Research On Thermal Comfort And Energy Saving Intelligent Optimization Control Of HVAC In Building

Heating,ventilation and air conditioning(HVAC)is a kind of equipment that consumes energy and usually causes greenhouse gas emissions.Similarly,HVAC is an important service to provide comfortable,healthy and efficient indoor environment for building occupants.With the increase of energy consumption in HVAC system and the increasing demand of residents for indoor thermal comfort,it is very important to study the control problem that can not only ensure the thermal comfort of indoor environment,but also reduce the energy consumption of HVAC.In order to solve this control problem,this paper uses data mining t echnology and machine learning algorithm to study indoor thermal comfort model and energy consumption prediction of HVAC system,puts forward the intelligent optimal control method of HVAC thermal comfort and energy saving,and simulates and develops the HVAC control system of intelligent building to achieve the goal of HVAC thermal comfort and energy saving.Firstly,aiming at the thermal comfort index,several commonly used thermal comfort indexes are listed,from which the predicted mean vote(PMV)is selected as the standard to measure the thermal comfort.However,the calculation of PMV equation is a complex nonlinear process,which is inconvenient for the real-time control application of air conditioning.The ASHRAE of American Society of heating,refrigeration and Air Conditioning Engineers is proposed Based on the actual thermal comfort data of RP-884 project,a BP network thermal comfort prediction model is established.The experimental results show that the method meets the accuracy requirements of thermal comfort and can be used for real-time control of HVAC.Secondly,the influencing factors of HVAC energy consumption are analyzed,and the influencing characteristics of the energy consumption model are established from the perspective of data availability.Combined with the quality characteristics of building energy consumption monitoring data,the data preprocessing of HVAC historical energy consumption data used in the establishment of the model is carried out,the missing value filling is carried out by the k-Nearest Neighbor algorithm,and the abnormal data identification and cleaning are carried out by the k-Means algorithm,so as to realize the availability of data.Then,three machine learning algorithms,support vector regression,random forest and artificial neural network,are used to build the energy consumption model of HVAC respectively.Through the experimental comparison,random forest algorithm is selected as the prediction method of energy consumption model.Finally,in order to solve the problems of modeling difficulty and high energy consumption in building HVAC control system,In this paper,a data-driven deep Q network(DQN)based optimal control framework for HVAC thermal comfort and energy consumption is proposed.The HVAC optimal control system is regarded as a Markov Decision Process,and the building environment state,HVAC system control action and reward function are defined.by using the established thermal comfort model and HVAC energy consumption model,the indoor comfort and HVAC energy consumption are predicted in real time,and the predicted results are used as the feedback of the set value action of the Reinforcement Learning controller,and then through DQN learning to optimize the energy consumption and thermal comfort control method,the set value of HVAC temperature and humidity is dynamically adjusted to achieve the purpose of comfort and energy saving.In order to verify the effectiveness of the control algorithm,the algorithm is applied to the simulation of the building environment using Energy Plus,and compared with the control effects of Agent-off baseline method and Q-learning control method.The experiment shows that the DQN control algorithm adopted in this paper has better control effect,can provide stable and comfortable environment conditions and achieve the purpose of energy saving.