Energy-saving diagnostic methods for air conditioning systems in large public buildings

Abstract

This invention discloses an energy-saving diagnostic method for air conditioning systems in large public buildings. The method includes: using weather and / or human activity as the basis for operating condition classification; employing a clustering method to divide preprocessed operating data into multiple data clusters; classifying operating condition types; comparing the building energy consumption data and / or indoor temperature data of each data cluster with data under typical operating conditions; determining abnormal indicator data for the current actual operating mode of the air conditioning system based on the consistency between the indicators of each data cluster and the indicators under typical operating conditions; performing symbolic approximation aggregation on the data; comparing the abnormal indicator data with the typical operating mode of the air conditioning system; determining whether the operating mode corresponding to the abnormal indicator data is abnormal; and providing a recommended operating mode or re-diagnosis result for controlling the HVAC system based on the judgment result. This invention can comprehensively analyze the operating status of HVAC systems.

Description

Technical Field

[0001] This invention relates to the field of building energy management technology, and in particular to an energy-saving diagnostic method for air conditioning systems in large public buildings. Background Technology

[0002] With the increase in public buildings, building energy consumption is rising, and the proportion of high-energy-consuming buildings in new construction is also increasing. The energy consumption per unit area of ​​public buildings shows little regional variation; the key factor influencing the energy intensity of public buildings is their size and scale. Large-scale public buildings mostly adopt glass curtain walls and other enclosed structures, are fully enclosed, and often use central air conditioning systems. However, due to the complexity of their HVAC systems, effective adjustment methods are lacking in actual operation. The actual operating conditions of the main energy-consuming equipment (chillers, circulating water pumps, air handling units, terminals, etc.) deviate significantly from their peak efficiency, resulting in unnecessary energy consumption. Secondly, in energy-saving diagnostics, data mining-based diagnostic methods require a large amount of operational or energy consumption monitoring data, with limited on-site testing. They often lack close integration with basic building information, and the various indicators are independent of each other. Connecting them requires extensive expert knowledge and experience, necessitating specific analysis of each problem. This process is difficult to quantify and mathematically describe, making the methods impractical. Secondly, the operation of air conditioning systems is dynamic and constantly changing. The previous diagnoses of equipment and system performance can be approximated as the instantaneous operating state of the air conditioner, which is insufficient for a dynamic evaluation of the air conditioning system. To connect these aspects, it is necessary to utilize a large amount of expert knowledge and experience for judgment, and to analyze specific problems on a case-by-case basis. This process is difficult to quantify and describe mathematically. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings and defects of the prior art and provide an energy-saving diagnostic method for air conditioning systems in large public buildings. The method of this invention is based on the historical continuous operation data of the air conditioning system and uses data mining technology to perform diagnosis from the outside to the inside and from the whole to the part, and finally realizes the diagnosis of the operation status of the air conditioning system.

[0004] A method for energy-saving diagnosis of air conditioning systems in large public buildings, comprising the following steps:

[0005] Operational data preprocessing includes standardizing the data types and sampling frequencies of discrete data in HVAC systems;

[0006] Operational data clustering: Based on weather and / or personnel activity as the basis for classifying operating conditions, a clustering method is used to divide the preprocessed operational data into multiple data clusters for classifying operating condition types;

[0007] Preliminary diagnosis of operational data: Compare the building energy consumption data and / or indoor temperature data of each data cluster with the data under typical operating conditions. Based on the degree of consistency between the indicators of each data cluster and the indicators under typical operating conditions, identify the abnormal indicator data of the current actual operating mode of the air conditioning system.

[0008] Runtime data symbol approximation aggregation processing: Process the preprocessed runtime data and convert it into a discrete format;

[0009] Operational data re-diagnosis: Compare abnormal indicator data with the typical operating mode of the air conditioning system to determine whether the operating mode corresponding to the abnormal indicator data is abnormal. Based on the judgment result, provide a recommended operating mode or re-diagnosis result for controlling the HVAC system.

[0010] Specifically, the data type of the unified discrete data of the HVAC system is to unify the discrete data of the HVAC system through a unified symbolic representation method; the sampling frequency of the unified data is to unify the data sampling frequency by using window moving average interpolation.

[0011] The personnel activities include personnel density and activity type; the weather includes temperature, humidity, and sunshine; the activity type includes whether the monitoring area is occupied or unoccupied; and the personnel activity status.

[0012] In the initial diagnosis of the operational data, the K-means algorithm is used to cluster the preprocessed operational data, and the silhouette coefficient algorithm is used to continuously iterate until the optimal number of clusters is obtained. Based on the optimal number of clusters, multiple data clusters are obtained to realize the classification of working conditions.

[0013] In the initial diagnosis of the operating data, a consistency evaluation percentage threshold is given. If the comparison result is less than the consistency evaluation percentage threshold, the current actual operating mode of the air conditioning system is considered to meet the requirements; otherwise, abnormal indicator data of the current actual operating mode of the air conditioning system is given for re-diagnosis.

[0014] The running data symbol approximation aggregation processing includes processing methods such as sequence normalization, sequence aggregation approximation, and sequence symbolization representation, which process the preprocessed running data and transform it into the required discrete format.

[0015] The normalization process refers to converting the continuous original time series data into a discrete format to generate a new sequence with a mean of 0 and a standard deviation of 1. The sequence clustering approximation process refers to shortening the length of the new sequence formed after normalization by using a preset sequence segmentation length, decomposing it into multiple subsequences. The sequence symbolization process refers to replacing the original time series with the letter strings of the subsequences of the new sequence of the running data obtained after the sequence clustering approximation process.

[0016] In the process of approximate aggregation of running data symbols, the SAX algorithm is used to process the preprocessed running data.

[0017] In the step of re-diagnosing the operating data, a frequent pattern mining method is used to mine the operating patterns of the air conditioning system; when an operating pattern repeats a preset number of times or more, the operating pattern is regarded as a typical operating pattern, and the typical operating pattern is one or more.

[0018] The abnormal indicator data obtained from the initial diagnosis of the operating data are matched with the typical operating mode, and a recommended operating mode or a re-diagnosis result is given based on the matching result.

[0019] The step of providing a re-diagnosis result based on the matching result includes:

[0020] If the abnormal indicator data obtained in the initial diagnosis of the operating data does not match one of the operating modes, the abnormal indicator data will continue to be matched with other operating modes. If there is no matching operating mode, it is assumed that the abnormal indicator data in the initial diagnosis is due to improper settings of the internal operating strategy of the air conditioning system. The distance between the actual operating mode of the air conditioning system and other operating modes is calculated, and the operating mode with the smallest distance from the actual operating mode is regarded as the recommended operating mode. If there is a matching operating mode, the cause and explanation are given using expert knowledge to complete the secondary diagnosis of the air conditioning system operation.

[0021] The method of this invention adopts a diagnostic sequence from the whole to the part. It first uses data mining methods to classify operating conditions and modes, and then uses certain professional knowledge to analyze and diagnose the classification results. This provides a systematic solution for subsequent air conditioning system operation diagnosis applications, and can comprehensively analyze the operating status of HVAC systems, providing a new data analysis approach for utilizing HVAC operation data.

[0022] The secondary diagnosis method based on frequent pattern mining of the present invention mines the typical operating modes of the air conditioning system, determines whether the operating mode of the air conditioning system is abnormal, and provides recommendations for operating modes. Compared with traditional rule-based or experience-based diagnostic methods, it is more adaptable to changes in different environments and has higher accuracy and reliability. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating the energy-saving diagnostic method for air conditioning systems in large public buildings according to the present invention.

[0024] Figure 2 This is a schematic diagram of the data clustering process of the present invention.

[0025] Figure 3This is a schematic diagram of the process for approximate aggregation of data symbols in the present invention. Detailed Implementation

[0026] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0027] See Figure 1 As shown, the energy-saving diagnostic method for air conditioning systems in large public buildings of the present invention includes the following steps:

[0028] S1. Operational data preprocessing: This includes standardizing the data type of discrete data in HVAC systems and standardizing the data sampling frequency;

[0029] S2. Operational Data Clustering: Based on weather and / or personnel activities as the basis for classifying operating conditions, a clustering method is used to divide the preprocessed operational data into multiple data clusters for classifying operating condition types.

[0030] S3. Preliminary diagnosis of operating data: Compare the building energy consumption data and / or indoor temperature data of each data cluster with the data under typical operating conditions. Based on the degree of consistency between the indicators of each data cluster and the indicators under typical operating conditions, determine the abnormal indicator data of the current actual operating mode of the air conditioning system.

[0031] S4. Symbolic Aggregate Approximation (SAX): Processes the preprocessed running data to convert it into a discrete format;

[0032] S5. Operational Data Re-diagnosis: Compare abnormal indicator data with the typical operating mode of the air conditioning system to determine whether the operating mode corresponding to the abnormal indicator data is abnormal. Based on the judgment result, provide a recommended operating mode or re-diagnosis result for controlling the HVAC system.

[0033] The operational data includes data on chillers, cooling towers, water pumps, air conditioning units, heat recovery units, and the indoor environment, including energy consumption monitoring data.

[0034] Specifically, the data type of the unified discrete data of the HVAC system is to unify the discrete data of the HVAC system through a unified symbolic representation method, that is, to present the data with high quality after cleaning and transformation of the operating data; the sampling frequency of the unified data is to unify the data sampling frequency by using window moving average interpolation.

[0035] Preprocessing of runtime data also includes data cleaning, data merging, data transformation, and data standardization. Through data preprocessing, the problem of inconsistent raw data quality is solved, and the problem of inconsistent sampling frequency is solved by using window moving average interpolation.

[0036] The personnel activities include personnel density and activity type; the weather includes temperature, humidity and sunshine; the activity type includes whether the monitoring area is occupied or unoccupied; and the personnel's activity status and condition.

[0037] In the initial diagnostic process of the operational data, the K-means algorithm is used to cluster the preprocessed operational data. Simultaneously, the silhouette coefficient algorithm is used for iterative calculation until the optimal number of clusters is obtained. Based on the optimal number of clusters, multiple data clusters are obtained, achieving the classification of operating conditions. Clustering groups data that initially lacked categories into different groups; such a collection of data objects is called a cluster, and each cluster is described. Clustering ensures that samples belonging to the same cluster are similar to each other, while samples from different clusters are sufficiently dissimilar.

[0038] K-means is a distance-based clustering algorithm that uses distance as a similarity metric, meaning that the closer two objects are, the greater their similarity. K-means considers clusters to be groups of objects that are close together, thus aiming to obtain compact and independent clusters as its final goal. It includes the following steps: Figure 2 As shown:

[0039] Step S11: Select the number of clusters K. The silhouette coefficient method is mainly used, and the cohesion (the degree of clustering of samples within a cluster) and the separation (the degree of separation of samples between clusters) are combined to determine whether the clustering is reasonable and effective, and to give the optimal K value.

[0040] Step S12: Input the value of K, i.e., the desired dataset D = {O1, O2, ..., O}. n After clustering, K categories or groups are obtained; K data points are randomly selected from dataset D as cluster centroids, each cluster centroid representing a cluster; the resulting cluster centroid basis is Centroid = {C p1 C p2 ,…,C pk}

[0041] Step S13: For each data point O in dataset D i Calculate O i With C pj We obtain a set of distance values ​​for (j = 1, 2, ..., k), and find the cluster centroid C corresponding to the smallest distance value from this set. ps , put data point O iDivided into C ps In a cluster whose centroid is [the center of mass].

[0042] Step S14: Based on the set of objects contained in each cluster, recalculate a new cluster centroid using the method in step S13. If |C x | is the x-th cluster C x The number of objects in m x If a vertex is the centroid of these objects, then it must satisfy the following relation:

[0043]

[0044] By following the steps above, multiple data families can be obtained using the K-means algorithm.

[0045] In the initial diagnosis of the operating data, a consistency evaluation percentage threshold is given. If the comparison result is less than the consistency evaluation percentage threshold, the current actual operating mode of the air conditioning system is considered to meet the requirements; otherwise, abnormal indicator data of the current actual operating mode of the air conditioning system is given for re-diagnosis.

[0046] Specifically, the building energy consumption data and indoor temperature data for each family of similar operating conditions, categorized by operating type, are compared and calculated with the corresponding data under preset typical operating conditions. The consistency between the data indicators under typical operating conditions and those under similar test conditions is analyzed. If the difference is less than 10%, the building is considered to be operating normally under that test condition. If the difference is greater than 10%, potential operational problems are considered to exist under that test condition, requiring further secondary or re-diagnosis.

[0047] The running data symbol approximation aggregation processing includes processing methods such as sequence normalization, sequence aggregation approximation, and sequence symbolization representation, which process the preprocessed running data and transform it into the required discrete format.

[0048] The normalization process refers to converting the continuous original time series data into a discrete format to generate a new sequence with a mean of 0 and a standard deviation of 1. The sequence clustering approximation process refers to shortening the length of the new sequence formed after normalization by using a preset sequence segmentation length, decomposing it into multiple subsequences. The sequence symbolization process refers to replacing the original time series with the letter strings of the subsequences of the new sequence of the running data obtained after the sequence clustering approximation process.

[0049] Preferably, in the running data symbol approximation aggregation processing, the SAX algorithm is used to process the preprocessed running data.

[0050] Specifically, the time range of the data processed in the re-diagnosis and the initial diagnosis in this invention is consistent, and the data for the re-diagnosis is the operating data of the air conditioning system on the date when the indicators obtained after the initial diagnosis are abnormal.

[0051] In specific implementation, during re-diagnosis, this invention reprocesses the target data in the preprocessed original running data time series, converting it into a discrete format while preserving its strong time-series characteristics and performing dimensionality reduction. This includes sequence normalization, sequence aggregation approximation, and sequence symbolization. Normalization refers to generating a new sequence with a mean of 0 and a standard deviation of 1 from the original sequence. Sequence aggregation approximation involves setting a certain segmentation length to shorten the length of the normalized sequence, decomposing it into subsequences. Sequence symbolization involves replacing the original time series with letter strings for the subsequences, converting them into letter symbols for storage. By combining letter symbols into strings to replace the original time series, subsequent mining and visualization techniques are supported. This includes the following steps, such as... Figure 3 As shown:

[0052] Step S21: Normalize the time series x(t) of the target data from the original running data to generate a new series Z(t) with a mean of 0 and a standard deviation of 1. The transformation process is as follows:

[0053]

[0054] Where μ is the mean of the original time series x(t), and σ is the standard deviation of x(t);

[0055] Step S22: Reduce the dimensionality of the normalized sequence Z(t), resulting in a new sequence Z(t) of length w = z1, z2…z w This is transformed into a sequence of length m, Y(t) = y1, y2…y m And the sequence Y(t) and the new sequence Z(t) satisfy the following relationship:

[0056]

[0057] Among them, the definition The segment length for SAX processing of running data;

[0058] Step S23: Replace the original sequence with a string of letters.

[0059] In this step, two parameters are first defined for SAX execution: word length W and number of symbols A. The letter delimiters are determined by referring to a normal distribution. For example, if the estimated time is 24 hours, then the daily air conditioning system operation data will form a subsequence, with the number of symbols A being the number of operating modes, and the word length W being 24 hours.

[0060] The air conditioning system operation data includes data on chillers, cooling towers, water pumps, air conditioning units, heat recovery units, and the indoor environment, including energy consumption monitoring data.

[0061] Through the above process, the target running data sequence in the original running dataset is processed by SAX to obtain a continuous string, thereby reducing the dimensionality of the original time series data while preserving the local "context" information of the data.

[0062] In the step of re-diagnosing the operating data, a frequent pattern mining method is used to mine the operating patterns of the air conditioning system; when an operating pattern repeats a preset number of times (e.g., 4 times) or more, the operating pattern is regarded as a typical operating pattern, and the typical operating pattern is one or more.

[0063] The abnormal indicator data obtained from the initial diagnosis of the operating data are matched with the typical operating mode, and a recommended operating mode or a re-diagnosis result is given based on the matching result.

[0064] In the above technical solution, the frequent pattern refers to a pattern that frequently appears in the dataset.

[0065] The step of providing a re-diagnosis result based on the matching result includes:

[0066] If the abnormal indicator data obtained in the initial diagnosis of the operating data does not match one of the operating modes, the abnormal indicator data will continue to be matched with other operating modes. If there is no matching operating mode, it is assumed that the abnormal indicator data in the initial diagnosis is due to improper settings of the internal operating strategy of the air conditioning system. The distance between the actual operating mode of the air conditioning system and other operating modes is calculated, and the operating mode with the smallest distance from the actual operating mode is regarded as the recommended operating mode. If there is a matching operating mode, the cause and explanation are given using expert knowledge to complete the secondary diagnosis of the air conditioning system operation.

[0067] The method of this invention adopts a diagnostic sequence from the whole to the part. It first uses data mining methods to classify operating conditions and modes, and then uses certain professional knowledge to analyze and diagnose the classification results. This provides a systematic solution for subsequent air conditioning system operation diagnosis applications, and can comprehensively analyze the operating status of HVAC systems, providing a new data analysis approach for utilizing HVAC operating data.

[0068] The secondary diagnosis method based on frequent pattern mining of the present invention mines the typical operating modes of the air conditioning system, determines whether the operating mode of the air conditioning system is abnormal, and provides recommendations for operating modes. Compared with traditional rule-based or experience-based diagnostic methods, it is more adaptable to changes in different environments and has higher accuracy and reliability.

[0069] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention.

[0070] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the claims be included within the invention.

[0071] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style 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 embodiments that can be understood by those skilled in the art.

Claims

1. A method for energy-saving diagnosis of air conditioning systems in large public buildings, characterized in that, Including the following steps: Operational data preprocessing includes standardizing the data types and sampling frequencies of discrete data in HVAC systems; Operational data clustering: Based on weather and / or personnel activity as the basis for classifying operating conditions, a clustering method is used to divide the preprocessed operational data into multiple data clusters for classifying operating condition types; the personnel activity includes personnel density and activity type, the weather includes temperature, humidity and sunshine, the activity type includes whether the monitoring area is occupied or unoccupied, and the activity status of personnel; Preliminary diagnosis of operational data: Compare the building energy consumption data and / or indoor temperature data of each data cluster with the data under typical operating conditions. Based on the degree of consistency between the indicators of each data cluster and the indicators under typical operating conditions, identify the abnormal indicator data of the current actual operating mode of the air conditioning system. Runtime data symbol approximation aggregation processing: Process the preprocessed runtime data and convert it into a discrete format; Operational data re-diagnosis: Compare abnormal indicator data with the typical operating mode of the air conditioning system to determine whether the operating mode corresponding to the abnormal indicator data is abnormal. Based on the judgment result, provide a recommended operating mode or re-diagnosis result for controlling the HVAC system. In the step of re-diagnosing the operating data, the frequent pattern mining method is used to mine the operating patterns of the air conditioning system. When a running mode is repeated a preset number of times or more, the running mode is considered as a typical running mode, and the typical running mode is one or more. The abnormal indicator data obtained from the initial diagnosis of the operating data are matched with the typical operating mode, and a recommended operating mode or a re-diagnosis result is given based on the matching result. The step of providing a re-diagnosis result based on the matching results includes: If the abnormal indicator data obtained in the initial diagnosis of the operating data does not match one of the operating modes, the abnormal indicator data will continue to be matched with other operating modes. If there is no matching operating mode, it is assumed that the abnormal indicator data in the initial diagnosis is due to improper settings of the internal operating strategy of the air conditioning system. The distance between the actual operating mode of the air conditioning system and other operating modes is calculated, and the operating mode with the smallest distance from the actual operating mode is regarded as the recommended operating mode. If there is a matching operating mode, the cause and explanation are given using expert knowledge to complete the secondary diagnosis of the air conditioning system operation.

2. The energy-saving diagnostic method for air conditioning systems in large public buildings according to claim 1, characterized in that, The data type of the unified HVAC system discrete data is to unify the discrete data of the HVAC system through a unified symbolic representation method; the sampling frequency of the unified data is to unify the data sampling frequency by using window moving average interpolation.

3. The energy-saving diagnostic method for air conditioning systems in large public buildings according to claim 1, characterized in that, In the operation data clustering, the K-means algorithm is used to cluster the preprocessed operation data, and the silhouette coefficient algorithm is used to continuously iterate until the optimal number of clusters is obtained. Based on the optimal number of clusters, multiple data clusters are obtained to realize the division of operating conditions.

4. The energy-saving diagnostic method for air conditioning systems in large public buildings according to claim 1, characterized in that, In the initial diagnosis of the operating data, a consistency evaluation percentage threshold is given. If the comparison result is less than the consistency evaluation percentage threshold, the current actual operating mode of the air conditioning system is considered to meet the requirements; otherwise, abnormal indicator data of the current actual operating mode of the air conditioning system is given for re-diagnosis.

5. The energy-saving diagnostic method for air conditioning systems in large public buildings according to claim 1, characterized in that, The running data symbol approximation aggregation processing includes processing methods such as sequence normalization, sequence aggregation approximation, and sequence symbolization representation to process the preprocessed running data and transform it into the required discrete format.

6. The energy-saving diagnostic method for air conditioning systems in large public buildings according to claim 5, characterized in that, The sequence normalization process refers to converting the continuous raw time series data into a discrete format to generate a new sequence with a mean of 0 and a standard deviation of 1. The sequence clustering approximation refers to shortening the length of the new sequence formed after normalization by using a preset sequence segmentation length, and decomposing it into multiple subsequences; the sequence symbolization refers to replacing the original time series with the letter strings of the subsequences of the new sequence of the running data obtained after the sequence clustering approximation.

7. The energy-saving diagnostic method for air conditioning systems in large public buildings according to claim 6, characterized in that, In the approximate aggregation of the running data symbols, the SAX algorithm is used to process the preprocessed running data.

Images