangular shaped membership function. At the inference stage, rules are generated and eliminated based on context-based guidelines, for example, heating and cooling actuators should not operate at the same time. At the defuzzification state, a combined method of “mean of maxima” and “center of area” are used. The system is tested against a simulator consisting of mathematical models of outdoor temperature and humidity, window, room, indoor RH and PMV. The result showed that the PMV is effectively regulated with the fuzzy control system in both winter and summer.

Another PMV-based system that involves user input is designed by Maher et al. (Hamdi & Lachiver, 1998). The input to the fuzzy model are the traditional six variables involved in PMV calculation and the user input of thermal comfort level measured in a categorical metric of Cold / OK / Hot. The output of the model are User Thermal Comfort Level (UTCL), which is the fuzzy version of PMV adjusted with user input, temperature set point and air velocity. The performance of the fuzzy control system is compared with two other systems: system-1 with constant set point, system-2 with night setback thermal stat. The fuzzy control system has achieved more hours of thermal comfort and consumed 20% less energy than system-1 and 8% less energy than system-2.

Another approach, proposed by Farrokh et al. (Farrokh, Ali, Burcin, Tatiana, & Michael, 2014) , is to directly acquire the thermal preference of occupants and operate on this preference without “translating” it into PMV. They adopted a user input scale, referred to as Thermal Perception Index TPI) with a discrete integral sensational scale from -50 to 50. The physical thermal conditions are accessed in each room with a sensor network in order to investigate the correlation between user perception and actual environment parameters. A typical data point is an ordered pair of (TPI, temperature). The TPI values are divided into M fuzzy sets and “WM fuzzy pattern recognition approach” is used in deciding the corresponding temperature range for each fuzzy set. The controlled parameter is the temperature set point for each zone. Based on the learned comfort temperature, the optimal temperature set point is determined by minimizing the average deviation from comfort temperature. The system is tested in an office building with an interview in assessing the user satisfaction and a measurement of the air flow rate of the VAV system in determine the energy performance. They discovered an improvement satisfaction of thermal conditions and a 39% reduction in average air flow rate.

However, one of a slight drawback of the classical fuzzy control system is that the membership functions and rules selected without solid optimization, limiting the performance of the fuzzy system. Upon this reflection, an integrated system of fuzzy control and neural network is introduced in the study of Jin et al (Jin, Sung, Youngchul, & Seung-Hoon, 2011). This approach tunes the system performance and optimizes the parameters of the membership function. The performance between the two systems are accessed with “percentage of comfort period” and “deviation from set point”. In terms of the former criterion, the two systems perform equally well. The standard deviation of the adaptive approach has better performance with a 0.13 standard deviation comparing to 0.22 of the fuzzy-only approach.

In this summary, we have briefly reviewed some approaches in building thermal environment control systems using fuzzy control logic. The fuzzy control approach is suitable for this application as a result of the fuzzy nature of the thermal environment itself. Some systems are PMV-based, others operates directly on user preferences. There are some drawback of the PMV-based system: 1) the two human determined variables, clothing and metabolic rate, are usually not measured for actual occupants but are assumptions. 2) It requires a complicated sensor network for the measurement of the other environment related input variables of PMV. Thus some recent studies tends to prefer a direct user preference metric. Integrations between fuzzy control system and other AI-based method are proposed as a way of increasing the accuracy of the fuzzy control method.

References Dounis, A., & Manolakis, M. (2001). Design of a Fuzzy System for Living Space Thermal-comfort Regulation.

*Applied Energy*, 119-144.Farrokh, J., Ali, G., Burcin, B.-G., Tatiana, K., & Michael, O. (2014). User-led decentralized thermal comfort driven HVAC operations forimproved efficiency in office buildings.

*Energy and Buildings*, 398-410.Hamdi, M., & Lachiver, G. (1998). A Fuzzy Control System Based on the Human Sensation of Thermal Comfort.

*IEEE World Congress on Computational Intelligence. Fuzzy Systems Proceedings*(pp. 487-492). IEEE.Jin, W. M., Sung, K. J., Youngchul, K. b., & Seung-Hoon, H. (2011). Comparative study of artificial intelligence-based building thermal control methods - Application of fuzzy, adaptive neuro-fuzzy inference system,and artificial neural network.

*Applied Thermal Engineering*, 2422-2429.Takashi, A., Shin-ichi, T., Takashi, Y., & Masato, S. (2010). Thermal comfort and productivity - Evaluation of workplace environment in a task conditioned office.

*Building and Environment*, 45-50.U.S. Energy Information Administration. (2015, 4 17).

*How much energy is consumed in residential and commercial buildings in the United States*. Retrieved from Independent Statistics and Analysis, U.S. Energy Information Administration: http://www.eia.gov/tools/faqs/faq.cfm?id=86&t=1Wang, L.-X. (2003). The WM Method Completed: A Flexible Fuzzy System Approach to Data Mining.

*IEEE Transactions of Fuzzy Systems*, 768-782. summary3-yujiexu.docx |