New Air Energy-saving Optimization Control Method for Terminal Based on Prediction of Passenger Spatiotemporal Distribution

Abstract

The present invention relates to the field of intelligent control technology for airport terminal environment control systems, and particularly to an energy-saving optimization control method for fresh air in the terminal based on passenger spatio-temporal distribution prediction, comprising the following steps: S1. Based on the chi-square distribution, establish a passenger arrival probability model for the airport terminal, and use the principle of fluid dynamics to realize the prediction of the spatio-temporal distribution of passenger flow on the passenger flow line in the terminal; S2. Based on the gated recurrent unit network and the prediction results of passenger spatio-temporal distribution, establish an indoor air quality prediction model for the airport terminal; S3. Based on the fuzzy control theory and the prediction results of indoor air quality, form a fresh air fuzzy control strategy for the air conditioning system in the airport terminal, and give the optimal control input sequence for the fresh air volume of the air conditioning system. The present invention predicts the passenger flow in the terminal through flight dynamic information, and uses it as an input parameter for predicting the indoor air quality in the terminal, thereby realizing the energy-saving optimization control of the fresh air volume of the air conditioning system in the terminal.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control of the environmental control system in airport terminals, and particularly to an optimized control method for fresh air energy saving in terminals based on the prediction of passenger spatio-temporal distribution. Background Art

[0002] As an important urban transportation hub, the airport terminal has important strategic significance for the sustainable development of the city and the urbanization construction. With the rapid development of China's civil aviation industry, the throughput and passenger volume of the whole country increased in 2019, which promoted the construction of new airports and the expansion of existing airports in China. As the core area serving passengers, the indoor environment of the airport terminal is crucial for the quality of passenger service. Especially since the outbreak of the Covid-19 pandemic, the indoor air quality has become a key factor affecting passenger health. On the other hand, due to the characteristics of large-scale personnel flow and long annual operating hours, the average energy consumption intensity of the terminal is 2.9 times that of ordinary public buildings and 8.0 times that of urban residential buildings. Among them, the energy consumption of the heating, ventilation and air conditioning (HVAC) system accounts for 40% - 80% of the total energy consumption of the terminal (DOI: 10.1016 / j.buildenv.2019.03.011; DOI: 10.1016 / j.scs.2021.103619; DOI: 10.1016 / j.enbenv.2022.06.006; DOI: 10.1016 / j.buildenv.2018.02.009). Therefore, the operation strategy of the terminal HVAC system not only affects the health of passengers during travel, but also is a key factor in energy consumption. Both are closely related to personnel activities. The personnel activities in the terminal are mainly centered around passengers and are travel-oriented, which are closely related to flight schedules and have obvious planning and predictability. Therefore, predicting the passenger flow in the terminal through flight dynamic information, monitoring the indoor air quality, and then adjusting the HVAC system is an effective way to ensure the safety, health, comfort and energy saving of the indoor environment in the terminal. The present invention discloses an optimized control method for fresh air energy saving in terminals based on the prediction of passenger spatio-temporal distribution, mainly by establishing a passenger arrival probability model for the terminal through chi-square distribution to realize the prediction of passenger spatio-temporal distribution in the terminal, using a gated recurrent unit network to establish a prediction model for the indoor air quality in the terminal, and finally adopting a fuzzy control strategy to realize the intelligent control of the terminal HVAC system. The concentration of CO 2 is an important indicator characterizing indoor air quality. This patent takes the CO 2 concentration in the terminal as a reference basis, and aims at the correlation characteristics of passenger flow - CO 2 concentration - fresh air volume of the air conditioning system in the terminal. By using statistical analysis theory and machine learning methods, an optimized control method for fresh air energy saving in terminals based on the prediction of passenger spatio-temporal distribution is proposed, which provides important technical support for creating a safe, healthy and low-carbon indoor environment in airport terminals and promoting the green and high-quality development of civil aviation. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a fresh air energy-saving optimization control method for airport terminals based on the prediction of passenger spatio-temporal distribution, which can create a safe, healthy and low-carbon indoor environment for airport terminals.

[0004] Technical solution of the present invention:

[0005] The fresh air energy-saving optimization control method for the terminal building based on the prediction of passenger spatio-temporal distribution is as follows:

[0006] S1. Prediction of passenger spatio-temporal distribution in the airport terminal building: Based on the chi-square distribution, establish a passenger arrival probability model for the airport terminal building, and use the idea of fluid dynamics to realize the prediction of the passenger flow spatio-temporal distribution of the passenger flow line in the terminal building. The specific steps are as follows:

[0007] S1.1 Chi-square distribution model of passenger arrival probability in the terminal building

[0008] The passenger flow in the airport terminal building is closely related to the flight schedule time, and has significant planning and predictability; by extracting flight information and passenger security inspection information, the earliest and latest arrival times of passengers are respectively expressed as t EA and t LA ; the sampling interval is set to ε, and the correlation between the passenger arrival probability in the terminal building and the flight schedule is analyzed; the chi-square distribution probability density function is used to establish a passenger arrival probability model for the terminal building;

[0009]

[0010] Among them, f(t) represents the percentage of passenger arrivals; Γ(·) represents the gamma function; t represents the passenger arrival time; t SD 、t EA and t LA respectively represent the scheduled departure time of the flight and the earliest and latest arrival times of passengers; d is the degree of freedom; s is the transformation factor;

[0011] S1.2 Passenger spatio-temporal distribution prediction model in the terminal building

[0012] Using the idea of fluid dynamics, based on the chi-square distribution model of passenger arrival probability in the terminal building, establish a passenger flow spatio-temporal distribution prediction model for the passenger flow line in the terminal building; set the passenger spatio-temporal distribution prediction range to 24 hours, the model uses the relative time of a day, and assumes that passenger boarding follows a uniform distribution;

[0013]

[0014]

[0015] Among them, Z jDenotes the number of the j-th spatial unit; G f,i Denotes the boarding gate number of flight i; Denotes the number of passengers in the j-th spatial unit at time t; C f,i Denotes the passenger capacity of flight i; L in,j And L out,j Denotes the distances from the entrance and exit of the j-th spatial unit to the security check channel; m is the total number of flights within the prediction range; p is the passenger attendance rate; v is the average walking speed of passengers; g(t) is the passenger boarding probability distribution model; t SB And t EB Denotes the start boarding time and the end boarding time;

[0016] If the boarding gate G of flight i f,i Is in the spatial unit Z j , then passengers enter the spatial unit until they board and leave; if the boarding gate G of flight i f,i Is not in the spatial unit Z j , then passengers only pass through the spatial unit; the start boarding time t of the flight SB May be earlier than the latest arrival time t of passengers LA .

[0017] S1.3, Identification and Calibration of Passenger Spatiotemporal Distribution Prediction Model

[0018] The passenger spatiotemporal distribution prediction model contains four unknown parameters: passenger attendance rate p, average walking speed v of passengers, start boarding time t of passengers SB And end boarding time t of passengers EB , and model identification and calibration are required; define the evaluation indexes of the passenger spatiotemporal distribution prediction model, including root mean square error RMSE, mean absolute error MAPE and correlation index R 2 ; monitor the actual passenger flow in the terminal building, and use the particle swarm optimization algorithm to solve the unknown parameters of the model;

[0019]

[0020]

[0021]

[0022]

[0023]

[0024] Among them, And Represent the position and velocity of the α-th particle in the τ-th iteration; And gbest τIndicate the individual optimal value and the global optimal value in the τ-th iteration; a 1 and a 2 are learning factors and can take the value of 2; r 1 and r 2 are random numbers between 0 and 1; w is the inertia weight; Y t and respectively represent the real-time value and the average value of the actual passenger flow in the terminal building; err represents the error vector between the actual passenger flow in the terminal building and the predicted value of the model; n represents the number of samples.

[0025] S2. Prediction of Indoor Air Quality in Airport Terminal Buildings: Based on the gated recurrent unit network and the prediction results of passenger spatio-temporal distribution, establish an indoor air quality prediction model for airport terminal buildings. The specific steps are as follows:

[0026] S2.1. Theoretical Analysis of Indoor Air Quality in Terminal Buildings

[0027] The indoor air quality in the terminal building is closely related to the passenger flow, the ventilation volume of the air conditioning system, and the outdoor air quality;

[0028]

[0029]

[0030] G v = G o - G e (11)

[0031] Among them, G v , G o and G e respectively represent the ventilation volume, the fresh air volume, and the exhaust air volume; Q p and δ p respectively represent the total indoor pollutant generation amount and the per capita pollutant generation amount; c i and c o respectively represent the allowable indoor pollutant concentration and the outdoor pollutant concentration;

[0032] S2.2. Indoor Air Quality Prediction Model in Terminal Buildings

[0033] Use the gated recurrent unit network to establish an indoor air quality prediction model; the gated recurrent unit network is a variant of the recurrent network, consisting of an update gate and a reset gate, and its long-term memory ability can characterize the non-linear relationship between the input variable and the target variable;

[0034] Γ r,t = σ(W r h t-1 + U r X i,t ) (12)

[0035] Γ u,t = σ(W u h t-1 + U u X i,t )(13)

[0036]

[0037]

[0038] Y o,t = g(W p h t )(16)

[0039] err t = T t - Y o,t (17)

[0040] Among them, Γ r,t and Γ u,t respectively represent the outputs of the reset gate and the update gate; and h t respectively represent the candidate hidden state and the hidden state; X i,t , Y o,t , T t and err t respectively represent the input parameter, output parameter, target parameter and error vector of the recurrent unit; W r , W u , W h , U r , U u , U h and W p respectively represent the weight matrices of the reset gate, the update gate and the output unit; σ(·) and g(·) respectively represent the sigmoid and linear activation functions; ⊙ represents the Hadamard product;

[0041] S3. Fresh air fuzzy control of the airport terminal: Based on the fuzzy control theory and the prediction results of the indoor air quality, form the fresh air fuzzy control strategy for the air conditioning system of the airport terminal. The specific steps are as follows:

[0042] S3.1. Indoor air quality indicators of the terminal

[0043] According to the "Green Terminal Standard" MH / T 5032-2017, the indoor air pollutant concentration should meet the relevant requirements of the "Indoor Air Quality Standard" GB / T 18882-2002; at the same time, referring to the "Code for Indoor Environmental Pollution Control of Civil Building Engineering" GB50325-2010, the airport terminal is a public transportation waiting room, belonging to Class II civil building engineering, and the indoor air quality indicators and limits are shown in Table 1.

[0044] Table 1 Indoor Air Quality Indicators and Limits

[0045]

[0046]

[0047] S3.2, Fresh Air Fuzzy Control Strategy for Terminal

[0048] The fuzzy rules consider the instantaneous value and instantaneous change rate of the input variables, and its control performance is better than that of simple linear rules;

[0049] ΔG o,t = M(e c,t , ec c,t ) (18)

[0050] e c,t = c p,t - c i,t (19)

[0051] ec c,t = e c,t - e c,t-1 (20)

[0052] Among them, ΔG o,t represents the fresh air adjustment amount; c p,t is the predicted value of the indoor pollutant concentration; e c,t and ec c,t are respectively the deviation and deviation change of the indoor pollutant concentration from its allowable value; M(·) is the Mamdani fuzzy rule;

[0053] The above control strategy can adjust the fresh air supply volume of the air conditioning system according to the predicted value of the indoor air quality.

[0054] This method is also applicable to the fresh air energy-saving optimization control of transportation hubs such as railway stations, high-speed railway stations, and bus stations.

[0055] Compared with the existing technology, the beneficial effects of the invention are:

[0056] The present invention provides a fresh air energy-saving optimization control method for terminal buildings associated with flight and passenger information. By predicting the passenger flow in the terminal building through flight dynamic information and using it as an input parameter for predicting the indoor air quality in the terminal building, the energy-saving optimization control of the fresh air volume of the air conditioning system in the terminal building is realized, providing important technical support for creating a safe, healthy, and low-carbon indoor environment in the airport terminal building and promoting the green and high-quality development of civil aviation. Description of the Drawings

[0057] Figure 1 It is a framework diagram of the fresh air energy-saving optimization control strategy for airport terminal buildings based on the prediction of passenger spatio-temporal distribution.

[0058] Figure 2 It is a logic diagram of the fresh air energy-saving optimization control for airport terminal buildings based on the prediction of passenger spatio-temporal distribution.

[0059] Figure 3 It is a flow chart of the fresh air energy-saving optimization control strategy for airport terminal buildings based on the prediction of passenger spatio-temporal distribution.

[0060] Figure 4 It is a chi-square distribution model diagram of the arrival probability of passengers in the airport terminal building.

[0061] Figure 5 It is a schematic diagram of the principle of the indoor air quality prediction model for the airport terminal building.

[0062] Figure 6 It is a diagram of the membership function of the fuzzy rule.

[0063] Figure 7 Fresh air fuzzy rule inference diagram for the airport terminal building. Detailed Implementation Modes

[0064] The following describes in detail the specific implementation modes of the present invention in combination with the invention content, the drawings in the specification, and the formulas.

[0065] Referring to Figure 2 , the present invention is a fresh air energy-saving optimization control method for terminal buildings based on the prediction of passenger spatio-temporal distribution. Taking the fresh air energy-saving optimization control of a certain airport terminal building in Guangzhou as an example, the specific steps are as follows:

[0066] S1. Prediction of the spatio-temporal distribution of passengers in the airport terminal building: Referring to Figure 3 , based on the chi-square distribution, establish a passenger arrival probability model for the airport terminal building, and use the idea of fluid dynamics to realize the prediction of the spatio-temporal distribution of passenger flow on the passenger flow line in the terminal building. The specific steps are as follows:

[0067] S1.1. Chi-square distribution model of the arrival probability of passengers in the terminal building

[0068] The passenger flow in the airport terminal is closely related to the flight schedule, with significant planning and predictability. By extracting flight information and passenger security inspection information, the earliest and latest arrival times of passengers are set to 400 minutes and 20 minutes before the scheduled departure time of the flight, respectively, that is, t SD -t EA = 400 min, t SD -t LA = 20 min; The sampling interval ε is set to 5 minutes to analyze the correlation between the passenger arrival probability in the terminal and the flight schedule. The chi-square distribution probability density function is used to establish a passenger arrival probability model for the terminal;

[0069]

[0070] where f(t) represents the percentage of passengers arriving; Γ(·) represents the gamma function; t represents the passenger arrival time; t SD 、t EA and t LA represent the scheduled departure time of the flight and the earliest and latest arrival times of passengers, respectively; d is the degree of freedom; s is the transformation factor;

[0071] The identification results of the degree of freedom d and the transformation factor s of the chi-square distribution model of a certain airport terminal in Guangzhou are 7 and 0.0974, respectively;

[0072] S1.2, Passenger Spatiotemporal Distribution Prediction Model of the Terminal

[0073] Using the idea of fluid dynamics, based on the chi-square distribution model of passenger arrival probability in the terminal, a passenger flow spatiotemporal distribution prediction model of the terminal passenger flow line is established; The prediction range of passenger spatiotemporal distribution is set to 24 hours. The model uses the relative time of a day and assumes that passenger boarding follows a uniform distribution;

[0074]

[0075]

[0076] where Z j represents the number of the j-th spatial unit; G f,i represents the boarding gate number of flight i; represents the number of passengers in the j-th spatial unit at time t; C f,i represents the passenger capacity of flight i; L in,j and L out,j represent the distances from the entrance and exit of the j-th spatial unit to the security inspection channel; m is the total number of flights within the prediction range; p is the passenger attendance rate; v is the average walking speed of passengers; g(t) is the passenger boarding probability distribution model; t SB and t EB represent the start boarding time and the end boarding time;

[0077] If the boarding gate G of flight i f,i is in the spatial unit Z j , then the passengers enter the spatial unit until they board and leave; if the boarding gate G of flight i f,i is not in the spatial unit Z j , then the passengers only pass through the spatial unit; the boarding start time t of the flight SB may be earlier than the latest arrival time t of the passengers LA .

[0078] S1.3, Identification and Calibration of Passenger Spatiotemporal Distribution Prediction Model

[0079] The passenger spatiotemporal distribution prediction model contains four unknown parameters: passenger attendance rate p, average passenger walking speed v, passenger boarding start time t SB and boarding cut-off time t EB . Model identification and calibration are required; define the evaluation indexes of the passenger spatiotemporal distribution prediction model, including root mean square error RMSE, mean absolute error MAPE and correlation index R 2 ; monitor the actual passenger flow in the terminal building, and use the particle swarm optimization algorithm to solve the unknown parameters of the model;

[0080]

[0081]

[0082]

[0083]

[0084]

[0085] Among them, and represent the position and velocity of the α-th particle in the τ-th iteration; and gbest τ represent the individual optimal value and the global optimal value in the τ-th iteration; a 1 and a 2 are learning factors, which can take 2; r 1 and r 2 are random numbers between 0 and 1; w is the inertia weight; Y t and represent the real-time value and the average value of the actual passenger flow in the terminal building respectively; err represents the error vector between the actual passenger flow in the terminal building and the model prediction value; n represents the number of samples;

[0086] In a certain airport terminal building in Guangzhou, the Wi-Fi indoor positioning technology is used to calibrate the assumed parameters of the passenger spatio-temporal distribution prediction model. The passenger attendance rate p is 0.84, the average walking speed v of passengers is 1.21 m / s, the time t when passengers start boarding SB and the cut-off boarding time t EB are 42 minutes and 23 minutes before the flight takes off respectively;

[0087] S2. Prediction of indoor air quality in the airport terminal building: Refer to Figure 3 , taking the CO 2 concentration prediction as an example, based on the gated recurrent unit network and the prediction results of passenger spatio-temporal distribution, an indoor CO 2 concentration prediction model is established. The specific steps are as follows:

[0088] S2.1. Theoretical analysis of indoor air quality in the terminal building

[0089] The indoor CO 2 concentration is closely related to the passenger flow, the ventilation volume of the air conditioning system and the outdoor CO 2 concentration;

[0090]

[0091]

[0092] G v = G o - G e (11)

[0093] Among them, G v , G o and G e represent the ventilation volume, the fresh air volume and the exhaust air volume respectively; Q p and δ p represent the total indoor pollutant generation amount and the per capita pollutant generation amount respectively; c i and c o represent the allowable indoor pollutant concentration and the outdoor pollutant concentration respectively;

[0094] The fresh air volume of a certain airport terminal building in Guangzhou is determined by the speed of the fresh air unit and the opening degree of the fresh air valve, and the exhaust air volume is determined by the start-stop state of the exhaust fan;

[0095] S2.2. Indoor air quality prediction model in the terminal building

[0096] Refer to Figure 5 , using the gated recurrent unit network, an indoor CO 2 concentration prediction model is established; The gated recurrent unit network is a variant of the recurrent network, consisting of an update gate and a reset gate, and its long-term memory ability can characterize the non-linear relationship between the input variable and the target variable;

[0097] Γ r,t = σ(W r h t-1 + U r X i,t ) (12)

[0098] Γ u,t = σ(W u h t-1 + U u X i,t ) (13)

[0099]

[0100]

[0101] Y o,t = g(W p h t ) (16)

[0102] err t = T t - Y o,t (17)

[0103] Among them, Γ t,t and Γ u,t respectively represent the outputs of the reset gate and the update gate; and h t respectively represent the candidate hidden state and the hidden state; X i,t , Y o,t , T t and err t respectively represent the input parameters, output parameters, target parameters and error vector of the recurrent unit; W r , W u , W h , U r , U u , U h and W p respectively represent the weight matrices of the reset gate, the update gate and the output unit; σ(·) and g(·) respectively represent the sigmoid and linear activation functions; ⊙ represents the Hadamard product;

[0104] The root mean square error of the CO 2 concentration prediction result in a certain airport terminal building in Guangzhou is less than 3 ppm, meeting the fresh air energy-saving control requirements of the air-conditioning system;

[0105] S3. Fresh air fuzzy control of the airport terminal building: Based on the fuzzy control theory and the indoor air quality prediction results, a fresh air fuzzy control strategy for the air-conditioning system of the airport terminal building is formed. The specific steps are as follows:

[0106] S3.1, Indoor Air Quality Index of the Terminal Building

[0107] According to the "Green Terminal Building Standard" MH / T 5032 - 2017, the concentration of indoor air pollutants should meet the relevant requirements of the "Indoor Air Quality Standard" GB / T 18882 - 2002; at the same time, referring to the "Code for Indoor Environmental Pollution Control of Civil Building Engineering" GB50325 - 2010, the airport terminal building is a public transportation waiting room, belonging to Class II civil building engineering. The indoor air quality index and limit values of the terminal building are shown in Table 1.

[0108] Table 1 Indoor Air Quality Index and Limit Values

[0109]

[0110] S3.2, New Air Fuzzy Control Strategy of the Terminal Building

[0111] Referring to Figure 6 and Figure 7 , the fuzzy rules consider the instantaneous value and instantaneous change rate of the input variables, and its control performance is better than that of simple linear rules;

[0112] ΔG o,t = M(e c,t , ec c,t ) (18)

[0113] e c,t = c i,t - c p,t (19)

[0114] ec c,t = e c,t - e c,t-1 (20)

[0115] Among them, ΔG o,t represents the new air regulation amount; c p,t is the predicted value of the indoor pollutant concentration; e c,t and ec c,t are respectively the deviation and deviation change between the indoor pollutant concentration and its allowable value; M(·) is the Mamdani fuzzy rule;

[0116] The optimization result of the fresh air volume of the air conditioning system in a certain airport terminal building in Guangzhou can maintain the indoor CO 2 concentration change range between 700 ppm and 800 ppm.

Claims

1. A new air energy-saving optimization control method for airport terminals based on passenger spatio-temporal distribution prediction, characterized in that, the steps are as follows: S1. Prediction of passenger spatio-temporal distribution in airport terminals: Based on the chi-square distribution, establish a passenger arrival probability model for airport terminals, and use the idea of fluid dynamics to realize the prediction of the passenger flow spatio-temporal distribution of the passenger flow line in the terminal. The specific steps are as follows: S1.

1. Chi-square distribution model of passenger arrival probability in the terminal By extracting flight information and passenger security inspection information, the earliest and latest arrival times of passengers are respectively expressed as t EA and t LA ; the sampling interval is set to ε, and the correlation between the arrival probability of passengers at the terminal and the number of flight schedules is analyzed; the probability density function of the chi-square distribution is used to establish a probability model for the arrival of passengers at the terminal; Among them, f(t) represents the percentage of passengers arriving at the port; Γ(·) represents the gamma function; t represents the time of passengers arriving at the port; t SD , t EA and t LA respectively represent the scheduled departure time of the flight and the earliest and latest arrival times of the passengers; d is the degree of freedom; s is the transformation factor; S1.

2. Prediction model of passenger spatio-temporal distribution in the terminal Using the idea of fluid dynamics, based on the chi-square distribution model of passenger arrival probability in the terminal, establish a prediction model of the passenger flow spatio-temporal distribution of the passenger flow line in the terminal; set the prediction range of passenger spatio-temporal distribution to 24 hours, the model uses the relative time of a day, and assume that passenger boarding follows a uniform distribution; Among them, Z j represents the number of the j-th spatial unit; G f,i represents the boarding gate number of flight i; N j t represents the number of passengers in the j-th spatial unit at time t; C f,i represents the passenger capacity of flight i; L in,j and L out,j represent the distances from the entrance and exit of the j-th spatial unit to the security check channel; m is the total number of flights within the prediction range; p is the passenger attendance rate; v is the average walking speed of passengers; g(t) is the passenger boarding probability distribution model; t SB and t EB represent the start boarding time and the end boarding time; If the boarding gate G of flight i f,i is in the spatial unit Z j , then the passenger enters the spatial unit until boarding and leaving; if the boarding gate G of flight i f,i is not in the spatial unit Z j , then the passenger only passes through the spatial unit; the boarding start time t of the flight SB may be earlier than the latest arrival time t of the passenger LA ; S1.

3. Identification and calibration of the passenger spatio-temporal distribution prediction model The passenger spatio-temporal distribution prediction model includes the passenger attendance rate p, the average walking speed v of passengers, the boarding start time t of passengers SB and the boarding cut-off time t EB There are four unknown parameters that need to be identified and calibrated in the model; define the evaluation indicators of the passenger spatio-temporal distribution prediction model, including the root mean square error RMSE, the mean absolute error MAPE, and the correlation index R 2 ; monitor the actual passenger flow in the terminal building, and use the particle swarm optimization algorithm to solve the unknown parameters of the passenger spatio-temporal distribution prediction model Among them, and represent the position and velocity of the α-th particle in the τ-th iteration; and gbest τ represent the individual optimal value and the global optimal value in the τ-th iteration; a 1 and a 2 are learning factors and can take the value of 2; r 1 and r 2 are random numbers between 0 and 1; w is the inertia weight; Y t and represent the real-time value and the average value of the actual passenger flow in the terminal building respectively; err represents the error vector between the actual passenger flow in the terminal building and the model prediction value; n represents the number of samples; S2. Prediction of indoor air quality in airport terminals: Based on the gated recurrent unit network and the prediction results of passenger spatio-temporal distribution, establish a prediction model of indoor air quality in airport terminals. The specific steps are as follows: S2.

1. Theoretical analysis of indoor air quality in the terminal The indoor air quality in the terminal is related to the passenger flow, the ventilation volume of the air conditioning system, and the outdoor air quality; G v = G o -G e (11) Among them, G v , G o and G e respectively represent the ventilation volume, fresh air volume, and exhaust air volume; Q p and δ p respectively represent the total indoor pollutant generation amount and the per capita pollutant generation amount; c i and c o respectively represent the allowable indoor pollutant concentration and the outdoor pollutant concentration; S2.

2. Prediction model of indoor air quality in the terminal Use the gated recurrent unit network to establish a prediction model of indoor air quality; Γ r,t = σ(W r h t-1 + U r X i,t ) (12) Γ u,t = σ(W u h t-1 + U u X i,t ) (13) Y o,t = g(W p h t ) (16) err t = T t - Y o,t (17) Among them, Γ r,t and Γ u,t represent the outputs of the reset gate and the update gate respectively; and h t represent the candidate hidden state and the hidden state respectively; X i,t 、Y o,t 、T t and err t represent the input parameter, output parameter, target parameter and error vector of the recurrent unit respectively; W r 、W u 、W h 、U r 、U u 、U h and W p represent the weight matrices of the reset gate, update gate and output unit respectively; σ(·) and g(·) represent the sigmoid and linear activation functions respectively; ⊙ represents the Hadamard product; S3. New air fuzzy control for airport terminals: Based on the fuzzy control theory and the prediction results of indoor air quality, form a new air fuzzy control strategy for the air conditioning system in airport terminals; ΔG o,t = M(e c,t , ec c,t ) (18) e c,t = c p,t - c i,t (19) ec c,t = e c,t - e c,t-1 (20) Among them, ΔG o,t represents the fresh air regulation amount; c p,t is the predicted value of the indoor pollutant concentration; e c,t and ec c,t are respectively the deviation and the deviation change between the indoor pollutant concentration and its allowable value; M(·) is the Mamdani fuzzy rule.

2. The new air energy-saving optimization control method for airport terminals based on passenger spatio-temporal distribution prediction according to claim 1, characterized in that, In the described step S1.3, a 1 and a 2 take the value 2.

Images