Metro air conditioner cooling load all-weather prediction method integrating time sequence and environment

Abstract

The invention relates to the technical field of air conditioner load prediction, in particular to a subway air conditioner cooling load all-weather prediction method fusing time sequence and environment, which can correct a result of an environment context load prediction model through a time sequence cooling load prediction model under an extreme weather condition so as to ensure the robustness and precision of cooling load prediction. And in order to reduce the influence of data volatility on the prediction of the cold load prediction model, the cold load time sequence data is decomposed into a relatively stable sequence by adopting an empirical mode decomposition algorithm, so that the cold load prediction difficulty is reduced, and the prediction accuracy is improved. Through the combination of an empirical mode decomposition algorithm, an adaptive inertial weight krill swarm algorithm and a support vector machine, the adaptive inertial weight krill swarm algorithm can prevent the model from falling into local optimum, so that the generalization ability of the cold load prediction model is improved, and a cold load prediction value provides a basis for an air conditioning unit control strategy; therefore, the ventilation and air-conditioning energy consumption of the subway electrical equipment room is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of air conditioning load prediction, and in particular to an all-weather prediction method for subway air conditioning cooling load integrating time series and environment. Background Art

[0002] With the vigorous development of the rail transit industry, how to effectively reduce the energy consumption of subway equipment has gradually become the focus of rail transit operation and management units. According to statistics, the energy consumption of the station ventilation and air-conditioning system accounts for nearly half of the energy consumption of the entire subway station, and the energy consumption of the ventilation and air-conditioning system of the equipment management room accounts for about 30% of the total energy consumption of the station ventilation and air-conditioning system. At present, many energy-saving studies on ventilation and air-conditioning systems are focused on the ventilation and air-conditioning systems in public areas, while energy-saving studies on equipment management rooms have been neglected. The energy consumption of the ventilation and air-conditioning system of the equipment management room mainly comes from the electrical equipment room. If the energy consumption of the electrical equipment room can be effectively reduced, it will be of great significance to the energy saving of the ventilation and air-conditioning system of the entire subway station.

[0003] The electrical equipment room is mainly composed of various electrical equipment with high power. The air conditioning cooling load in the room is mainly the heating load of the equipment. Since the electrical equipment room is the nerve center of the entire station, it ensures the efficient and safe operation of the subway station. Therefore, the design value of the air conditioning cooling load of the electrical equipment room will be much larger than the actual cooling load at the beginning of the design. At the same time, the air conditioning in the electrical equipment room needs to be kept running all year round, and there is obvious waste of energy consumption in the ventilation and air conditioning system.

[0004] In response to the problems of mismatch between cooling supply and demand and unreasonable operation strategies in the electrical equipment rooms of subway stations, many patents have disclosed cooling load prediction methods. For example, patent CN202311754298.0 discloses a subway air-conditioning load prediction method based on air enthalpy, which takes outdoor temperature, passenger flow, etc. as input for prediction. Patent CN202410562681.4 discloses an air-conditioning load prediction method based on similar days and long short-term memory networks, which puts four features, namely maximum temperature, minimum temperature, average temperature, and average humidity, into the data set.

[0005] However, the prediction methods in the existing technology have not considered that extreme weather with sudden changes in temperature will cause model prediction failure. In addition, since most subway stations are underground, the sudden changes in the outside temperature have a large time delay on the subway stations, which leads to the problem of unsatisfactory prediction results. Summary of the invention

[0006] The purpose of the present invention is to provide a method for all-weather prediction of subway air-conditioning cooling load that integrates time series and environment, so as to solve the problem that the prediction methods in the prior art do not take into account extreme weather with sudden changes in temperature, which will lead to model prediction failure, and because most subway stations are underground, the sudden changes in external temperature have a large time delay on the subway stations, which in turn leads to unsatisfactory prediction results.

[0007] To achieve the above object, the present invention provides a method for predicting subway air conditioning cooling load all-weather by integrating time series and environment. The method for predicting subway air conditioning cooling load all-weather by integrating time series and environment comprises the following steps:

[0008] S1: Construct the time series cooling load prediction model dataset and the environmental context load prediction model dataset;

[0009] S2: Use the empirical mode decomposition algorithm to decompose the cooling load value and decompose the original cooling load time series into multiple relatively flat time series;

[0010] S3: Based on the time series cooling load prediction model data set, the combination of empirical mode decomposition algorithm, adaptive inertia weight krill swarm algorithm and support vector machine is used to jointly construct the time series cooling load prediction model. Based on the environmental context load prediction model data set, the combination of adaptive inertia weight krill swarm algorithm and support vector machine is used to jointly construct the environmental context load prediction model.

[0011] S4: Utilize the predicted values ​​of the time series cooling load prediction model and the environmental context load prediction model to obtain the optimal cooling load value at the next moment.

[0012] Among them, in step S1, the specific steps of constructing the time series cooling load prediction model data set are: measuring the air volume, temperature and humidity of the supply and return air in the room every hour, and calculating the air conditioning cooling load in the equipment room every hour by the following formula, and the cooling load value is the required cooling load value;

[0013] Q=Gρ(h h -h s )

[0014] Among them, Q is the air conditioning cooling load to be calculated, G is the air supply volume of the air conditioner, m 3 / s; ρ is the air density, which is 1.2kg / m 3 ;h h is the return air enthalpy of the air conditioning system, kg / m 3 ;h s is the air supply enthalpy of the air conditioning system, kg / m 3 ,h h and h s It can be obtained by looking up the temperature and humidity table.

[0015] Among them, in step S1, the specific steps of constructing the environmental context load prediction model data set are: measuring the outdoor temperature and humidity every hour and the weighted average of the temperatures of several cabinets in the computer room.

[0016] In step S2, when the original cooling load time series is decomposed into multiple relatively gentle time series, the subsequences of the decomposed wind are as follows:

[0017]

[0018]

[0019] …

[0020]

[0021] in, represents the cold negative value of the Nth subsequence at the tth moment, and satisfies

[0022] Wherein, in step S3, the adaptive inertia weight krill swarm algorithm is used to optimize the parameters of the support vector machine;

[0023] Each subsequence needs to build a corresponding cooling load prediction model, and take the first three cooling load values ​​as input and the fourth cooling load value as output;

[0024] The support vector machine consists of two parameters: the regularization parameter c and the kernel parameter g.

[0025] In step S3, the process of optimizing the regularization parameter c and the kernel parameter g of the support vector machine using the adaptive inertia weight krill swarm algorithm is as follows:

[0026] Step 1: Using the individual krill positions as the regularization parameter c and kernel parameter g of the support vector machine, randomly initialize the krill population and set the krill population parameters;

[0027] Step 2: Set the fitness function and determine the initial optimal krill individual position and the number of iterations t, t = 1;

[0028] Step 3: Using the current induction speed, foraging speed, and diffusion speed of each krill individual, update the movement speed and position of each krill individual;

[0029] Step 4: Perform a crossover operation based on a genetic algorithm on each updated krill individual position, and the value range of the crossover factor is [0.3, 0.5];

[0030] Step 5: Calculate the fitness of each krill individual position in the current krill population, taking the maximum fitness in the current krill population as The corresponding krill individual position is taken as the current optimal krill individual position

[0031] Step 6: Based on the fitness value of the current optimal krill individual and the fitness value of the optimal krill individual in the previous iteration Dynamically adjust the inertia weight of individual krill, and update the induced speed and foraging speed of individual krill;

[0032] Step 7: Determine whether the search accuracy or the maximum number of iterations is met. If not, set t=t+1 and return to step 3. Otherwise, output the current optimal krill individual position and obtain the regularization parameter c and kernel parameter g of the cold load prediction model based on support vector machine.

[0033] Among them, in step S3, when the combination of the adaptive inertia weight krill swarm algorithm and the support vector machine is used to construct the environmental context load prediction model, the input of the model is the weighted average of the outdoor temperature, humidity and several cabinet temperatures at the previous moment, the weighted average of the outdoor temperature, humidity and several cabinet temperatures at the current moment, and the cold negative value at the current moment, which is specifically expressed as follows:

[0034]

[0035] in, is the outdoor temperature at the previous moment, is the outdoor temperature at the current moment, is the outdoor humidity at the previous moment, is the outdoor humidity at the current moment, is the weighted average value of the cabinet temperature at the previous moment, is the weighted average value of the cabinet temperature at the current moment, Q t is the cooling load value at the current moment, and the output is: Q t+1 , which is the cooling load value at the next moment.

[0036] Among them, under normal climatic conditions, that is, when the absolute value of the temperature difference between the current moment and the previous moment is less than the threshold, the predicted values ​​of the time series cooling load prediction model and the environmental context load prediction model are weighted averaged to obtain the optimal cooling load value at the next moment;

[0037] Under abnormal climatic conditions, that is, when the absolute value of the temperature difference between the current moment and the previous moment is greater than or equal to the threshold, the environmental context load prediction model is more sensitive to temperature, and there is a significant time delay for the actual ground temperature to be transmitted to the underground. Therefore, the predicted value of the time series cooling prediction model is used as the optimal cooling load value at the next moment.

[0038] The present invention discloses a subway air-conditioning cooling load all-weather prediction method integrating time series and environment, which can correct the result of the environmental context load prediction model through the time series cooling load prediction model under extreme weather conditions, thereby ensuring the robustness and accuracy of the cooling load prediction, and in order to reduce the influence of data volatility on the prediction of the cooling load prediction model, the empirical mode decomposition algorithm is used to decompose the cooling load time series data into a relatively stable sequence, thereby reducing the difficulty of cooling load prediction and improving the accuracy of prediction, and through the combination of the empirical mode decomposition algorithm + the adaptive inertia weight krill swarm algorithm + the support vector machine, the adaptive inertia weight krill swarm algorithm can avoid the model from falling into the local optimum, thereby improving the generalization ability of the cooling load prediction model, and the cooling load prediction value provides a basis for the air-conditioning unit control strategy, thereby effectively reducing the ventilation and air-conditioning energy consumption of the subway electrical equipment room. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 It is a flowchart of the steps of the all-weather prediction method for subway air conditioning cooling load integrating time series and environment provided by the present invention.

[0041] Figure 2 It is a schematic diagram of the training of the time series cooling load prediction model provided by the present invention.

[0042] Figure 3 It is a schematic diagram of the environmental context load prediction model training provided by the present invention.

[0043] Figure 4 It is a flow chart of the dual-mode fusion prediction process provided by the present invention. DETAILED DESCRIPTION

[0044] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0045] See also Figures 1 to 4 The present invention provides a method for predicting subway air conditioning cooling load all-weather by integrating time series and environment. The method for predicting subway air conditioning cooling load all-weather by integrating time series and environment comprises the following steps:

[0046] S1: Construct the time series cooling load prediction model dataset and the environmental context load prediction model dataset;

[0047] S2: Use the empirical mode decomposition algorithm to decompose the cooling load value and decompose the original cooling load time series into multiple relatively flat time series;

[0048] S3: Based on the time series cooling load prediction model data set, the combination of empirical mode decomposition algorithm, adaptive inertia weight krill swarm algorithm and support vector machine is used to jointly construct the time series cooling load prediction model. Based on the environmental context load prediction model data set, the combination of adaptive inertia weight krill swarm algorithm and support vector machine is used to jointly construct the environmental context load prediction model.

[0049] S4: Utilize the predicted values ​​of the time series cooling load prediction model and the environmental context load prediction model to obtain the optimal cooling load value at the next moment.

[0050] In this embodiment, the result of the environmental context load prediction model can be corrected through the time series cooling load prediction model under extreme weather conditions, thereby ensuring the robustness and accuracy of the cooling load prediction, and in order to reduce the impact of data volatility on the prediction of the cooling load prediction model, the empirical mode decomposition algorithm is used to decompose the cooling load time series data into a relatively stable sequence, thereby reducing the difficulty of cooling load prediction and improving the accuracy of the prediction. Through the combination of the empirical mode decomposition algorithm + the adaptive inertia weight krill swarm algorithm + the support vector machine, the adaptive inertia weight krill swarm algorithm can avoid the model from falling into the local optimum, thereby improving the generalization ability of the cooling load prediction model, and the cooling load prediction value provides a basis for the air-conditioning unit control strategy, thereby effectively reducing the ventilation and air-conditioning energy consumption of the subway electrical equipment room.

[0051] Among them, in step S1, the specific steps of constructing the time series cooling load prediction model data set are: measuring the air volume, temperature and humidity of the supply and return air in the room every hour, and calculating the air conditioning cooling load in the equipment room every hour by the following formula, and the cooling load value is the required cooling load value;

[0052] Q=Gρ(h h -h s )

[0053] Among them, Q is the air conditioning cooling load to be calculated, G is the air supply volume of the air conditioner, m 3 / s; ρ is the air density, which is 1.2kg / m 3 ;h h is the return air enthalpy of the air conditioning system, kg / m 3 ;h s is the air supply enthalpy of the air conditioning system, kg / m 3 ,h h and hs It can be obtained by looking up the temperature and humidity table.

[0054] Among them, in step S1, the specific steps of constructing the environmental context load prediction model data set are: measuring the outdoor temperature and humidity every hour and the weighted average of the temperatures of several cabinets in the computer room.

[0055] In step S2, when the original cooling load time series is decomposed into multiple relatively gentle time series, the subsequences of the decomposed wind are as follows:

[0056]

[0057] …

[0058]

[0059] in, represents the cold negative value of the Nth subsequence at the tth moment, and satisfies

[0060] Among them, in step S3, the adaptive inertia weight krill swarm algorithm is used to optimize the parameters of the support vector machine; each subsequence needs to build a corresponding cooling load prediction model, and the first three cooling load values ​​are used as input and the fourth cooling load value is used as output; the support vector machine includes two parameters: regularization parameter c and kernel parameter g.

[0061] In step S3, the process of optimizing the regularization parameter c and the kernel parameter g of the support vector machine using the adaptive inertia weight krill swarm algorithm is as follows:

[0062] Step 1: Using the individual krill positions as the regularization parameter c and kernel parameter g of the support vector machine, randomly initialize the krill population and set the krill population parameters;

[0063] Step 2: Set the fitness function and determine the initial optimal krill individual position and the number of iterations t, t = 1;

[0064] Step 3: Using the current induction speed, foraging speed, and diffusion speed of each krill individual, update the movement speed and position of each krill individual;

[0065] Step 4: Perform a crossover operation based on a genetic algorithm on each updated krill individual position, and the value range of the crossover factor is [0.3, 0.5];

[0066] Step 5: Calculate the fitness of each krill individual position in the current krill population, taking the maximum fitness in the current krill population as The corresponding krill individual position is taken as the current optimal krill individual position

[0067] Step 6: Based on the fitness value of the current optimal krill individual and the fitness value of the optimal krill individual in the previous iteration Dynamically adjust the inertia weight of individual krill, and update the induced speed and foraging speed of individual krill;

[0068] Step 7: Determine whether the search accuracy or the maximum number of iterations is met. If not, set t=t+1 and return to step 3. Otherwise, output the current optimal krill individual position and obtain the regularization parameter c and kernel parameter g of the cold load prediction model based on support vector machine.

[0069] Furthermore, in step 1, the krill population contains multiple krill individuals, the krill population size ranges from [200, 200], and the foraging inertia weight w f The value range is [0.4, 0.7], and the induced inertia weight w n The value range of is [0.15, 0.75], the value range of the maximum induced speed is [0.03, 0.09], the value range of the maximum foraging speed is [0.03, 0.09], the value range of the maximum diffusion speed is [0.005, 0.01], the value range of the step scaling factor is [0.3, 0.7], the value range of the maximum number of iterations T is [400, 1000], and the value range of the search accuracy is [0.005, 0.15];

[0070] In step 2, the regularization parameter c and kernel parameter g corresponding to the krill individual position are substituted into the prediction model of the support vector machine, and the state score of the target individual is calculated using the support vector machine prediction model determined by the krill position. The inverse of the mean square error MSE between the predicted value and the actual value of the cooling load is used as the fitness function f1(x); the fitness function is used to calculate the fitness of each krill individual position, and the maximum fitness is used as the maximum fitness. The corresponding krill individual position is taken as the initial optimal krill individual position

[0071] In step 6, the formula for dynamically adjusting the inertia weight of individual krill is as follows:

[0072]

[0073] Among them, the inertia weight of krill individuals includes the induction speed inertia weight and the foraging speed inertia weight, w n and w f They represent the initialized induced speed inertia weight and foraging speed inertia weight, and They represent the induced speed inertia weight and foraging speed inertia weight of the krill individual after the tth iteration respectively;

[0074] In step 7, the search accuracy is achieved when the difference between the fitness value of the current optimal krill individual and the fitness value of the optimal krill individual in the previous iteration is less than the set search accuracy. The predicted values ​​of N models are added together to obtain the final predicted cooling load value. The schematic diagram of the training of the time series cooling load prediction model is shown in FIG. Figure 2 shown.

[0075] Among them, in step S3, when the combination of the adaptive inertia weight krill swarm algorithm and the support vector machine is used to construct the environmental context load prediction model, the input of the model is the weighted average of the outdoor temperature, humidity and several cabinet temperatures at the previous moment, the weighted average of the outdoor temperature, humidity and several cabinet temperatures at the current moment, and the cold negative value at the current moment, which is specifically expressed as follows:

[0076]

[0077] in, is the outdoor temperature at the previous moment, is the outdoor temperature at the current moment, is the outdoor humidity at the previous moment, is the outdoor humidity at the current moment, is the weighted average value of the cabinet temperature at the previous moment, is the weighted average value of the cabinet temperature at the current moment, Q t is the cooling load value at the current moment, and the output is: Q t+1 , that is, the cooling load value at the next moment, and so on. The model is trained and the adaptive inertia weight krill swarm algorithm is used to optimize the parameters of the support vector machine. The same as above, the schematic diagram of the environmental context load prediction model training is shown in Figure 3 shown.

[0078] Among them, in step S4, under normal climatic conditions, that is, when the absolute value of the temperature difference between the current moment and the previous moment is less than the threshold, the predicted values ​​of the time series cooling load prediction model and the environmental context load prediction model are weighted averaged to obtain the optimal cooling load value at the next moment;

[0079] Under abnormal climatic conditions, that is, when the absolute value of the temperature difference between the current moment and the previous moment is greater than or equal to the threshold, since the environmental context load prediction model is more sensitive to temperature, and there is a significant delay in the actual ground temperature being transmitted to the underground, the predicted value of the time series cold prediction model is used as the optimal cold load value at the next moment, where the threshold can be flexibly adjusted according to the local climate and model usage experience. The dual-mode fusion prediction process flow chart is as follows: Figure 4 shown.

[0080] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for predicting subway air conditioning cooling load all-weather by integrating time series and environment, characterized in that: The steps include: S1: Construct the time series cooling load prediction model dataset and the environmental context load prediction model dataset; S2: Use the empirical mode decomposition algorithm to decompose the cooling load value and decompose the original cooling load time series into multiple relatively flat time series; S3: Based on the time series cooling load prediction model data set, the combination of empirical mode decomposition algorithm, adaptive inertia weight krill swarm algorithm and support vector machine is used to jointly construct the time series cooling load prediction model. Based on the environmental context load prediction model data set, the combination of adaptive inertia weight krill swarm algorithm and support vector machine is used to jointly construct the environmental context load prediction model. S4: Utilize the predicted values ​​of the time series cooling load prediction model and the environmental context load prediction model to obtain the optimal cooling load value at the next moment.

2. The all-weather prediction method for subway air conditioning cooling load integrating time series and environment as claimed in claim 1 is characterized in that: In step S1, the specific steps of constructing the time series cooling load prediction model data set are: measuring the air volume, temperature and humidity of the supply and return air in the room every hour, and calculating the air conditioning cooling load in the equipment room every hour by the following formula, and the cooling load value is the required cooling load value; Q=Gρ(h h -h s ) Among them, Q is the air conditioning cooling load to be calculated, G is the air supply volume of the air conditioner, m 3 / s; ρ is the air density, which is 1.2kg / m 3 ;h h is the return air enthalpy of the air conditioning system, kg / m 3 ;h s is the air supply enthalpy of the air conditioning system, kg / m 3 ,h h and h s It can be obtained by looking up the temperature and humidity table.

3. The all-weather prediction method for subway air conditioning cooling load integrating time series and environment as claimed in claim 2 is characterized in that: In step S1, the specific steps of constructing the environmental context load prediction model data set are: measuring the outdoor temperature and humidity every hour and the weighted average temperature of several cabinets in the computer room.

4. The all-weather prediction method for subway air conditioning cooling load integrating time series and environment as claimed in claim 3 is characterized in that: In step S2, when the original cooling load time series is decomposed into multiple relatively gentle time series, the subsequences of decomposed wind are as follows: in, represents the cold negative value of the Nth subsequence at the tth moment, and satisfies 5. The all-weather prediction method for subway air conditioning cooling load integrating time series and environment as claimed in claim 4 is characterized in that: In step S3, the adaptive inertia weight krill swarm algorithm is used to optimize the parameters of the support vector machine; Each subsequence needs to build a corresponding cooling load prediction model, and take the first three cooling load values ​​as input and the fourth cooling load value as output; The support vector machine consists of two parameters: the regularization parameter c and the kernel parameter g.

6. The all-weather prediction method for subway air conditioning cooling load integrating time series and environment as claimed in claim 5 is characterized in that: In step S3, the process of optimizing the regularization parameter c and the kernel parameter g of the support vector machine using the adaptive inertia weight krill swarm algorithm is as follows: Step 1: Using the individual krill positions as the regularization parameter c and kernel parameter g of the support vector machine, randomly initialize the krill population and set the krill population parameters; Step 2: Set the fitness function and determine the initial optimal krill individual position and the number of iterations t, t = 1; Step 3: Using the current induction speed, foraging speed, and diffusion speed of each krill individual, update the movement speed and position of each krill individual; Step 4: Perform a crossover operation based on a genetic algorithm on each updated krill individual position, and the value range of the crossover factor is [0.3, 0.5]; Step 5: Calculate the fitness of each krill individual position in the current krill population, taking the maximum fitness in the current krill population as The corresponding krill individual position is taken as the current optimal krill individual position Step 6: Based on the fitness value of the current optimal krill individual and the fitness value of the optimal krill individual in the previous iteration Dynamically adjust the inertia weight of individual krill, and update the induced speed and foraging speed of individual krill; Step 7: Determine whether the search accuracy or the maximum number of iterations is met. If not, set t=t+1 and return to step 3. Otherwise, output the current optimal krill individual position and obtain the regularization parameter c and kernel parameter g of the cold load prediction model based on support vector machine.

7. The all-weather prediction method for subway air conditioning cooling load integrating time series and environment as claimed in claim 6 is characterized in that: In step S3, when the combination of the adaptive inertia weighted krill swarm algorithm and the support vector machine is used to construct the environmental context load prediction model, the input of the model is the weighted average of the outdoor temperature, humidity and several cabinet temperatures at the previous moment, the weighted average of the outdoor temperature, humidity and several cabinet temperatures at the current moment, and the cold negative value at the current moment, which is specifically expressed as follows: in, is the outdoor temperature at the previous moment, is the outdoor temperature at the current moment, is the outdoor humidity at the previous moment, is the outdoor humidity at the current moment, is the weighted average value of the cabinet temperature at the previous moment, is the weighted average value of the cabinet temperature at the current moment, Q t is the cooling load value at the current moment, and the output is: Q t+1 , which is the cooling load value at the next moment.

8. The all-weather prediction method for subway air conditioning cooling load integrating time series and environment as claimed in claim 7 is characterized in that: In step S4, under normal climatic conditions, that is, when the absolute value of the temperature difference between the current moment and the previous moment is less than the threshold, the prediction values ​​of the time series cooling load prediction model and the environmental context load prediction model are weighted averaged to obtain the optimal cooling load value at the next moment; Under abnormal climatic conditions, that is, when the absolute value of the temperature difference between the current moment and the previous moment is greater than or equal to the threshold, the environmental context load prediction model is more sensitive to temperature, and there is a significant time delay for the actual ground temperature to be transmitted to the underground. Therefore, the predicted value of the time series cooling prediction model is used as the optimal cooling load value at the next moment.

Images