A Mall Central Air Conditioning Load Prediction Method, System and Medium Based on EMD-PSO-LSTM

Abstract

This invention discloses a method, system, and medium for predicting central air conditioning load in shopping malls based on EMD-PSO-LSTM. The method involves collecting temperature and flow data from the central air conditioning chilled water system, as well as outdoor temperature and humidity data, to calculate the load; constructing an LSTM model and training it using the EMD decomposition method, optimizing its parameters using the PSO method; and finally, predicting the load based on historical loads and outdoor temperature and humidity data. The empirical mode decomposition (EMD) method employed in this invention can transform non-stationary sequences into stationary ones, reducing nonlinearity and improving prediction accuracy and speed. Furthermore, the method utilizes historical data to identify correlations between loads, thereby enhancing model prediction accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of building energy conservation, and in particular relates to a method, system and medium for predicting the load of central air conditioning in a shopping mall based on EMD-PSO-LSTM. Background Art

[0002] Under the "dual carbon" strategic goal, the urban building energy conservation improvement project calls for the implementation of green and efficient cooling initiatives, the upgrading of cooling technologies and equipment, the optimization of load matching between supply and demand, and the significant improvement of cooling system energy efficiency. As a key component of large public buildings, large shopping malls feature large floor areas, high window-to-wall ratios, high occupancy density, long operating hours, a high density of various lighting appliances, and high energy consumption of central air conditioning. Their energy consumption per unit area is significantly higher than that of other large public buildings, presenting significant potential for energy savings.

[0003] During the design phase of the mall's central air conditioning system, designers based the system on a maximum cooling load to ensure it could meet the mall's comfort cooling needs during extreme weather conditions or peak customer traffic. However, during actual operation, the mall's cooling load fluctuates dynamically due to multiple factors, including outdoor meteorological parameters and indoor customer traffic density. Consequently, the central air conditioning system operates at partial load 90% of the time. The lack of appropriate management measures, cooling capacity adjustment methods, and control devices prevented the mall's central air conditioning system from dynamically optimizing and adjusting to the cooling load, resulting in an "oversupply" of cooling capacity, low energy efficiency, and significant energy waste. Therefore, accurate prediction of the mall's central air conditioning load is essential for dynamic air conditioning system regulation. Summary of the Invention

[0004] The main purpose of this invention is to overcome the shortcomings and deficiencies of existing technologies by providing a method, system, and medium for predicting central air conditioning load in shopping malls based on EMD-PSO-LSTM. This method combines historical data with real-time outdoor meteorological parameters to achieve accurate load prediction. Compared with existing methods for predicting central air conditioning load in shopping malls, this method can further improve prediction accuracy.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] One aspect of the present invention provides a method for predicting the load of central air conditioning in a shopping mall based on EMD-PSO-LSTM, comprising the following steps:

[0007] Collect temperature and flow data of the central air-conditioning chilled water system, as well as outdoor temperature and humidity data, to perform load calculations;

[0008] Build an LSTM model and train it based on the EMD decomposition method, and use the PSO method to optimize parameters;

[0009] Load forecasting is performed based on historical load and outdoor temperature and humidity data.

[0010] As a preferred technical solution, it is characterized in that the temperature and flow data of the central air-conditioning chilled water system are collected, specifically:

[0011] Arrange temperature sensors in the central air conditioning chilled water system, and set the temperature sensor at the water distributor to the supply water temperature T1, and the temperature sensor at the water collector to the return water temperature T2;

[0012] Arrange flow sensors in the central air-conditioning chilled water system and set the flow at the chilled water main as the chilled water flow rate q.

[0013] As a preferred technical solution, it is characterized in that the load calculation is specifically performed as follows:

[0014] Q=q×(T2-T1)×c×ρ

[0015] Where Q is the load at the current moment, c is the specific heat capacity of chilled water, and ρ is the density of chilled water.

[0016] As a preferred technical solution, it is characterized in that the construction of the LSTM model is specifically as follows:

[0017] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0018] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0019]

[0020]

[0021] O t =σ(W O ·[h t-1 ,x t ]+b O )

[0022] g t =O t *tan h(C t )

[0023] Among them, W is the weight term, b is the bias term, and is the sigmoid function; the input variable at each moment contains the unit state C at the previous momentt-1 , the intermediate state h at the previous moment t-1 , and the current input x t , the intermediate variables include the output f of the forget gate t , the output i of the input-output gate t and O t , and the output of the input node Output variables include the cell state C t and the intermediate state h t .

[0024] As a preferred technical solution, it is characterized in that the training of the LSTM model is specifically as follows:

[0025] Use EMD to decompose the load sequence x(t) to obtain several IMFs and a residual;

[0026] The load, outdoor temperature and outdoor humidity at the previous two moments and the previous one are spliced into a data sequence according to a, b and c;

[0027] Convert the data sequence into n×3×3 format and input it into the LSTM model. Use the current load Q as the predicted value for training.

[0028] The particle swarm algorithm is used to optimize the learning rate, number of hidden layers, number of nodes per layer, and number of iterations of the LSTM model;

[0029] For each IMF, make a forecast according to the above steps, and obtain the forecast value 0 to the forecast value n, where n is the number of IMFs;

[0030] The load forecast value is obtained by superimposing the forecast value 0 to the forecast value n and the residual;

[0031] Save the model in the above steps.

[0032] As a preferred technical solution, the load sequence x(t) is decomposed using EMD, specifically:

[0033] Find all extreme points in the load sequence x(t);

[0034] Use the envelope to connect all the maximum points to form e max Similarly, connect all the minimum points to form e min ;

[0035] Find the average value e of the upper and lower envelopes mean , and subtract it from the original sequence to get the new sequence H:

[0036]

[0037] H=x(t)-emean

[0038] Determine whether the new sequence H is an intrinsic mode function IFM based on the following criteria:

[0039] a) The difference between the number of extreme points and the number of zero points in the intrinsic mode function IFM is no more than 1;

[0040] b) At any time, the mean of the upper and lower envelopes of the intrinsic mode function IFM is 0;

[0041] c) If not, take H as the new x(t) and repeat the above steps until the criterion is met and IFM0 is obtained;

[0042] Each time an intrinsic mode function IFM is obtained, it is removed from x(t):

[0043] x(t)=x(t)-IFM0

[0044] Repeat the above steps until the remaining residual Res is a monotonic sequence or a constant sequence, that is:

[0045] x(t)=IFM0+IMF1+IMF2+…+IMFn+Res.

[0046] As a preferred technical solution, the particle swarm algorithm is specifically as follows:

[0047] Initialize each particle randomly;

[0048] Evaluate each particle and obtain the global optimum;

[0049] Determine whether the end condition is met, if so, end;

[0050] If the end condition is not met, the position and velocity of each particle are updated;

[0051] Update the fitness function of each particle;

[0052] Update the historical optimal position of each particle;

[0053] Go back to the step of evaluating each particle and obtaining the global optimum;

[0054] Determine whether the end condition is met.

[0055] As a preferred technical solution, the load forecasting based on historical load and outdoor temperature and humidity data is specifically as follows:

[0056] The actual load Q at the previous moment is calculated based on the historically collected data. t-1 and the actual load Q at the previous two moments t-2 , and the outdoor temperature T at the previous 2 moments 3,t-2, outdoor humidity w at the previous 2 moments t-2 , outdoor temperature T at the moment before 3,t-1 , outdoor humidity w at the moment before t-1 As an input item, it is input into the LSTM model for prediction to obtain the load forecast value Q at the current moment t .

[0057] Another aspect of the present invention provides a shopping mall central air conditioning load forecasting system based on EMD-PSO-LSTM, which is applied to the above-mentioned shopping mall central air conditioning load forecasting method based on EMD-PSO-LSTM, including a data acquisition and calculation module, a model building and training module, and a prediction module;

[0058] The data acquisition and calculation module is used to collect temperature and flow data of the central air-conditioning chilled water system, as well as outdoor temperature and humidity data, to perform load calculation;

[0059] The model building and training module is used to build an LSTM model and perform training based on the EMD decomposition method, and use the PSO method to optimize parameters;

[0060] The prediction module is used to perform load prediction based on historical load and outdoor temperature and humidity data.

[0061] Another aspect of the present invention provides a storage medium storing a program, which, when executed by a processor, implements the above-mentioned shopping mall central air-conditioning load prediction method based on EMD-PSO-LSTM.

[0062] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0063] (1) Empirical mode decomposition (EMD) can transform non-stationary sequences into stationary sequences, reduce nonlinearity, and help improve prediction accuracy and speed.

[0064] (2) A method for quickly predicting the central air-conditioning load in shopping malls is proposed, which can shorten the prediction time.

[0065] (3) Use historical data to mine the correlation between loads and improve the prediction accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of a method for predicting central air-conditioning load in a shopping mall based on EMD-PSO-LSTM according to an embodiment of the present invention;

[0067] Figure 2 This is a schematic structural diagram of a central air-conditioning system for a shopping mall according to an embodiment of the present invention;

[0068] Figure 3This is a schematic diagram of data sequence generation according to an embodiment of the present invention;

[0069] Figure 4 Schematic diagram of the structure of LSTM according to an embodiment of the present invention;

[0070] Figure 5 This is a schematic diagram of the LSTM prediction structure of an embodiment of the present invention;

[0071] Figure 6 is a flow chart of the PSO algorithm according to an embodiment of the present invention;

[0072] Figure 7 This is a schematic diagram of the central air-conditioning load prediction for a shopping mall according to an embodiment of the present invention;

[0073] Figure 8 This is a structural diagram of a shopping mall central air-conditioning load forecasting system based on EMD-PSO-LSTM in an embodiment of the present invention;

[0074] Figure 9 It is a schematic structural diagram of a storage medium according to an embodiment of the present invention. DETAILED DESCRIPTION

[0075] In order to enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0076] Example

[0077] This embodiment uses a data acquisition system to construct a central air-conditioning load sequence for a shopping mall. EMD (Empirical Mode Decomposition) is used to decompose the load sequence into several IMF (Intrinsic Mode Function) sequences and residuals to reduce nonlinearity. The IMF and outdoor meteorological parameters form an input sequence, which is input into an LSTM (Long Short-Term Memory) neural network for load forecasting. At the same time, PSO (Particle Swarm Optimization) is used to optimize LSTM-related hyperparameters to speed up the forecast. Finally, the outputs are superimposed to obtain the load forecast value.

[0078] like Figure 1 As shown, this embodiment provides a method for predicting the central air-conditioning load of a shopping mall based on EMD-PSO-LSTM, including the following steps:

[0079] S1. Data collection and calculation.

[0080] Arrange temperature sensors in the central air-conditioning chilled water system, and set the temperature sensor at the water distributor to the supply water temperature T1, and the temperature sensor at the water collector to the return water temperature T2 (as shown in the figure). Figure 1 ).

[0081] Arrange flow sensors in the central air-conditioning chilled water system and set the flow rate at the chilled water main to the chilled water flow rate q (e.g. Figure 2 ).

[0082] Place a temperature and humidity sensor outdoors, set the outdoor temperature to T3, and the outdoor humidity to w.

[0083] Data is collected every 20 minutes, and the current load Q is calculated according to the formula Q = q × (T2-T1) × c × ρ (c is the specific heat capacity of chilled water, ρ is the density of chilled water).

[0084] The calculated load Q, outdoor temperature T3, and outdoor humidity w are spliced into a sequence according to a, b, and c (such as Figure 3 shown).

[0085] S2. Build an LSTM model and train it.

[0086] (1) LSTM model structure;

[0087] like Figure 4 As shown, the LSTM model in this embodiment is specifically:

[0088] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0089] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0090]

[0091]

[0092] O t =σ(W O ·[h t-1 ,x t ]+b O )

[0093] ht =O t *tan h(C t )

[0094] Among them, W is the weight term, b is the bias term, and is the sigmoid function; the input variable at each moment contains the unit state C at the previous moment t-1 , the intermediate state h at the previous moment t-1 , and the current input x t , the intermediate variables include the output f of the forget gate t , the output i of the input-output gate t and O t , and the output of the input node Output variables include the cell state C t and the intermediate state h t .

[0095] (2) The LSTM model training process is as follows:

[0096] S2.1. Use EMD to decompose the load sequence x(t) and obtain several IMFs and a residual.

[0097] The load sequence x(t) is obtained by combining the loads collected in step S1 from time t=0 to time t=n into a sequence x(t).

[0098] Furthermore, in this embodiment, the load sequence x(t) is decomposed using EMD, specifically as follows:

[0099] Find all extreme points in the load sequence x(t);

[0100] Use the envelope to connect all the maximum points to form e max Similarly, connect all the minimum points to form e min ;

[0101] Find the average value e of the upper and lower envelopes mean , and subtract it from the original sequence to get the new sequence H:

[0102]

[0103] H=x(t)-e mean

[0104] Determine whether the new sequence H is an intrinsic mode function IFM based on the following criteria:

[0105] a) The difference between the number of extreme points and the number of zero points in the intrinsic mode function IFM is no more than 1;

[0106] b) At any time, the mean of the upper and lower envelopes of the intrinsic mode function IFM is 0;

[0107] c) If not, take H as the new x(t) and repeat the above steps until the criterion is met and IFM0 is obtained;

[0108] Each time an intrinsic mode function IFM is obtained, it is removed from x(t):

[0109] x(t)=x(t)-IFM0

[0110] Repeat the above steps until the remaining residual Rea is a monotonic sequence or a constant sequence, that is:

[0111] z(t)=IFM0+IMF1+IMF2+…+IMFn+Res.

[0112] S2.2, convert the data sequence into n×3×3 format, input it into LSTM model, and use the current load Q as the predicted value for training (e.g. Figure 5 shown).

[0113] S2.3, use the particle swarm algorithm to optimize the LSTM learning rate, number of hidden layers, number of nodes per layer, and number of iterations (such as Figure 6 shown).

[0114] Furthermore, Figure 6 As shown, the particle swarm algorithm is specifically as follows:

[0115] Initialize each particle randomly;

[0116] Evaluate each particle and obtain the global optimum;

[0117] Determine whether the end condition is met, if so, end;

[0118] If the end condition is not met, the position and velocity of each particle are updated;

[0119] Update the fitness function of each particle;

[0120] Update the historical optimal position of each particle;

[0121] Go back to the step of evaluating each particle and obtaining the global optimum;

[0122] Determine whether the end condition is met.

[0123] S2.4, such as Figure 7 As shown, for each IMF, the above steps are used to predict and obtain predicted values 0 to n.

[0124] The input to IMF0 is:

[0125] <![CDATA[Q t-2,0 ]]> <![CDATA[T 3,t-2 ]]> <![CDATA[w t-2 ]]> <![CDATA[Q t-1,0 ]]> <![CDATA[T 3,t-1 ]]> <![CDATA[w t-1 ]]>

[0126] Its output is: Q t,0 ;

[0127] The input of IMFn is:

[0128] <![CDATA[Q t-2,n ]]> <![CDATA[T 3,t-2 ]]> <![CDATA[w t-2 ]]> <![CDATA[Q t-1,n ]]> <![CDATA[T 3,t-1 ]]> <![CDATA[w t-1 ]]>

[0129] Its output is: Q t,n .

[0130] S2.5. Superimpose the predicted value 0, predicted value 1…predicted value n and the residual to obtain the load prediction value.

[0131] S2.6. Save the model in the above steps.

[0132] S3. Load forecast (taking time t as an example).

[0133] The actual load Q at the previous moment is calculated from the data collected in step S1 t-1 and the actual load Q at the previous two moments t-2 , and the outdoor temperature T at the previous 2 moments 3,t-2 , outdoor humidity w at the previous 2 moments t-2 , outdoor temperature T at the moment before 3,t-1 , outdoor humidity w at the moment before t-1 As an input item, it is input into the LSTM model for prediction to obtain the load forecast value Q at the current moment t .

[0134] like Figure 8 As shown, in another embodiment of the present application, a shopping mall central air-conditioning load forecasting system based on EMD-PSO-LSTM is provided, which includes a data acquisition and calculation module, a model building and training module, and a prediction module;

[0135] The data acquisition and calculation module is used to collect temperature and flow data of the central air-conditioning chilled water system, as well as outdoor temperature and humidity data, to perform load calculation;

[0136] The model building and training module is used to build an LSTM model and perform training based on the EMD decomposition method, and use the PSO method to optimize parameters;

[0137] The prediction module is used to perform load prediction based on historical load and outdoor temperature and humidity data.

[0138] It should be noted here that the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure can be divided into different functional modules to complete all or part of the functions described above. The system is a shopping mall central air-conditioning load prediction method based on EMD-PSO-LSTM applied to the above embodiment.

[0139] like Figure 9 As shown, in another embodiment of the present application, a storage medium is further provided, storing a program, which, when executed by a processor, implements a shopping mall central air conditioning load forecasting based on EMD-PSO-LSTM in the above embodiment, specifically:

[0140] Collect temperature and flow data of the central air-conditioning chilled water system, as well as outdoor temperature and humidity data, to perform load calculations;

[0141] Build an LSTM model and train it based on the EMD decomposition method, and use the PSO method to optimize parameters;

[0142] Load forecasting is performed based on historical load and outdoor temperature and humidity data.

[0143] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0144] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A shopping mall central air conditioning load forecasting method based on EMD-PSO-LSTM, characterized by: The steps include: Collect temperature and flow data of the central air-conditioning chilled water system, as well as outdoor temperature and humidity data, to perform load calculations; Build an LSTM model and train it based on the EMD decomposition method, and use the PSO method to optimize parameters; The construction of the LSTM model is specifically as follows: ； ； ； ； ； ； Among them, W is the weight item, b is the bias term, which is the sigmoid function; the input variable at each moment contains the unit state at the previous moment C t-1 , the intermediate state of the previous moment h t-1 , and the current input x t , the intermediate variables include the output of the forget gate f t , the output of the input-output gate i t and O t , and the output of the input node ; Output variables include unit status C t and intermediate states h t ; The training of the LSTM model is as follows: Using EMD for load series x ( t ) is decomposed to obtain several IMF and a residual; The load, outdoor temperature and outdoor humidity at the previous 2 moments and the previous 1 moment are calculated according to a 、 b 、 c Splice into data sequence; Convert the data series into n ×3×3 format, input to the LSTM model, with the current moment load Q As the predicted value, training is performed; The particle swarm algorithm is used to optimize the learning rate, number of hidden layers, number of nodes per layer, and number of iterations of the LSTM model; For each IMF Make predictions according to the above steps and get predicted values 0 to predicted values n , n for IMF The number of Set the predicted value from 0 to the predicted value n The load forecast value is obtained by superimposing the residual; Save the above steps model; The use of EMD on the load sequence x ( t ) is decomposed into: Find the load sequence x ( t ) all extreme points in ; Use the envelope line to connect all the maximum points to form e max Similarly, connect all the minimum points to form e min ; Find the average of the upper and lower envelopes e mean and subtract it from the original sequence to get the new sequence H : ； H = x ( t )- e mean ； The new sequence is judged according to the following criteria H Is it an eigenmode function? IMF : a) Intrinsic mode function IMF The difference between the number of extreme points and the number of zero points in is no more than 1; b) Eigenmode function at any time IMF The mean of the upper and lower envelopes is 0; c) If not satisfied, H For the new x ( t ), repeat the above steps until the criterion is met IMF 0; Each time the intrinsic mode function is obtained IMF , it will be removed from x ( t ) from the elimination: x ( t )= x ( t )- IMF 0； Repeat the above steps until the remaining residual Res is a monotonic sequence or a constant sequence, that is: x ( t )= IFM 0+ IMF 1+ IMF 2+⋯+ IMFn + Res ； Load forecasting is performed based on historical load and outdoor temperature and humidity data.

2. A shopping mall central air conditioning load forecasting method based on EMD-PSO-LSTM according to claim 1, characterized in that: The temperature and flow data of the central air-conditioning chilled water system are collected as follows: Arrange temperature sensors in the central air-conditioning chilled water system and set the temperature sensor at the water distributor to the water supply temperature. T 1. Set the temperature sensor at the water collector to the return water temperature T 2; Arrange flow sensors in the central air conditioning chilled water system and set the flow at the chilled water main to the chilled water flow rate. q .

3. A method for predicting central air-conditioning load in a shopping mall based on EMD-PSO-LSTM according to claim 2, characterized in that: The load calculation is specifically as follows: Q = q ×( T 2- T 1)× c × ρ ； in, Q is the current moment load, c is the specific heat capacity of chilled water, ρ is the density of frozen water.

4. A method for predicting central air-conditioning load in a shopping mall based on EMD-PSO-LSTM according to claim 1, wherein the particle swarm algorithm is specifically: Initialize each particle randomly; Evaluate each particle and obtain the global optimum; Determine whether the end condition is met, if so, end; If the end condition is not met, the position and velocity of each particle are updated; Update the fitness function of each particle; Update the historical optimal position of each particle; Go back to the step of evaluating each particle and obtaining the global optimum; Determine whether the end condition is met.

5. The method for predicting central air-conditioning load in a shopping mall based on EMD-PSO-LSTM according to claim 1, characterized in that: The load forecasting based on historical load and outdoor temperature and humidity data is specifically as follows: The actual load at the previous moment is calculated based on the historically collected data Q t-1 and the actual load at the previous two moments Q t-2 , and the outdoor temperature at the previous two moments T 3,t-2 , Outdoor humidity 2 hours ago w t-2 , outdoor temperature at the previous moment T 3,t-1 , outdoor humidity at the previous moment w t-1 As an input item, it is input into the LSTM model for prediction to obtain the load forecast value at the current moment Q t .

6. A shopping mall central air conditioning load forecasting system based on EMD-PSO-LSTM, characterized by: A method for predicting central air-conditioning load in a shopping mall based on EMD-PSO-LSTM, applied to any one of claims 1-5, comprising a data acquisition and calculation module, a model building and training module, and a prediction module; The data acquisition and calculation module is used to collect temperature and flow data of the central air-conditioning chilled water system, as well as outdoor temperature and humidity data, to perform load calculation; The model building and training module is used to build an LSTM model and perform training based on the EMD decomposition method, and use the PSO method to optimize parameters; The prediction module is used to perform load prediction based on historical load and outdoor temperature and humidity data.

7. A storage medium storing a program, characterized in that: When the program is executed by the processor, the shopping mall central air-conditioning load forecasting method based on EMD-PSO-LSTM described in any one of claims 1 to 5 is implemented.

Images