Temperature and humidity prediction method for solution dehumidification air conditioning system and application of temperature and humidity prediction method

Abstract

The invention provides a solution dehumidification air conditioning system temperature and humidity prediction method and application thereof, and the method comprises the steps: constructing a hybrid neural network air temperature and humidity prediction model, collecting the operation data of a solution dehumidification air conditioning system, and training the hybrid neural network air temperature and humidity prediction model until the model is stable; acquiring real operation data of the solution dehumidification air conditioning system, and inputting the real operation data into the hybrid neural network air temperature and humidity prediction model to obtain a prediction result; the method is applied to temperature and humidity prediction of the solution dehumidification air-conditioning system. Local features and time dependence in a sequence are extracted through convolution operation by using a time convolution network, data are processed on a higher level by using a bidirectional gating loop unit, the front-back relation and remote dependence of the sequence are captured, key features in the sequence are identified by using a self-attention mechanism module, and the self-attention mechanism module is used for identifying the key features in the sequence. Different weights are distributed for each time step, and hyper-parameters most suitable for the model are found out through an ITSO algorithm; the model has higher prediction precision and lays a good foundation for subsequent real-time control.

Description

Technical Field

[0001] The present invention relates to the technical field of calculation, estimation or counting, and in particular to a method for predicting temperature and humidity of a solution dehumidification air-conditioning system based on the fusion of ITSO and TCN-BiGRU-SA and an application thereof. Background Art

[0002] According to relevant statistics, the building industry plays an important role in the energy field, and its energy consumption accounts for 40% of the world's total energy consumption. In fact, among all types of loads in buildings, the energy consumption of air conditioning systems accounts for almost 40% to 50% of the total energy consumption. Therefore, improving the energy-saving efficiency of building air conditioning systems is crucial to reducing global energy consumption.

[0003] In recent years, the solution dehumidification air conditioning system has attracted widespread attention due to its excellent dehumidification capacity and low energy consumption. It is an active dehumidification air treatment technology that uses dehumidification solution as a desiccant to adjust the air humidity and water as a refrigerant to adjust the air temperature. It utilizes the physical properties of the dehumidifier and achieves efficient dehumidification above the dew point temperature through innovative solution dehumidification and regeneration methods.

[0004] In order to achieve more accurate and efficient real-time control of the liquid dehumidification air-conditioning system and improve energy efficiency, it is particularly important to establish an accurate temperature and humidity prediction model. However, due to the complex internal structure of the liquid dehumidification air-conditioning system and the strong coupling and hysteresis between the refrigeration and dehumidification processes, there are many challenges in establishing a real-time prediction model for the heat and mass transfer performance of the system.

[0005] In the prior art, the heat and mass transfer models of liquid dehumidification air conditioning mainly include finite difference models, efficiency-transfer unit number models, neural network models and other models. In recent years, neural network modeling methods have been widely used, but these methods only rely on the input parameters of the system to predict the temperature and humidity of the system output air. Ibrahim Oleolo and other researchers used the Long Short-Term Memory (LSTM) model to predict the indoor air temperature in a multi-loop central air conditioning system. The results showed that the LSTM model performed better than the Recurrent Neural Network (RNN) model in terms of evaluation indicators such as R², mean square error and mean absolute error. In order to further reduce the number of parameters, Cho et al. proposed a simplified LSTM network that only retains two gate structures, namely the Gated Recurrent Unit (GRU) network. In addition, they innovatively developed a GRU neural network with a hierarchical attention mechanism for indoor temperature prediction. The simulation results show that the prediction accuracy of the algorithm is as high as 98.4%, significantly exceeding the performance of other methods such as support vector machines, RNNs and LSTMs. Li et al. modeled and predicted the indoor air temperature in a variable air volume air conditioning system based on a time delay neural network (TDNN) and an extended neural network (ENN). The results showed that ENN has more advantages in predicting indoor air temperature because its network structure is simpler, occupies less storage space, and has higher prediction accuracy. In order to improve the prediction accuracy, more and more hybrid models combined with multiple deep learning modules have been widely used in the prediction field. For example, Elmaz et al. designed a CNN-LSTM model for indoor temperature prediction, in which convolutional neural networks (CNN) are used as local feature extractors. The extracted features are then input into the LSTM pipeline for multi-step temperature prediction. Experimental results show that the model is stable during a prediction period of up to 2 hours, with an R² value of more than 0.9.

[0006] However, the above models also have some limitations. The hybrid model contains multiple adjustable hyperparameters, which will have a direct impact on the prediction results. How to find the most suitable hyperparameters for the model is a key issue. Therefore, integrating optimization algorithms into hybrid models is also the current mainstream trend. In order to optimize the air-conditioning system of residential buildings, Guo et al. also considered the feasibility of optimizing the MLP model with four metaheuristic algorithms, including salp swarm optimization (SS), whale optimization (WO), spotted hyena optimization (SHO) and wind-driven optimization (WDO). The results showed that the accuracy of heating load prediction increased by 83.85% to 92.48%, and the accuracy of cooling load prediction increased by 60.61% to 89.29%.

[0007] It can be seen that the pursuit of prediction accuracy has always been pursued in the field of temperature and humidity prediction of liquid dehumidification air-conditioning systems. Summary of the invention

[0008] Based on this, the present invention provides a solution dehumidification air-conditioning system temperature and humidity prediction method and application thereof which integrates an improved Tuna Swarm Optimization (ITSO), a temporal convolutional network (TCN), a bidirectional gated recurrent unit (BiGRU) and a self-attention mechanism (SA).

[0009] The technical solution adopted by the present invention is a method for predicting temperature and humidity of a liquid dehumidification air conditioning system, wherein the method constructs a hybrid neural network air temperature and humidity prediction model, collects operation data of the liquid dehumidification air conditioning system, and trains the hybrid neural network air temperature and humidity prediction model until it is stable;

[0010] The actual operation data of the liquid dehumidification air conditioning system is collected and input into this hybrid neural network air temperature and humidity prediction model to obtain the prediction results.

[0011] Preferably, the hybrid neural network air temperature and humidity prediction model includes a sequentially connected temporal convolutional network, a bidirectional gated recurrent unit, a self-attention mechanism module and a fully connected layer; an ITSO optimization module is arranged between the output end of the fully connected layer and the input end of the temporal convolutional network.

[0012] Preferably, the temporal convolutional network comprises a plurality of residual modules connected in sequence; any of the residual modules comprises an input layer, and the output end of the input layer is respectively connected to the CNN branch and the causal convolution branch;

[0013] The CNN branch includes a convolutional layer;

[0014] The causal convolution branch includes two connected dilated causal convolution layers, and each of the dilated causal convolution layers is sequentially connected with a corresponding weight normalization layer, an activation function layer and a Dropout layer;

[0015] The output of the second Dropout layer is added to the output of the convolutional layer and then output to the output layer.

[0016] Compared with the traditional one-dimensional convolutional neural network CNN, TCN combines causal convolution, dilated convolution, and residual connection, and is better than traditional CNN in time series prediction. Compared with the traditional model CNN-LSTM, the accuracy of the present invention is better.

[0017] Preferably, the dilated causal convolution layer is a composite of causal convolution and dilated convolution; wherein each time step of the causal convolution output depends only on the current and past inputs, and the dilated convolution inserts a dilation factor between the elements of the convolution kernel, thereby increasing the receptive field of the model without increasing the number of parameters. The composite here integrates the advantages of the two convolutions to optimize the model performance.

[0018] Preferably, the tuna algorithm optimization module generates an initial population with an improved Circle chaos map, introduces a Levy operator to modify the population update strategy, and obtains the optimal (hyper) parameters of the model based on an improved weight adjustment system.

[0019] Preferably, the collected operation data of the solution dehumidification air-conditioning system includes a data group corresponding to temperature and a data group corresponding to humidity; all the data are normalized.

[0020] Preferably, the data set corresponding to temperature includes the inlet air temperature of the system , Inlet air humidity , Inlet air mass flow , Cooling water temperature , cooling water mass flow , Dehumidifier solution temperature , Dehumidifier solution mass flow rate and system outlet temperature .

[0021] Preferably, the data set corresponding to humidity includes the inlet air temperature of the system , Inlet air humidity , Inlet air mass flow , Cooling water temperature , cooling water mass flow , Dehumidifier solution temperature , Dehumidifier solution mass flow rate , Dehumidifier solution concentration And system outlet humidity .

[0022] Preferably, the prediction accuracy of the hybrid neural network air temperature and humidity prediction model corresponding to temperature and humidity is evaluated respectively.

[0023] An application of the method for predicting temperature and humidity in a liquid dehumidification air-conditioning system is applied to predicting temperature and humidity in a liquid dehumidification air-conditioning system.

[0024] The present invention provides a method for predicting the temperature and humidity of a liquid dehumidification air-conditioning system and its application, constructing a hybrid neural network air temperature and humidity prediction model, collecting operating data of the liquid dehumidification air-conditioning system and training the hybrid neural network air temperature and humidity prediction model until it is stable; collecting real operating data of the liquid dehumidification air-conditioning system, inputting the hybrid neural network air temperature and humidity prediction model, and obtaining prediction results; the method is applied to predicting the temperature and humidity of a liquid dehumidification air-conditioning system.

[0025] The beneficial effect of the present invention is that a hybrid model is provided, which utilizes a temporal convolutional network to extract local features and time dependencies in a sequence by convolution operations, processes data at a higher level by a bidirectional gated recurrent unit, and further captures the context and long-range dependencies of the sequence, identifies key features in the sequence by a self-attention mechanism module, assigns different weights to each time step, and uses an ITSO algorithm to find the most suitable hyperparameters for the model; by evaluating the model of the present invention and comparing it with other models, the results show that the model of the present invention has higher prediction accuracy, which lays a good foundation for subsequent real-time control. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the method of the present invention;

[0027] Figure 2 It is a schematic diagram of the model structure of the present invention;

[0028] Figure 3 Schematic diagram of the structure of the residual module of the temporal convolutional network in the model of the present invention;

[0029] Figure 4 Schematic diagram of the structure of a bidirectional gated recurrent unit in the model of the present invention;

[0030] Figure 5 Schematic diagram of the structure of the self-attention mechanism module in the model of the present invention;

[0031] Figure 6 is a system principle diagram of the present invention in an embodiment;

[0032] Figure 7 A prediction curve for predicting temperature using a model in an embodiment;

[0033] Figure 8 It is a prediction curve of humidity predicted by the model in the embodiment. DETAILED DESCRIPTION

[0034] The present invention is further described in detail below in conjunction with embodiments, but the protection scope of the present invention is not limited thereto.

[0035] like Figure 1 As shown, the present invention provides a method for predicting temperature and humidity of a liquid dehumidification air conditioning system, wherein the method constructs a hybrid neural network air temperature and humidity prediction model, collects operation data of the liquid dehumidification air conditioning system, and trains the hybrid neural network air temperature and humidity prediction model until it is stable;

[0036] The actual operation data of the liquid dehumidification air conditioning system is collected and input into this hybrid neural network air temperature and humidity prediction model to obtain the prediction results.

[0037] The method of the present invention mainly includes model building, data collection and prediction based on the model.

[0038] 1. Model building

[0039] like Figure 2 As shown, the hybrid neural network air temperature and humidity prediction model includes a sequentially connected temporal convolutional network TCN, a bidirectional gated recurrent unit BiGRU, a self-attention mechanism module SA and a fully connected layer; a tuna algorithm optimization module is set between the output end of the fully connected layer and the input end of the temporal convolutional network.

[0040] Among them, TCN uses convolution operation to extract local features and time dependencies in the sequence; BiGRU processes data at a higher level to capture the previous and next relationships and long-range dependencies of the sequence; SA is used to identify key features in the sequence and assign different weights to each time step. By defining the network structure such as the number of layers of the network and the number of neurons in each layer. The present invention uses the Adam optimizer, and the activation function uses the Adam optimizer, the activation function uses the ReLU function, the single-head self-attention mechanism, the convolution kernel size is set to 5, the Dropout coefficient is 0.1, and the maximum number of training times is 100 times; after the model training is completed, it is deployed in the operating environment of the actual solution dehumidification air-conditioning system to improve the prediction accuracy of temperature and humidity, and provide a basis for subsequent real-time control.

[0041] Specifically, the temporal convolutional network (TCN) includes several residual modules connected in sequence. Based on the traditional one-dimensional CNN, it combines causal convolution, dilated convolution and residual connection. It captures complex patterns and long-term dependencies in time series data by combining the efficiency of convolution, causal and dilated convolution design. Compared with recurrent neural networks, it can solve the problems of gradient disappearance and explosion, reflecting the superior prediction ability of TCN.

[0042] TCN introduces residual modules to avoid information loss and instability caused by excessive network depth. The core idea is to introduce one or more layers of "skip connection" operations.

[0043] like Figure 3 As shown in Figure 1, each residual module includes two dilated causal convolution layers and a nonlinear mapping (activation function layer, generally ReLu). In the causal convolution branch, the weight norm (weight normalization layer) and the Dropout layer are combined to regularize the network. In order to correct the difference in input and output widths, an additional one-dimensional convolution layer (CNN branch) is used to ensure that the two tensors involved in the summation operation maintain consistent shapes. The output of each residual module is , x is the input;

[0044] The dilated causal convolution layer is a combination of causal convolution and dilated convolution.

[0045] In the present invention, causal convolution is a time-constrained model that satisfies,

[0046]

[0047] in, is the predicted probability, is the total time, It is an orthogonal operation; the output sequence of TCN has the same length as the input sequence, and the output of the network is only related to the previous input, avoiding future information leakage; and unlike ordinary convolution, dilated convolution increases the dilation rate of the convolution layer exponentially, enabling TCN to achieve the largest possible perception field with fewer network layers.

[0048] like Figure 4 As shown in the figure, the BiGRU layer is connected to the end of the TCN layer, and a gating mechanism is used to control the flow of information. The input sequence is processed through two independent GRU layers (one processes from front to back, i.e., forward GRU, and the other processes from back to front, i.e., backward GRU), capturing the before-and-after relationship and long-range dependency of time series data, so that the model can understand the time series information more comprehensively. The gated recurrent unit GRU is a recurrent neural network, a variant of LSTM, which is proposed to solve problems such as long-term memory and gradient in back propagation. Compared with LSTM, GRU has fewer parameters and almost the same accuracy.

[0049] like Figure 5 As shown in the figure, the self-attention mechanism SA identifies important features of the time series, assigns different weights to each time step, helps the model focus on important information at different time scales, alleviates problems such as gradient disappearance in long sequences, and selectively focuses on specific time steps based on current input and past information, enabling the model to better understand the previous and next relationships during prediction or classification tasks.

[0050] In the present invention, the tuna algorithm optimization module ITSO uses an improved Circle chaotic map to replace the randomly generated initial population in the original algorithm. The formula is as follows:

[0051]

[0052] in, It is A chaotic particle, It is chaotic particles; the improved Circle chaotic mapping operator is used to modify TSO to obtain a more uniform candidate solution. At the beginning of the algorithm, tuna individuals are evenly distributed in the search space, which significantly increases the population diversity of TSO;

[0053] Furthermore, ITSO also introduces the Levy operator to modify the group update strategy of TSO. The mathematical model of the improved spiral foraging strategy based on the Levy operator and nonlinear adaptive weight strategy is as follows:

[0054]

[0055] Among them, t and t max Represent the number of iterations and the maximum number of iterations respectively, is the current best individual position, is the position of the random individual after t iterations, and are the positions of the i-th and i-1-th individuals after t iterations, is the distance parameter between the best individual position and the i-th individual position after t iterations; an improved weight adjustment system is established, and is the weight coefficient, satisfying,

[0056]

[0057]

[0058] in, , express The initial value of express The final value of express The initial value of express The final value of .

[0059] In the present invention, the ITSO optimization algorithm is finally introduced into the TCN-BiGRU-SA model, and the learning rate, TCN expansion coefficient, number of filters, number of BiGRU hidden layer neurons, key value of the self-attention mechanism and regularization parameters that have a greater impact on the model are optimized to find the most suitable parameters for the model, so that the hybrid prediction model has higher prediction accuracy.

[0060] 2. Data collection and prediction based on models

[0061] The collected operation data of the liquid dehumidification air conditioning system includes a data group corresponding to temperature and a data group corresponding to humidity; all the data are normalized.

[0062] The temperature data set includes the system inlet air temperature , Inlet air humidity , Inlet air mass flow , Cooling water temperature , cooling water mass flow , Dehumidifier solution temperature , Dehumidifier solution mass flow rate and system outlet temperature .

[0063] The data set corresponding to humidity includes the system inlet air temperature , Inlet air humidity , Inlet air mass flow , Cooling water temperature , cooling water mass flow , Dehumidifier solution temperature , Dehumidifier solution mass flow rate , Dehumidifier solution concentration And system outlet humidity .

[0064] In the present invention, sensors are used to collect real-time operating data of the solution dehumidification air-conditioning system during the refrigeration and dehumidification process. Since the temperature and humidity of the air at the system outlet are affected by multiple factors, the temperature and humidity are predicted separately according to the characteristics of the heat and mass transfer model.

[0065] In the sample data finally obtained, generally speaking, the interval time between each group of sample data is 10 seconds, and the training set and test set are divided into 8:2; in order to eliminate the dimension of the data and improve the calculation efficiency, the sample data is normalized and then input into the model for training and testing.

[0066] The prediction accuracy of the hybrid neural network air temperature and humidity prediction model corresponding to temperature and humidity is evaluated respectively.

[0067] Specifically, the prediction accuracy of the model for temperature and humidity is evaluated on the test set, and the root mean square error, mean square error, mean absolute error, mean relative percentage error, and determination coefficient are generally used for evaluation.

[0068] The present invention also relates to an application of the method for predicting temperature and humidity of a liquid dehumidification air-conditioning system, which is applied to predicting temperature and humidity of a liquid dehumidification air-conditioning system.

[0069] like Figure 6 As shown, an embodiment of the present invention is provided.

[0070] (1) According to the design, a liquid dehumidification air conditioning system experimental platform was built. The experimental platform used lithium chloride aqueous solution as the dehumidifier and mainly included cooling coils, dehumidifiers, coolers, regenerators, heaters, heat pipe heat exchangers, data acquisition and control systems, etc.;

[0071] (2) Use sensors to collect real-time operating data of the liquid dehumidification air conditioning system during the refrigeration and dehumidification process, including the flow rate, temperature and humidity of the air and the flow rate and temperature of the dehumidifier solution. Since the temperature and humidity of the air at the outlet of the system are affected by multiple factors, the temperature and humidity are predicted separately according to the characteristics of the heat and mass transfer model; the temperature model selects the system inlet air temperature and humidity and mass flow , Cooling water temperature With mass flow , Dehumidifier solution temperature and mass flow As input, the system outlet temperature As output; humidity model selects system inlet air temperature and humidity and mass flow , Cooling water temperature With mass flow , Dehumidifier solution temperature With mass flow and concentration As input, the system outlet humidity As output;

[0072] The experimental platform measured a total of 3268 sets of sample data, with an interval of 10 seconds between each set of sample data. The data was divided into 8:2, with the first 80% as the training set and the last 20% as the test set. The sample data was normalized.

[0073] (3) Using the training set of the processed 3268 sets of data as input, the hybrid neural network air temperature and humidity prediction model was trained;

[0074] (4) The prediction accuracy of the model for temperature and humidity was evaluated on the test set using the following five evaluation indicators:

[0075] Root Mean Square Error: ;

[0076] Mean Square Error: ;

[0077] Mean absolute error: ;

[0078] Average relative percentage error: ;

[0079] Coefficient of determination: ;

[0080] in, is the actual value, is the predicted value;

[0081] Through experimental verification, the above embodiment is applied to predict the temperature and humidity of the liquid dehumidification air conditioning system through the above steps, and the following is obtained: Figure 7 , Figure 8 The results of the evaluation model for temperature and humidity prediction accuracy are as follows:

[0082] Temperature prediction accuracy , , , , ;

[0083] Humidity prediction accuracy , , , , ;

[0084] The results show that the hybrid neural network air temperature and humidity prediction model of the present invention has strong applicability and high accuracy;

[0085] (5) Save and output the prediction results to provide a basis for subsequent system control and energy consumption optimization.

[0086] The present invention also relates to a computer-readable storage medium in its application, on which a temperature and humidity prediction program for a liquid dehumidification air-conditioning system is stored. When the program is executed by a processor, the temperature and humidity prediction method for a liquid dehumidification air-conditioning system is implemented.

[0087] The present invention also relates to a computer device in its application, comprising a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the temperature and humidity prediction method of the solution dehumidification air conditioning system is implemented.

[0088] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0089] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0090] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0091] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1A step that specifies a function in one or more boxes.

[0092] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0093] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for predicting temperature and humidity of a liquid dehumidification air conditioning system, characterized in that: The method constructs a hybrid neural network air temperature and humidity prediction model, collects operation data of a liquid dehumidification air conditioning system, and trains the hybrid neural network air temperature and humidity prediction model until it is stable; The actual operation data of the liquid dehumidification air conditioning system is collected and input into this hybrid neural network air temperature and humidity prediction model to obtain the prediction results.

2. The method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 1, characterized in that: The hybrid neural network air temperature and humidity prediction model includes a sequentially connected temporal convolutional network, a bidirectional gated recurrent unit, a self-attention mechanism module and a fully connected layer; a tuna algorithm optimization module is arranged between the output end of the fully connected layer and the input end of the temporal convolutional network.

3. The method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 2, characterized in that: The temporal convolutional network includes a plurality of residual modules connected in sequence; any of the residual modules includes an input layer, and the output end of the input layer is respectively connected to the CNN branch and the causal convolution branch; The CNN branch includes a convolutional layer; The causal convolution branch includes two connected dilated causal convolution layers, and each of the dilated causal convolution layers is sequentially connected with a corresponding weight normalization layer, an activation function layer and a Dropout layer; The output of the second Dropout layer is added to the output of the convolutional layer and then output to the output layer.

4. The method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 3, characterized in that: The dilated causal convolution layer is a combination of causal convolution and dilated convolution.

5. The method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 2, characterized in that: The tuna algorithm optimization module generates an initial population using an improved Circle chaos map, introduces a Levy operator to modify the population update strategy, and obtains the optimal parameters of the model based on an improved weight adjustment system.

6. The method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 1, characterized in that: The collected operation data of the liquid dehumidification air conditioning system includes a data group corresponding to temperature and a data group corresponding to humidity; all the data are normalized.

7. A method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 6, characterized in that: The temperature data set includes the system inlet air temperature , Inlet air humidity , Inlet air mass flow , Cooling water temperature , cooling water mass flow , Dehumidifier solution temperature , Dehumidifier solution mass flow rate and system outlet temperature .

8. The method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 6, characterized in that: The data set corresponding to humidity includes the system inlet air temperature , Inlet air humidity , Inlet air mass flow , Cooling water temperature , cooling water mass flow , Dehumidifier solution temperature , Dehumidifier solution mass flow rate , Dehumidifier solution concentration And system outlet humidity .

9. The method for predicting temperature and humidity of a liquid dehumidification air conditioning system according to claim 1, characterized in that: The prediction accuracy of the hybrid neural network air temperature and humidity prediction model corresponding to temperature and humidity is evaluated respectively.

10. An application of the method for predicting temperature and humidity in a liquid dehumidification air conditioning system according to any one of claims 1 to 9, characterized in that: Applied to temperature and humidity prediction of liquid dehumidification air conditioning system.

Images