A control method for a rotary dehumidification air conditioning system

Abstract

The application relates to a control method of a rotary dehumidification air conditioning system, which comprises the following steps: obtaining state variables, controllable variables and target variables and a relationship model satisfied by the target variables under typical working conditions; constructing a control model, generating a training set and a verification set by using the relationship model, training and verifying the control model by using the training set and the verification set, and obtaining an optimal control model; adjusting the controllable variables by taking minimization of the target variables as an optimization target, updating parameters in the optimal control model according to a predicted target variable obtained by inputting the adjusted controllable variables and the state variables into the optimal control model and the minimized target variables, and obtaining a final control model; inputting the state variables and the controllable variables in multiple to-be-controlled working conditions into the final control model, selecting a minimum value from target variables corresponding to all the to-be-controlled working conditions, and controlling the rotary dehumidification air conditioning system by using a controllable variable corresponding to the minimum value. The control method can reduce energy consumption.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of air conditioning systems, and in particular to a control method for a rotary dehumidification air conditioning system. BACKGROUND

[0002] With the increasing global energy demand and environmental pressure, energy saving and emission reduction has become the common goal of all industries. As the core component of new energy vehicles, the production process of lithium batteries has extremely harsh requirements on environmental conditions. In particular, humidity control is crucial to the production quality and safety of lithium batteries; excessive humidity can cause battery materials to absorb moisture, thereby affecting the electrochemical performance of the battery, and even possibly causing safety accidents. Therefore, in some key processes of lithium battery production, the environmental humidity must be strictly controlled to ensure that the produced batteries have high capacity, high stability and long life.

[0003] There are many traditional dehumidification technologies, such as refrigeration dehumidification and direct expansion heat pump dehumidification, etc. However, traditional dehumidification technologies often fall short when faced with extremely low humidity requirements. Existing rotary dehumidification technology gradually becomes the mainstream dehumidification solution for lithium battery production workshops, as it can handle air humidity to a dew point temperature of -60℃ or lower. However, the high energy consumption problem of rotary dehumidification systems also becomes prominent. How to reduce energy consumption while ensuring humidity control accuracy has become a difficult problem that needs to be solved in the industry.

[0004] Traditional dehumidification system control methods rely heavily on pre-set parameters based on experience and rules. Although this method is simple and direct, it lacks the ability to respond flexibly to actual operating environments. In actual applications, production environments vary greatly, making it difficult for traditional methods to adapt to complex operating conditions, resulting in low system operating efficiency and high energy consumption. In particular, in the context of lithium battery production, which requires extremely high humidity control accuracy, the limitations of traditional control methods are more apparent. Therefore, further improvements to existing technology are needed. SUMMARY

[0005] The technical problem to be solved by the present application is to provide a control method for a rotary dehumidification air conditioning system that can reduce energy consumption, in view of the above existing technology.

[0006] The technical solution adopted by the present application to solve the above technical problem is as follows: a control method for a rotary dehumidification air conditioning system, characterized by comprising the following steps:

[0007] Step 1: Collecting the working environment parameters of the rotary dehumidification air conditioning system and preprocessing the collected data to form a data set from the preprocessed data;

[0008] Step 2: Cluster analysis of the data set to obtain typical operating conditions of the rotary dehumidification air conditioning system;

[0009] Step 3, parameter extraction and classification are performed on the typical working conditions of the rotary dehumidification air conditioning system in step 2 to obtain state variables x, controllable variables μ and target variables y;

[0010] Step 4, the state variables x, controllable variables μ and target variables y are analyzed to obtain a relationship model satisfied between the state variables x, controllable variables μ and target variables y;

[0011] Step 5, a control model of the rotary dehumidification air conditioning system is constructed, a sample set composed of multiple different working conditions is generated using the relationship model, the sample set is divided into a training set and a verification set, the training set is used to train the constructed control model to obtain a trained control model, and the verification set is used to verify the trained control model to obtain an optimal control model;

[0012] Step 6, under the condition that the state variables x meet the requirements, the controllable variables μ are adjusted with the optimization objective of minimizing the target variables of the rotary dehumidification air conditioning system, the adjusted controllable variables μ and the state variables x are input into the optimal control model to obtain a predicted target variable, and the parameters in the optimal control model are updated according to the difference between the predicted target variable and the minimized target variable to obtain a final control model;

[0013] Step 7, the state variables x and controllable variables μ in multiple to-be-controlled working conditions are input into the final control model to obtain target variables corresponding to each to-be-controlled working condition, and the minimum value is selected from all target variables corresponding to the to-be-controlled working conditions, and then the controllable variables μ corresponding to the minimum value are used to control the rotary dehumidification air conditioning system.

[0014] Preferably, the data collected in step 1 includes at least one or more of humidity, outdoor temperature, outdoor dew point temperature, chilled water temperature, steam flow, cooling capacity, steam temperature and energy consumption of the place where the rotary dehumidification air conditioning system is located.

[0015] In order to eliminate abnormal data caused by data acquisition failure or external interference, the preprocessing process in step 1 includes one or more of outlier detection, missing value filling, data smoothing and standardization processing.

[0016] In order to ensure the scientificity and practicability of the lithium battery production process working condition analysis, the specific process for obtaining the typical working conditions of the rotary dehumidification air conditioning system in step 2 is as follows:

[0017] At least two clustering methods are selected to perform clustering analysis on the data set, the clustering results of each clustering method are compared, the better clustering result is selected as the final analysis result, and the outliers in the final analysis result are deleted, i.e. the typical working conditions of the rotary dehumidification air conditioning system are obtained.

[0018] Preferably, the clustering method is two, respectively K-means clustering method and DBSCAN clustering method.

[0019] Preferably, the specific process in step 3 is: according to the time period corresponding to each cluster typical working condition, all parameters related thereto are extracted; and all extracted parameters are divided into three categories according to functions, i.e. state variables x, controllable variables and target variables y are obtained.

[0020] In the above scheme, the generalized additive model is used in step 4 to analyze the state variable x, the controllable variable μ and the target variable y, and the fitted generalized additive model is obtained.

[0021] It is judged whether the result of the fitted generalized additive model is effective, if yes, the fitted generalized additive model is taken as the relationship model satisfied between the state variable x, the controllable variable μ and the target variable y; if no, the regression analysis method is used to analyze the state variable x, the controllable variable μ and the target variable y, so as to obtain the relationship model satisfied between the state variable x, the controllable variable μ and the target variable y.

[0022] Preferably, the specific process of using the regression analysis method to analyze the state variable x, the controllable variable μ and the target variable y is:

[0023] If the state variable x, the controllable variable μ and the target variable y present a linear relationship, the multiple linear regression method is used to analyze the state variable x, the controllable variable μ and the target variable y;

[0024] If the state variable x, the controllable variable μ and the target variable y present a nonlinear relationship, the multiple nonlinear regression method or the support vector machine regression method is used to analyze the state variable x, the controllable variable μ and the target variable y.

[0025] Preferably, the control model in step 5 adopts a neural network or a multiple regression model.

[0026] Compared with the prior art, the application has the advantages that: by deeply mining the collected data, the rules and patterns hidden in the data can be found, i.e. the relationship model satisfied between the state variable x, the controllable variable μ and the target variable y is obtained, so that the key factors affecting the performance of the rotary dehumidification air conditioning system are extracted, and the control model of the rotary dehumidification air conditioning system is established, and then the control model is optimized and controlled according to the working condition of the rotary dehumidification air conditioning system, so that the energy saving and consumption reduction of the rotary dehumidification air conditioning system are realized. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 The flow chart of the typical working condition acquisition method in the embodiment of the application;

[0028] Figure 2 The flow chart of the relationship model acquisition method in the embodiment of the present application. DETAILED DESCRIPTION

[0029] The present application will be further described in detail below with reference to the embodiments of the drawings.

[0030] The control method of the rotary dehumidification air conditioning system in the embodiment includes the following steps:

[0031] Step 1, collecting the working environment parameters of the rotary dehumidification air conditioning system, and pre-processing the collected data to form a data set;

[0032] The collected data at least includes one or more of the humidity, outdoor temperature, outdoor dew point temperature, chilled water temperature, steam flow, cooling capacity, steam temperature and energy consumption of the place where the rotary dehumidification air conditioning system is located;

[0033] The rotary dehumidification air conditioning system in the embodiment is located in a lithium battery production workshop, and the collected data in the embodiment includes but is not limited to the humidity, outdoor temperature, outdoor dew point temperature, chilled water temperature, steam flow, cooling capacity, steam temperature and energy consumption data of the lithium battery production workshop; the collected data can be recorded at high frequency and continuously through a sensor network and an industrial Internet of Things system to ensure the timeliness and integrity of the data;

[0034] The pre-processing process in the embodiment includes one or more of outlier detection, missing value filling, data smoothing and standardization processing;

[0035] Step 2, performing cluster analysis on the data set to obtain the typical working conditions of the rotary dehumidification air conditioning system;

[0036] The specific process of obtaining the typical working conditions of the rotary dehumidification air conditioning system is as follows:

[0037] At least two clustering methods are selected to perform cluster analysis on the data set, the clustering results of each clustering method are compared, the better clustering result is selected as the final analysis result, and the outliers in the final analysis result are deleted, i.e. the typical working conditions of the rotary dehumidification air conditioning system are obtained;

[0038] As shown in the figure, the clustering methods in the embodiment are two, which are K-means clustering method and DBSCAN clustering method; the K-means clustering method divides the data into several clusters by minimizing the distance from the sample to the cluster center, each cluster represents a typical working condition, and the objective function of the K-means clustering is: Figure 1

[0039]

[0040] Wherein, C​i For the i-th cluster, μ i Let x be the centroid of the cluster and x be the sample point of the cluster. In order to determine the optimal number of clusters k in K-means clustering, the Elbow Method was used. By observing the change of the sum of squared errors under different numbers of clusters, the number of clusters at the inflection point was selected as the optimal number of clusters.

[0041] DBSCAN clustering, a density-based clustering method, effectively handles noisy data and identifies clusters of arbitrary shapes. DBSCAN does not require pre-specifying the number of clusters; instead, it automatically divides high-density regions into clusters and marks low-density points as noise by setting radius and minimum point count parameters. The formula for the noise point condition is:

[0042] Noise point condition = {x1 | MinPts(x1) < ∈}

[0043] Where x1 represents a specific data point in the dataset, ∈ represents the neighborhood radius of a given point x1. Within this radius, the algorithm checks how many other data points fall within this range. MinPts represents the minimum number of data points required within the neighborhood radius ∈. If the number of points within this range is less than MinPts, then the point x1 is marked as a noise point.

[0044] By comparing the clustering results of K-means and DBSCAN clustering methods, we can evaluate their performance in handling data noise and identifying typical working conditions. K-means clustering performs excellently when handling large-scale data, quickly processing and generating multiple clusters, but it is slightly insufficient when dealing with noisy data. Conversely, DBSCAN clustering has significant advantages in handling noise and identifying complex cluster shapes. Ultimately, based on the stability, accuracy, and representativeness of the clustering results, the superior clustering result can be selected as the final analysis result.

[0045] Step 3: Extract and classify the parameters of the typical operating conditions of the rotary dehumidification air conditioning system in Step 2, and obtain the state variable x, the controllable variable μ and the target variable y;

[0046] like Figure 2As shown, the specific process is as follows: for the clustered typical working conditions, first, detailed analysis is performed to check whether there are abnormal values in each type of typical working condition; for the identified abnormal values, removal and deletion should be performed, and then a plurality of final typical working conditions are obtained; after the cleaned typical working conditions are obtained, according to the time period corresponding to each cluster of typical working conditions, all related data are extracted one by one; specifically, all other parameters related to each cluster corresponding time period need to be extracted, including outdoor temperature, outdoor dew point temperature, chilled water temperature, chilled water side power consumption, steam temperature, cooling capacity and steam flow, etc.; and all the extracted parameters are divided into three categories according to functions, that is, state variables x, controllable variables μ and target variables y; the state variables x in this embodiment include outdoor temperature and outdoor dew point temperature, the controllable variables μ include chilled water temperature, cooling capacity and steam temperature; and the target variables y include chilled water side power consumption and steam flow;

[0047] Step 4, analyze the state variables x, controllable variables μ and target variables y to obtain a relationship model satisfied between the state variables x, controllable variables μ and target variables y;

[0048] In this embodiment, the state variables x, controllable variables μ and target variables y are analyzed by using a generalized additive model to obtain a fitted generalized additive model;

[0049] It is judged whether the result of the fitted generalized additive model is valid, if yes, the fitted generalized additive model is taken as the relationship model satisfied between the state variables x, controllable variables μ and target variables y; if no, regression analysis method is used to analyze the state variables x, controllable variables μ and target variables y to obtain a relationship model satisfied between the state variables x, controllable variables μ and target variables y;

[0050] The generalized additive model in this embodiment is a prior art, which will not be described here. The generalized additive model can capture linear and nonlinear relationships at the same time without preassuming the relationship form of data, so the generalized additive model is an ideal choice for preliminary analysis. If the analysis result shows good fitting effect, the result can be directly used to explain the relationship between variables. However, if the result is not ideal and cannot accurately capture the pattern in data, a suitable analysis method needs to be further selected according to the relationship (linear or nonlinear) of data;

[0051] The specific process of using regression analysis method to analyze the state variables x, controllable variables μ and target variables y is as follows:

[0052] If the state variables x, controllable variables μ and target variables y present a linear relationship, multivariate linear regression method is used to analyze the state variables x, controllable variables μ and target variables y;

[0053] If the state variable x, the controllable variable μ and the target variable y present a nonlinear relationship, a multivariate nonlinear regression method or a support vector machine regression method is used to analyze the state variable x, the controllable variable μ and the target variable y;

[0054] This step-by-step analysis method ensures a comprehensive understanding of the variable relationship, which can more effectively capture and explain the complex relationship between x-μ-y by first exploring the possible relationship form using the generalized additive model and then selecting other methods according to the data characteristics, thereby providing a solid theoretical basis for the optimization of production processes and the formulation of control strategies;

[0055] Step 5, a control model of the rotary dehumidification air conditioning system is constructed, a sample set composed of multiple different working conditions is generated using the relationship model, the sample set is divided into a training set and a validation set, the constructed control model is trained using the training set to obtain a trained control model, and the trained control model is verified using the validation set to obtain an optimal control model;

[0056] In this embodiment, the state variables and control variables under different typical working conditions are parameterized and adjusted to obtain the changes of the target variables such as energy consumption and humidity control effect under specific environmental and control conditions. For each typical working condition, the control model can accurately reproduce the energy consumption-control-environment interaction relationship in a specific season and time, reflecting the energy consumption variation law of the system under different conditions. The control model not only has strong robustness, can adapt to different working conditions, and maintains stable prediction accuracy, but also can realize effective energy consumption optimization and regulation in the production process. For each working condition, the control model uses a neural network or a multivariate regression model. The input variables include state variables (outdoor temperature and outdoor dew point temperature) and controllable variables (chilled water temperature, cooling capacity and steam temperature), and the output variables are target variables (chilled water side power consumption and steam flow);

[0057] The constructed control model is trained and verified through the training set and the validation set, and the control model can accurately predict the energy consumption under each working condition. When further optimized, techniques such as Bayesian optimization and genetic algorithm can be used for parameter tuning to improve the generalization ability of the control model and the prediction accuracy under multiple working conditions;

[0058] Step 6, under the condition that the state variable x meets the requirements, the controllable variable μ is adjusted to minimize the target variable of the rotary dehumidification air conditioning system, and the adjusted controllable variable μ and the state variable x are input into the optimal control model to obtain the predicted target variable. The parameters in the optimal control model are updated according to the difference between the predicted target variable and the minimized target variable to obtain the final control model;

[0059] The optimization target in this embodiment is:

[0060]

[0061] T chill is the chilled water temperature, T steam is the steam temperature; E(T chill , T steam ) represents the power consumption of the rotary dehumidification air conditioning system under the chilled water temperature and steam temperature setting conditions;

[0062] Constraints:

[0063] H indoor = H(T chill , T steam , x)

[0064] H indoor ≤ H min

[0065] wherein x is the outdoor temperature, H(T chill , T lteam , x) represents the indoor relative humidity under the outdoor temperature, chilled water temperature and steam temperature setting conditions; H min represents the target humidity of the lithium battery production workshop under the target scenario;

[0066] When the controllable variable μ is adjusted, the above constraints need to be met; in actual operation, a balance point between energy consumption and humidity control effect is found through a multi-objective optimization algorithm or a heuristic search method; since the humidity control effect is closely related to the adjustment of the chilled water temperature and the steam temperature, and the adjustment of these parameters directly affects the energy consumption of the rotary dehumidification air conditioning system, the chilled water temperature and the steam temperature of different parameters need to be constantly adjusted through the control model to gradually approach the optimal balance point between energy consumption and humidity control under the target scenario;

[0067] Step 7, input the state variable x and the controllable variable μ in multiple to-be-controlled working conditions into the final control model to obtain the target variable corresponding to each to-be-controlled working condition, and select the minimum value from the target variables corresponding to all to-be-controlled working conditions, then use the controllable variable μ corresponding to the minimum value to control the rotary dehumidification air conditioning system.

[0068] In this embodiment, the control model serves as a virtual test platform to test different regulation schemes in a simulated environment, thereby reducing the trial-and-error cost of the actual system, and finding a regulation combination that meets the humidity requirement and has the lowest energy consumption through iteration; in addition, when the working condition changes, the control model can predict the influence of different environments and regulation conditions on humidity control and energy consumption in real time, automatically adjusts the chilled water temperature and the steam temperature through dynamic programming and reinforcement learning algorithm to adapt to environmental changes and uncertainties, and realizes the optimal dynamic balance between energy consumption and humidity control.

[0069] The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, several improvements and refinements can be made without departing from the technical principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A control method for a rotary dehumidification air conditioning system, characterized in that... Includes the following steps: Step 1: Collect the working environment parameters of the rotary dehumidification air conditioning system, preprocess the collected data, and construct a dataset from the preprocessed data. Step 2: Perform cluster analysis on the dataset to obtain typical operating conditions of the rotary dehumidifier air conditioning system; Step 3: Extract and classify parameters for typical operating conditions of the rotary dehumidification air conditioning system in Step 2, and obtain state variables. χ Controllable variables μ and target variable y ; Based on the time period corresponding to each cluster of typical operating conditions, all relevant data were extracted one by one. The extracted parameters included outdoor temperature, outdoor dew point temperature, chilled water temperature, chilled water side power consumption, steam temperature, cooling capacity, and steam flow rate. These extracted parameters were then categorized into three types according to function: state variables. χ Including outdoor temperature and outdoor dew point temperature, controllable variables μ This includes chilled water temperature, cooling capacity, and steam temperature; Target variable y This includes power consumption on the chilled water side and steam flow rate; Step 4: Adjust the state variables χ Controllable variables μ and target variable y Perform analysis to obtain state variables χ Controllable variables μ and target variable y The relationship model satisfied between them; Step 5: Construct a control model for the rotary dehumidification air conditioning system. Use a relational model to generate a sample set, which consists of multiple different operating conditions. Divide the sample set into a training set and a validation set. Use the training set to train the constructed control model to obtain the trained control model. Use the validation set to validate the trained control model to obtain the optimal control model. Step 6, in the state variables χ Under the condition of meeting the requirements, the optimization objective is to minimize the target variable of the rotary dehumidification air conditioning system, and the controllable variables are optimized accordingly. μ Make adjustments and change the controllable variables. μ and state variables χ The parameters in the optimal control model are updated based on the difference between the predicted target variable and the minimized target variable, thus obtaining the final control model. Step 7: Transfer the state variables from multiple controllable operating conditions. χ and controllable variables μ The input is fed into the final control model to obtain the target variable corresponding to each controllable condition. The minimum value is then selected from the target variables corresponding to all controllable conditions, and the controllable variable corresponding to this minimum value is used. μ Control the rotary dehumidification air conditioning system.

2. The control method according to claim 1, characterized in that: The data collected in step 1 includes at least one or more of the following: humidity of the location where the rotary dehumidifying air conditioning system is located, outside temperature, outside dew point temperature, chilled water temperature, steam flow rate, cooling capacity, steam temperature, and energy consumption.

3. The control method according to claim 1, characterized in that: The preprocessing process in step 1 includes one or more of outlier detection, missing value imputation, data smoothing, and standardization.

4. The control method according to any one of claims 1 to 3, characterized in that: The specific process for obtaining the typical operating conditions of the rotary dehumidification air conditioning system in step 2 is as follows: At least two clustering methods are selected to perform cluster analysis on the dataset. The clustering results of each method are compared, and the better clustering result is selected as the final analysis result. Outliers in the final analysis result are deleted, thus obtaining the typical operating conditions of the rotary dehumidifying air conditioning system.

5. The control method according to claim 4, characterized in that: The clustering methods are of two types: K-means clustering and DBSCAN clustering.

6. The control method according to claim 1, characterized in that: In step 4, a generalized additive model is used to analyze the state variables. χ Controllable variables μ and target variable y Perform analysis to obtain the fitted generalized additive model; Determine whether the fitted generalized additive model is valid. If so, use the fitted generalized additive model as the state variable. χ Controllable variables μ and target variable y The relationship model between them must be satisfied; if not, regression analysis should be used to analyze the state variables. χ Controllable variables μ and target variable y Perform analysis to obtain state variables χ Controllable variables μ and target variable y The relationship model satisfied between them.

7. The control method according to claim 6, characterized in that: Using regression analysis on state variables χ Controllable variables μ and target variable y The specific process of analysis is as follows: If state variables χ Controllable variables μ and target variable y If a linear relationship is observed, then multiple linear regression is used to analyze the state variables. χ Controllable variables μ and target variable y Perform analysis; If state variables χ Controllable variables μ and target variable y If a nonlinear relationship is observed, then multiple nonlinear regression or support vector machine regression methods should be used to analyze the state variables. χ Controllable variables μ and target variable y Perform the analysis.

8. The control method according to claim 1, characterized in that: The control model in step 5 uses a neural network or a multiple regression model.

Images