Research On Forecasting Method Of Dynamic Air Conditioning Loads Of Metro Stations Based On Passenger Flow Characteristics

The rapid development of urban rail transit has made its energy consumption problem increasingly obvious.If it is possible to predict the hourly air-conditioning load of the subway station to guide the optimization of the air-conditioning system operation plan,it can reduce the problem of the traditional passive control response lag and the cooling supply and demand mismatch,and achieve the purpose of energy-saving operation.First,a survey of the passenger flow characteristics of an island-style subway station in Xi'an revealed that the passengers waited longer in queue during the peak passenger flow,and there was a characteristic that the passenger's stay time in the station hall and platform changed significantly with the passenger flow and the frequency of departure.Therefore,combined with the research results and relevant scholars' research,pedestrian simulation software was used to simulate and reveal the passenger's stay time and passenger flow distribution law under different passenger flow and departure frequency,and a fitting relationship between passenger stay time and passenger flow and departure interval was proposed.Then,analyze the influencing factors of the air-conditioning load of the subway station,finely calculate the load related to the air-conditioning calculation number in the station according to the passenger stay time,and then analyze and sort out the other loads to obtain the hourly air-conditioning load of the subway station.Finally,according to the influencing factors of the air-conditioning load of the station,the required relevant parameters are collected and sorted out.Using factor analysis and correlation analysis and other statistical methods to filter and determine the appropriate historical moment data that the model should use as input parameters,the BP neural network air conditioning load prediction model and the BP air conditioning load prediction model optimized by genetic algorithm were established.Compare and verify the load forecast results.The results of the study on passengers' stay time show that with the increase of passenger flow,the queuing time of passengers at the security inspection facilities increases,resulting in the change of the dwelling time of inbound passengers in the station hall from 1min to 10 min.Station passengers stay on the platform for a shorter time,from 6min to 2min.Comparing the static and dynamic passenger residence time calculations at the station,the comparison of staff load shows that the air conditioning load of station hall staff based on dynamic dwelling time is highly volatile,and the difference between the two can reach 83.1kW at the peak of passenger flow.Platform staff load The fluctuation is relatively gentle,with a maximum load difference of-11.4kW.Screening the input parameters can improve the accuracy of the load prediction results.Compared with the BP model,the genetic algorithm optimized model load prediction results are more accurate,and the cold load prediction errors of the station hall and platform layers of the example stations are Less than 30 kW and 5kW,the forecast error of wet load is less than 20kg/h and 5kg/h,respectively.