Resarch On Control Strategy Of Air Conditioning System Of High-Speed Railway Station Based On Machine Learning

At present,the indoor air-conditioning system in the waiting hall of high-speed railway stations generally adopts the manual method to set the control parameters,and drives the airconditioning system to operate in a fixed working condition for a long time,so that the airconditioning system often can not satisfy the comfort experience of passengers and consumes a lot of energy.For example,over-adjustment of the air-conditioning system(the set temperature of the air-conditioning system is too low in summer)not only causes energy waste,but also affects the comfort of passengers,while the under-adjustment of the air-conditioning system(the set temperature of the air-conditioning system in summer is too high or the number of air-conditioning units turned on is insufficient)will greatly reduce the comfort experience of passengers.In view of the above problems,this paper uses a data-driven approach to study the control strategy of the air-conditioning system by considering the overall energy consumption of the air-conditioning system and the demand for human comfort.On the premise of ensuring the comfort of indoor passengers,reduce the energy consumption of air conditioning and achieve the purpose of overall energy saving.The main research contents of this thesis are as follows:1.The co-simulation platform of MATLAB-TRNSYS(Transient System Simulation Program)is built.Through the field investigation of the actual situation of the waiting hall of a high-speed railway station in Sichuan Province,and using it as the research object to build a simulation model of the building with TRNSYS software,at the same time,in order to realize real-time data communication between the simulation environment and the control algorithm,it also wrote the interactive file between MATLAB and TRNSYS,provides a good simulation environment for the comparison of experimental results of different control methods.2.A temperature control method of air conditioner based on PMV(Predicted Mean Vote)inverse calculation is proposed.Firstly,the thermal comfort PMV index is studied,and the solution method of PMV is given according to the mathematical expression of PMV,and analyzed the influence of four environmental variables on PMV,including indoor temperature,indoor relative humidity,indoor wind speed,and indoor average radiant temperature.Then,on the basis of the research on thermal comfort index,an air-conditioning temperature control method based on PMV back calculation is proposed.This method takes the PMV index as the control target of the air conditioner,and comprehensively considers six parameters related to the thermal comfort of the human body,of which some parameters related to the environment are collected by indoor sensors,the parameters related to the human body are manually set according to different weather or working conditions.Then use the PMV formula to iteratively calculate the most comfortable indoor temperature.Finally,seven different air-conditioning temperature setting modes are compared using the built building simulation environment.The results show that compared with the traditional constant temperature control method,this method can better meet the thermal comfort needs of the human body.3.An energy-saving and thermal comfort control algorithm for air-conditioning based on DDQN(Double Deep Q Network)is designed.Since the energy consumption of the airconditioning system is jointly affected by the indoor and outdoor environments,the control method based on PMV inverse calculation only considers the indoor environment.Although it can improve the indoor comfort,the energy-saving effect is not good.In order to solve this problem,an air-conditioning energy-saving and thermal comfort control algorithm based on deep reinforcement learning is designed.First,the optimal control of the air-conditioning system is modeled as an Markov model,and the environmental state,control actions,and reward and punishment functions are defined,and then the air-conditioning system makes control actions based on the current environmental state,and takes the thermal comfort in the next state and air-conditioning energy consumption as actions the real-time feedback,through the DDQN algorithm to optimize the control strategy to achieve the purpose of comfort and energy saving.Finally,the algorithm is trained and verified through the built building simulation environment,and compared with other control methods.The results show that compared with the control method based on DQN,the control method based on Q-learning,the control method based on PMV inverse calculation and the traditional constant temperature control method,the energy consumption of air conditioners is reduced by 1.68%,5.39%,6.42%,2.26%.