Air conditioner terminal temperature and humidity data splitting method and system based on K-means
By using a K-means-based data splitting method, combined with professional knowledge and machine learning models, the problem of poor data quality at the air conditioning terminal was solved, achieving efficient and accurate indoor temperature and humidity data splitting and improving the data support capability of air conditioning energy-saving control algorithms.
Patent Information
- Application Number
- CN202510973127.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-15
- Publication Date
- 2025-11-21
AI Technical Summary
In existing technologies, the quality of air conditioning terminal data is poor, with the presence of unlabeled data and a large amount of manually modified erroneous data, which affects the development of global optimization energy-saving algorithms and the effectiveness of energy-saving analysis.
The K-means-based data splitting method is adopted, which includes data cleaning, integration, visualization, aggregation and splitting steps. Combining HVAC expertise and machine learning models, effective indoor temperature and humidity data are identified and split through cluster similarity calculation.
It significantly improves the accuracy and rationality of end-point data, and can accurately separate indoor temperature and humidity data from mixed data, improving the accuracy of data processing by about 60%, and providing high-quality data support for air conditioning energy-saving control algorithms.
Smart Images
Figure CN120995141A_ABST
Abstract
Description
Technical Field
[0001] This invention specifically relates to a method and system for splitting temperature and humidity data of air conditioning terminals based on K-means, belonging to the field of building energy conservation technology. Background Technology
[0002] Building operation energy consumption accounts for approximately 21.3% of the national total energy consumption, while central air conditioning in large public buildings accounts for 50%-70% of the building operation energy consumption in summer, making it a key area for building energy conservation and carbon reduction. As the main energy-consuming component of central air conditioning systems in large buildings, the central air conditioning cooling source system accounts for more than half of the total air conditioning energy consumption. Its equipment is relatively centralized, easy to manage and control, and possesses significant energy-saving potential. Optimizing its energy-saving performance is crucial for saving building energy consumption. Because the central air conditioning cooling source system is a highly nonlinear system, traditional local optimization energy-saving control methods cannot fully tap its energy-saving potential. However, global optimization energy-saving control methods based on deep learning and reinforcement learning models can establish the intrinsic relationship between various operating parameters and equipment and cooling source system energy consumption. Starting from overall operational energy efficiency, they optimize controllable parameters of the equipment, enabling the equipment's operating status to dynamically change with the load demand of end users, thereby further improving the operational energy efficiency of the central air conditioning system while reducing reliance on the experience of building engineer (BA) engineers.
[0003] Deep learning and reinforcement learning technologies rely on a large amount of high-quality critical operational data from air conditioning systems. This critical operational data consists of water system data and terminal data. Terminal data includes indoor temperature and humidity, supply air temperature and humidity, and return air temperature and humidity. Indoor temperature and humidity data is often used to assess indoor thermal comfort levels and is crucial data for developing energy-saving algorithms. However, in engineering practice, due to factors such as cost control by general contractors and the professional capabilities of property management companies, the collected terminal data often lacks clear labels (e.g., temperature 1, humidity 2), and BA engineers arbitrarily modify the terminal data. This makes it impossible to distinguish indoor environmental parameters, supply air parameters, or return air parameters from the air conditioning terminal data, and the terminal data contains a large amount of erroneous data generated by human modification. This severely affects the quality of the terminal data and prevents the provision of usable high-quality indoor temperature and humidity data for the development of global optimization energy-saving algorithms or energy-saving analysis for air conditioning systems. Summary of the Invention
[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a K-means-based method for splitting temperature and humidity data at air conditioning terminals, as well as a system and electronic equipment. This method enables the identification and splitting of valid indoor temperature and humidity data from unlabeled terminal datasets containing a large amount of abnormal data, thereby solving the problem of poor terminal data quality that restricts the development, application, and energy-saving analysis of air conditioning energy-saving control algorithms.
[0005] This invention discloses a method for splitting temperature and humidity data at air conditioning terminals based on K-means, comprising the following steps: (1) data cleaning based on the term_data_preprocess method; (2) data integration based on the term_data_merge method; (3) data visualization; and (4) data aggregation and splitting. Step (4) includes: selecting a temperature dataset; transforming and standardizing the data dimensions; selecting a model and training it; adding clustering results to the temperature dataset; and then adding labels to the clustering results for further filtering, splitting the temperature dataset to extract the room temperature item and the temperature and humidity dataset to extract the indoor relative humidity item.
[0006] Furthermore, step (1) includes deleting noisy data and missing values contained in the original dataset or generated during data structure transformation by calling the drop method of the DataFrame and the columns attribute in the term_data_preprocess method, and filling missing values by calling the backfill and forwardfill methods of the pandas library; and calling the custom date conversion method convert_date in the term_data_preprocess method to process the date format. The convert_date method corrects the date and time data format errors contained in the original dataset by comprehensively calling isinstance and replace, and generates a complete timestamp.
[0007] Furthermore, in step (2), during the data integration process based on the term_data_merge method, all multi-source terminal temperature and humidity data are merged by data item through comprehensive calls to the join, drop, and merge methods, which facilitates subsequent analysis.
[0008] Furthermore, in the data visualization process of step (3), the toolkits of the pandas library and matplotlib library in the Python environment are used to draw curves of all terminal temperature and humidity data changing over time, so as to intuitively present the trend of all terminal temperature and humidity data changing over time.
[0009] Furthermore, in step (4), the specific steps include: sub-step (4-1) reading the integrated temperature and humidity dataset; sub-step (4-2) filtering the temperature dataset to achieve data dimensionality reduction; sub-step (4-3) data dimensionality transformation, converting each temperature item into a label; sub-step (4-4) data standardization to achieve dimensionlessness; sub-step (4-5) model selection and training, selecting the K-means model to cluster the terminal temperature data; and sub-step (4-6) model evaluation, using the Davies-Bouldin model. The index scores the clustering effect of the model and selects the best clustering result; sub-step (4-7) adds clustering results to the temperature dataset and visualizes the clustering results; sub-step (4-8) adds labels to the clustering results, including indoor temperature, supply air temperature and return air temperature containing a certain amount of manually modified supply / return air temperature data; sub-step (4-9) calculates and compares the similarity between the indoor temperature cluster and each temperature item in the temperature dataset, and filters and splits the room temperature data / room temperature item from the temperature dataset; sub-step (4-10) splits the indoor relative humidity item from the temperature and humidity dataset according to the correspondence between each temperature item and the relative humidity item.
[0010] The present invention also provides a system for splitting air conditioning terminal temperature and humidity data based on K-means, including: a data cleaning module for data cleaning based on the term_data_preprocess method;
[0011] The data integration module is used for data integration based on the term_data_merge method;
[0012] The data visualization processing module is used for data visualization processing;
[0013] The data aggregation and splitting module is used to aggregate and split data, filter out the temperature dataset, transform and standardize the data dimensions; select a model and train it, add clustering results to the temperature dataset; then add labels to the clustering results for filtering, split the room temperature item from the temperature dataset, and split the indoor relative humidity item from the temperature and humidity dataset.
[0014] Furthermore, the data cleaning module is also used to remove noisy data and missing values contained in the original dataset or generated during data structure transformation by calling the DataFrame's drop method and column attribute in the term_data_preprocess method, and to fill missing values by calling the backfill and forwardfill methods of the pandas library; the term_data_preprocess method also calls the custom date conversion method convert_date to process date formats. The convert_date method corrects date and time data format errors contained in the original dataset by comprehensively calling isinstance and replace, and generates complete timestamps.
[0015] The data integration module is also used to merge all multi-source terminal temperature and humidity data by data item by comprehensively calling the join, drop, and merge methods during the data integration process based on the term_data_merge method.
[0016] Furthermore, the data aggregation and splitting module is also used to perform sub-steps: (4-1) reading the integrated temperature and humidity dataset; (4-2) filtering the temperature dataset to achieve data dimensionality reduction; (4-3) data dimensionality transformation, converting each temperature item into a label; (4-4) data standardization to achieve dimensionlessness; (4-5) model selection and training, selecting the K-means model to cluster the terminal temperature data; and (4-6) model evaluation, using the Davies-Bouldin model. The index scores the clustering effect of the model and selects the best clustering result; sub-step (4-7) adds clustering results to the temperature dataset and visualizes the clustering results; sub-step (4-8) adds labels to the clustering results, including indoor temperature, supply air temperature, return air temperature, and manually modified temperature; sub-step (4-9) calculates and compares the similarity between the indoor temperature cluster and each temperature item in the temperature dataset, and filters and splits the room temperature data / room temperature item from the temperature dataset; sub-step (4-10) splits the indoor relative humidity item from the temperature and humidity dataset according to the correspondence between each temperature item and the relative humidity item.
[0017] As an implementation method, the present invention also employs an electronic device using the above method, including a memory and a processor. The memory is used to store a computer-executable application program, and the processor is used to execute the application program to implement the steps of the K-means-based method for splitting temperature and humidity data at air conditioning terminals.
[0018] The present invention also provides a computer-readable storage medium having computer-executable instructions stored thereon, the computer-executable instructions implementing the steps of the above-described method for splitting temperature and humidity data of air conditioning terminals based on K-means.
[0019] Compared with the prior art, the method, system, and electronic device of the present invention have the following advantages:
[0020] 1. The present invention's solution uses a custom term_data_preprocess method for data cleaning, efficiently handling missing values, noisy data, and time format errors in time-series air conditioning terminal temperature and humidity data; it combines data integration and data visualization based on a custom term_data_merge method, providing users with an intuitive display of the time-varying trends of all terminal temperature and humidity; data aggregation and splitting, combined with HVAC professional knowledge and the k-means clustering model, achieves the aggregation and labeling of terminal data, and accurately identifies and splits indoor temperature and humidity data through cluster similarity calculation, greatly improving the rationality and accuracy of data screening.
[0021] 2. For scenarios where terminal temperature and humidity data processing is chaotic and lacks clear labels in engineering practice, this invention can combine data preprocessing technology and machine learning data clustering models. Based on the terminal data analysis results, the clustering results can be labeled, such as room temperature clustering, supply air temperature clustering, and return air temperature clustering, which contain an indefinite number of manually modified data. This can accurately distinguish between them. The data is processed by algorithms multiple times to separate indoor temperature and relative humidity data from the mixed temperature and humidity data of air conditioning terminals. The results are then presented as visual and intuitive curves, significantly improving the accuracy of data splitting. The accuracy is about 60% higher than that of existing technologies.
[0022] 3. Existing air conditioning system data preprocessing methods can usually remove abnormal, duplicate, and noisy data from real terminal data with clear labels, but they are not applicable to situations where terminal data has no clear labels or contains a large amount of manually modified erroneous data, resulting in low accuracy.
[0023] The method and system of this invention combine HVAC professional knowledge with machine learning data clustering models to cluster and label mixed terminal data. By comparing the similarity between the labeled room temperature cluster and each terminal temperature dataset pair, the similarity between each unlabeled terminal temperature dataset (such as temperature 1 and temperature 2 in Figure 2(a)) and the room temperature cluster is evaluated by calculating the mean or standard deviation. Based on the similarity, the indoor temperature and humidity data are accurately identified and separated from the air conditioning terminal temperature and humidity dataset, solving the problem that existing air conditioning data preprocessing methods cannot handle unlabeled terminal datasets containing a large amount of manually modified data.
[0024] Taking a central air conditioning project as an example, existing technologies can only roughly determine that temperatures 3, 5, 6, 7, and 8 are potential room temperature datasets based on the air conditioning design room temperature range and the PID control characteristics of the air conditioning system. This method yields unclear, low-quality, and large-scale results, limiting the computational load for subsequent integration into air conditioning energy-saving control algorithms. However, by applying the method and system of this invention, temperature 3 can be accurately identified as a room temperature dataset, improving data processing accuracy by approximately 60% compared to existing technologies. This provides usable, high-quality indoor temperature and humidity data for the development of global optimization energy-saving algorithms or energy-saving analysis for central air conditioning systems. Attached Figure Description
[0025] Figure 1 This is a schematic diagram of the terminal temperature and humidity data splitting method based on K-means of the present invention;
[0026] Figure 2(a) is a schematic diagram of the change of terminal temperature over time in a certain case;
[0027] Figure 2(b) is a schematic diagram of the change of terminal relative humidity over time in a certain case;
[0028] Figure 3 This is a schematic diagram of the terminal temperature clustering density results in a case study where the method of this invention was applied.
[0029] Figure 4 This is a schematic diagram illustrating the verification of room temperature data splitting results using the method of this invention in a specific case.
[0030] In Figure 2(a): the horizontal axis represents the time of data collection for the terminal temperature in a certain case, the vertical axis represents the temperature (°C), and the 8 curves represent 8 sets of terminal temperature data without clear labels (including manually modified data) at the corresponding collection time.
[0031] In Figure 2(b): the horizontal axis represents the data collection time of the terminal relative humidity in a certain case, the vertical axis represents the relative humidity, and the 8 curves represent 8 sets of terminal relative humidity data without clear labels (including manually modified data) at the corresponding collection time.
[0032] Figure 3 In the diagram: the horizontal axis represents the terminal temperature (°C) in a specific case, and the vertical axis represents the kernel density value (1 / °C) corresponding to the temperature in each cluster. The red curve represents the temperature kernel density distribution curve in Cluster 0 (mainly indoor temperature data, with a very small amount of artificially modified supply / return air temperature data), the blue curve represents the temperature kernel density distribution curve in Cluster 1 (mainly return air temperature data, with a large amount of artificially modified supply air temperature data), and the green curve represents the temperature kernel density distribution curve in Cluster 2 (mainly supply air temperature data, with a small amount of artificially modified return air temperature data).
[0033] Figure 4 In the diagram: the horizontal axis represents the cluster temperature and terminal temperature (°C) of Cluster 0 in a certain case, and the vertical axis represents the kernel density value (1 / °C) corresponding to the temperature. The green, light green, pink, orange, red, cyan, yellow, light blue, and purple curves are the kernel density distribution curves of the Cluster 0 cluster temperature (room temperature cluster), temperature 1, temperature 2, temperature 3, temperature 4, temperature 5, temperature 6, temperature 7, and temperature 8 datasets, respectively. Specific implementation methods
[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0035] Please see Figure 1 As shown, the present invention provides a method for splitting air conditioning terminal temperature and humidity data based on K-means, comprising: step (1) data cleaning based on the term_data_preprocess method, step (2) data integration based on the term_data_merge method, step (3) data visualization, and step (4) data aggregation and splitting; wherein step (4) includes filtering out the temperature dataset, transforming and standardizing the data dimensions; selecting a model and training it, adding clustering results to the temperature dataset; then adding labels to the clustering results for filtering, splitting the room temperature item from the temperature dataset, and splitting the indoor relative humidity item from the temperature and humidity dataset.
[0036] Step (1) includes deleting noisy data and missing values contained in the original dataset or generated during data structure transformation by calling the drop method of DataFrame and the columns attribute in the term_data_preprocess method, and filling missing values by calling the backfill and forwardfill methods of pandas library; and calling the custom date conversion method convert_date in the term_data_preprocess method to handle date format issues. The convert_date method corrects the date and time data format errors contained in the original dataset by comprehensively calling isinstance and replace, and generates a complete timestamp.
[0037] Please see Figure 1As shown in Figures 2(a)-(b), in a certain case, step (1) includes sub-step (1-1) calling the drop method and column attribute of the DataFrame in the term_data_preprocess method to delete noisy data columns contained in the original dataset or generated during the data structure conversion process, as well as blank columns without any data, and using backfill and forwardfill methods to fill missing values; sub-step (1-2) for the case of mixed Chinese characters and numbers in the original date data, calling the custom date conversion method convert_date in the term_data_preprocess method to handle date format issues. The convert_date method converts Chinese characters into numbers by comprehensively calling isinstance, replace, etc. to correct the date and time data format errors contained in the original dataset, and generates a complete timestamp to facilitate subsequent time series data processing.
[0038] Specifically, in step (2), during the data integration process based on the term_data_merge method, all multi-source end temperature and humidity data are merged by data item through comprehensive calls to join, drop, merge, and other methods, which facilitates subsequent analysis.
[0039] Specifically, in the data visualization process of step (3), the toolkits of the pandas library and matplotlib library in the Python environment are used to draw curves of the change of terminal temperature and relative humidity data over time, as shown in Figure 2(a)-(b).
[0040] Please see Figures 1 to 4As shown, step (4) specifically includes sub-step (4-1) using the pandas library toolkit to read the integrated end-point temperature and humidity data. Sub-step (4-2) removing time series data and relative humidity data, and filtering out the temperature dataset to achieve data dimensionality reduction. This can reduce the amount of computation and improve analysis efficiency. Sub-step (4-3) data dimension transformation, using the melt method in the pandas library to transform the data dimensions, converting each temperature item into a label; where the frame attribute is set to the name of the Dataframe composed of the end-point temperature dataset, the value_name attribute is set to 'temperature', and the var_name attribute is set to 'label'; after the transformation, the drop method of the end-point temperature Dataframe is applied to remove the label column. Sub-step (4-4) applying the drop method to discard the label column, and using the StandardScaler method of the sklearn database to standardize the data to achieve dimensionlessness; where both the drop method and the StandardScaler method use the system default parameter settings. Sub-steps (4-5) involve model selection and training. The K-means model from the sklearn database is selected to divide the terminal temperature dataset into three clusters: room temperature cluster, return air temperature cluster, and supply air temperature cluster, which contain varying amounts of artificially modified data. This is a unique data processing method developed by our team, which can scientifically and reasonably classify data, avoid confusion, and facilitate subsequent model building for evaluation and effective data selection.
[0041] Sub-steps (4-6) involve model evaluation. The davies_bouldin Index method from the sklearn library is used to score the model's clustering performance. A lower score indicates better clustering. The lowest score (0.33) is achieved when the number of clusters is 3, indicating that a clustering number of 3 is reasonable. Our team has found through experimental research that this model evaluation scoring method is superior to existing variance ratio criteria and external evaluation indicators.
[0042] Sub-steps (4-7) add the clustering results to the temperature dataset and visualize the clustering results, as shown in the following diagram. Figure 3 As shown. Sub-steps (4-8) add labels to the clustering results, including indoor temperature, supply air temperature, return air temperature, and manually modified temperatures. For example... Figure 3 Cluster 0 is mainly composed of indoor temperature data, with a very small amount of manually modified supply / return air temperature data mixed in, and is called room temperature cluster; Cluster 1 is mainly composed of return air temperature data, with a large amount of manually modified supply air temperature data mixed in, and is called return air temperature cluster; Cluster 2 is mainly composed of supply air temperature data, with a small amount of manually modified return air temperature data mixed in, and is called supply air temperature cluster.
[0043] Sub-steps (4-9) calculate the similarity between the room temperature cluster and each temperature dataset using the mean or standard deviation (see Formulas 1 and 2 below). The temperature dataset with the highest similarity to the room temperature cluster is the indoor temperature dataset, thus completing the labeling and splitting of the room temperature data. The visualization results of the similarity calculation are shown below. Figure 4 As shown, temperature 3 is significantly more likely to be room temperature data than the other seven temperature datasets, and is listed as the best result.
[0044]
[0045] In the formula above, Cluster x and Cluster room These represent clustering at arbitrary terminal temperatures and clustering at room temperature, respectively; N x and N room , respectively, represent the data size of the arbitrary end-temperature dataset and the data size of the room temperature cluster; t represents any temperature data value in the arbitrary end-temperature cluster, in °C; μ x and μ room , respectively, are the temperature mean values for arbitrary end temperature clustering and room temperature clustering, in °C; α is a user-defined constant, typically 0.5, but can also be set according to actual conditions.
[0046] Sub-step (4-10) Based on the correspondence between each temperature item and the relative humidity item, the indoor relative humidity item is extracted from the temperature and humidity dataset. It is accurately determined that Humidity 3 is the best indoor relative humidity dataset, which is significantly higher than the other seven relative humidity datasets.
[0047] This invention replaces existing air conditioning data preprocessing methods. While achieving basic functions such as removing abnormal, duplicate, and noisy data and completing missing data, it first combines professional domain knowledge with machine learning data clustering models to cluster and label the end data. It uses a special clustering method to process the data to avoid confusion. Then, it calculates the similarity between each unknown label temperature dataset and the room temperature cluster by using the mean or standard deviation. Finally, it can identify and separate indoor temperature and humidity data from the mixed end temperature and humidity data that has no clear labels and contains a lot of manually modified data, so as to achieve the goal of fast and accurate data analysis.
[0048] This invention also provides a system, electronic device, and running program employing the above method to implement the steps of the K-means-based method for splitting air conditioning terminal temperature and humidity data. The terminal data is clustered using the k-means clustering model, and the Davies-Boudin Index is used to evaluate the rationality of the clustering. This significantly improves the rationality and accuracy of data selection.
[0049] Furthermore, it should be understood that although this specification describes the implementation methods, not every implementation method contains only one independent technical solution. This method of description is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.
Claims
1. A method for splitting temperature and humidity data at air conditioning terminals based on K-means, characterized in that, include: Step (1) Data cleaning based on the term_data_preprocess method, Step (2) Data integration based on the term_data_merge method, Step (3) Data visualization, Step (4) Data aggregation and splitting; wherein Step (4) includes filtering out the temperature dataset, transforming and standardizing the data dimensions; selecting a model and training it, adding clustering results to the temperature dataset; then adding labels to the clustering results for filtering, splitting the room temperature item from the temperature dataset, and splitting the indoor relative humidity item from the temperature and humidity dataset.
2. The method for splitting air conditioning terminal temperature and humidity data based on K-means according to claim 1, characterized in that: Step (1) includes deleting noisy data and missing values contained in the original dataset or generated during data structure conversion by calling the drop method of the DataFrame and the columns attribute in the term_data_preprocess method; and calling the custom date conversion method convert_date in the term_data_preprocess method to process the date format. The convert_date method corrects the date and time data format errors contained in the original dataset by comprehensively calling isinstance and replace, and generates a complete timestamp.
3. The method for splitting air conditioning terminal temperature and humidity data based on K-means according to claim 1, characterized in that: In step (2), during the data integration process based on the term_data_merge method, all multi-source terminal temperature and humidity data are merged by data item through comprehensive calls to the join, drop, and merge methods, which facilitates subsequent analysis.
4. The method for splitting air conditioning terminal temperature and humidity data based on K-means according to claim 1, characterized in that: In the data visualization process of step (3), the toolkits of pandas and matplotlib libraries in the Python environment are used to draw curves of all terminal temperature and humidity data changing over time, so as to intuitively present the trend of all terminal temperature and humidity data changing over time.
5. The method for splitting air conditioning terminal temperature and humidity data based on K-means as described in claim 1, characterized in that: Step (4) specifically includes sub-step (4-1) reading the integrated temperature and humidity dataset; sub-step (4-2) filtering the temperature dataset to achieve data dimensionality reduction; sub-step (4-3) data dimensionality transformation, converting each temperature item into a label; sub-step (4-4) data standardization to achieve dimensionlessness; sub-step (4-5) model selection and training, selecting the K-means model to cluster the terminal temperature data; and sub-step (4-6) model evaluation, using the Davies-Bouldin model. The index scores the clustering effect of the model and selects the best clustering result; sub-step (4-7) adds clustering results to the temperature dataset and visualizes the clustering results; sub-step (4-8) adds labels to the clustering results, including indoor temperature, supply air temperature, return air temperature, and manually modified temperature; sub-step (4-9) calculates and compares the similarity between the indoor temperature cluster and each temperature item in the temperature dataset, and filters and splits the room temperature data / room temperature item from the temperature dataset; sub-step (4-10) splits the indoor relative humidity item from the temperature and humidity dataset according to the correspondence between each temperature item and the relative humidity item.
6. A system for splitting temperature and humidity data at air conditioning terminals based on K-means, characterized in that, include: The data cleaning module is used for data cleaning based on the term_data_preprocess method; The data integration module is used for data integration based on the term_data_merge method; The data visualization processing module is used for data visualization processing; The data aggregation and splitting module is used to aggregate and split data, filter out temperature datasets, and perform data dimension transformation and standardization. Select a model and training, add clustering results to the temperature dataset; then add labels to the clustering results for filtering, and split the room temperature item from the temperature dataset and the indoor relative humidity item from the temperature and humidity dataset.
7. The system for splitting air conditioning terminal temperature and humidity data based on K-means as described in claim 6, characterized in that, The data cleaning module is also used to remove noisy data and missing values contained in the original dataset or generated during data structure transformation by calling the drop method and column attribute of the DataFrame in the term_data_preprocess method; and to call the custom date conversion method convert_date in the term_data_preprocess method to process the date format. The convert_date method corrects the date and time data format errors contained in the original dataset by comprehensively calling isinstance and replace, and generates a complete timestamp. The data integration module is also used to merge all multi-source terminal temperature and humidity data by data item by comprehensively calling the join, drop, and merge methods during the data integration process based on the term_data_merge method.
8. A system for splitting air conditioning terminal temperature and humidity data based on K-means as described in claim 6, characterized in that, The data aggregation and splitting module is also used to execute sub-steps: (4-1) reading the integrated temperature and humidity dataset; (4-2) filtering the temperature dataset to achieve data dimensionality reduction; (4-3) data dimensionality transformation, converting each temperature item into a label; (4-4) data standardization to achieve dimensionlessness; (4-5) model selection and training, selecting the K-means model to cluster the terminal temperature data; and (4-6) model evaluation, using the Davies-Bouldin model. The index scores the clustering effect of the model and selects the best clustering result; sub-step (4-7) adds clustering results to the temperature dataset and visualizes the clustering results; sub-step (4-8) adds labels to the clustering results, including indoor temperature, supply air temperature and return air temperature containing a certain amount of manually modified supply / return air temperature data; sub-step (4-9) calculates and compares the similarity between the indoor temperature cluster and each temperature item in the temperature dataset, and filters and splits the room temperature data / room temperature item from the temperature dataset; sub-step (4-10) splits the indoor relative humidity item from the temperature and humidity dataset according to the correspondence between each temperature item and the relative humidity item.
9. An electronic device employing the method of any one of claims 1-5, comprising a memory and a processor, characterized in that: The memory is used to store computer-executable applications, and the processor is used to execute the applications to implement the steps of the K-means-based method for splitting temperature and humidity data at air conditioning terminals.
10. A computer-readable storage medium having computer-executable instructions stored thereon, characterized in that: The computer-executable instructions implement the steps of the K-means-based air conditioning terminal temperature and humidity data splitting method as described in any one of claims 1-5.