CN-120910731-B - Building equipment abnormality identification method and system based on machine learning

CN120910731BCN 120910731 BCN120910731 BCN 120910731BCN-120910731-B

Abstract

The application relates to the technical field of machine learning, and provides a building equipment abnormality identification method and system based on machine learning, which are used for realizing accurate detection and precise early warning of building equipment abnormality. The method comprises the steps of obtaining a continuous operation data set of target building equipment, wherein the continuous operation data set comprises a multi-section equipment state recording unit with a timestamp mark, conducting time-frequency domain feature extraction processing on the continuous operation data set to obtain a time-frequency domain feature set of the equipment state recording unit, invoking a pre-built hybrid machine learning model to conduct abnormality detection processing on the time-frequency domain feature set to generate an abnormality identification result of the equipment state recording unit, determining an abnormality type of the target building equipment and distribution feature information of the abnormality type in a time dimension according to the abnormality identification result, generating a target early warning instruction comprising a time positioning identifier based on the abnormality type and the time distribution feature information, and sending the equipment early warning instruction to a target equipment management terminal.

Inventors

XUE SHIWEI
ZHANG ZHIJIE
YAO ZIXUAN
SUN JIAN
WANG HAIBIN

Assignees

中建生态环境集团有限公司
中建碳科技有限公司

Dates

Publication Date: 20260505
Application Date: 20250703

Claims (8)

1. A building equipment anomaly identification method based on machine learning, the method comprising: acquiring a continuous operation data set of target building equipment, wherein the continuous operation data set comprises a multi-section equipment state recording unit with a timestamp mark; performing time-frequency domain feature extraction processing on the continuous operation data set to obtain a time-frequency domain feature set of the equipment state recording unit; Invoking a pre-built hybrid machine learning model to perform anomaly detection processing on the time-frequency domain feature set, and generating an anomaly identification result of the equipment state recording unit; Determining the abnormal type of the target building equipment and the distribution characteristic information of the abnormal type in the time dimension according to the abnormal recognition result, wherein the method comprises the steps of analyzing the abnormal characteristic descriptors in the abnormal recognition result, extracting a time-frequency domain characteristic mode corresponding to the abnormal characteristic descriptors, carrying out matching processing on the time-frequency domain characteristic mode and a preset abnormal type characteristic library, determining the abnormal type corresponding to the abnormal characteristic descriptors, extracting timestamp information of the abnormal characteristic descriptors in the abnormal recognition result, counting occurrence frequencies of different abnormal types in different time intervals, carrying out time sequence analysis processing on the occurrence frequencies of the abnormal types, and identifying an active period and a non-active period of the abnormal type in the time dimension; The method comprises the steps of carrying out time sequence analysis processing on occurrence frequency of an abnormal type, identifying an active period and a non-active period of the abnormal type in a time dimension, carrying out sliding window statistics processing on the occurrence frequency time sequence of the abnormal type, calculating average occurrence frequency in each time window, comparing average occurrence frequency differences of adjacent time windows, identifying a time point with occurrence frequency increase as an active period starting point, identifying a time point with occurrence frequency decrease as an active period ending point, carrying out boundary smoothing processing on the active period and the non-active period, eliminating period division errors caused by accidental fluctuation, and storing the smoothed active period and the non-active period in association with the abnormal type to form time distribution characteristic information of the abnormal type; Generating a target early warning instruction containing a time positioning identifier based on the abnormal type and the time distribution characteristic information, and sending the target early warning instruction to a target equipment management terminal; The method comprises the steps of analyzing a preset early warning rule base corresponding to an abnormal type, extracting early warning priority identification and maintenance strategy codes associated with the abnormal type, extracting active time periods and non-active time periods in the time distribution characteristic information, determining the current active time periods of the abnormal type, calculating the probability of re-occurrence of the abnormal type in a specified follow-up time period according to the current active time periods, carrying out association weighting processing on the probability of re-occurrence and the early warning priority identification of the abnormal type, generating dynamic early warning grade parameters, carrying out information fusion processing on the dynamic early warning grade parameters, maintenance strategy codes and the current active time periods, generating a target early warning instruction comprising the time positioning identification, carrying out format standardization processing on the target early warning instruction, and obtaining a standardized early warning instruction, wherein the standardized early warning instruction is matched with the communication protocol requirements of a target equipment management terminal.
2. The machine learning based building equipment anomaly identification method according to claim 1, wherein the performing time-frequency domain feature extraction processing on the continuous operation data set to obtain a time-frequency domain feature set of the equipment state recording unit includes: performing time window division processing on the continuous operation data set to obtain a plurality of data segmentation units with continuous time sequence relations, wherein each data segmentation unit corresponds to a device state record with fixed duration; carrying out time domain feature extraction processing on each data segment unit, and extracting fluctuation amplitude features, change rate features and periodic fluctuation features of equipment state parameters in the data segment units as time domain feature subsets; Performing frequency domain conversion processing on each data segment unit, and converting a time sequence signal of the data segment unit into a frequency domain signal representation; Performing energy distribution analysis processing on the frequency domain signal representation, and extracting energy duty ratio features, main frequency component features and frequency component stability features of different frequency intervals as frequency domain feature subsets; And carrying out association matching processing on the time domain feature subsets and the features of the corresponding data segmentation units in the frequency domain feature subsets to generate a time-frequency domain feature set of each equipment state recording unit.
3. The machine learning based building equipment anomaly identification method according to claim 2, wherein the performing a time domain feature extraction process on each data segment unit extracts a fluctuation amplitude feature, a change rate feature, and a periodic fluctuation feature of equipment status parameters in the data segment unit as a time domain feature subset, and includes: Calculating the difference value between the maximum value and the minimum value of the equipment state parameters in the data segmentation unit to obtain the fluctuation amplitude characteristic of the data segmentation unit; Calculating the ratio of the difference value of the equipment state parameters at adjacent time points in the data segmentation unit to the time interval to obtain the change rate characteristic of the data segmentation unit; performing autocorrelation analysis processing on the equipment state parameter sequence of the data segmentation unit, and identifying the period length of a repeated mode in the sequence to obtain the periodic fluctuation characteristic of the data segmentation unit; Normalizing the fluctuation amplitude characteristic, the change rate characteristic and the periodic fluctuation characteristic to eliminate the dimension difference among different parameters; Combining the normalized fluctuation amplitude characteristic, the change rate characteristic and the periodic fluctuation characteristic into a time domain characteristic subset of the data segmentation unit; The frequency domain conversion processing is performed on each data segment unit, and the time sequence signal of the data segment unit is converted into a frequency domain signal representation, which comprises the following steps: Windowing is carried out on the time sequence signals of the data segmentation units so as to inhibit the frequency spectrum leakage phenomenon of the signal edges, and the windowed time sequence signals are obtained; Processing the windowed time series signal based on a fast Fourier transform algorithm, and converting the time domain signal into a frequency domain signal; Extracting real part and imaginary part information of the frequency domain signal to generate a frequency distribution list containing frequency values and corresponding amplitude values; Performing smoothing filtering processing on the frequency distribution list to eliminate high-frequency noise interference and obtain a smooth frequency domain signal representation; and storing the smooth frequency domain signal representation and the time stamp information of the data segmentation unit in an associated mode to form a frequency domain signal representation with a time-frequency corresponding relation.
4. The machine learning based building equipment anomaly identification method according to claim 1, wherein the invoking the pre-built hybrid machine learning model to perform anomaly detection processing on the time-frequency domain feature set, generating an anomaly identification result of the equipment state recording unit includes: Inputting the time-frequency domain feature set into an LSTM neural network of the hybrid machine learning model, and modeling the time sequence dependency of the time-frequency domain feature set to generate a sequence feature representation with time sequence association information; inputting the sequence feature representation into an isolated forest algorithm of the hybrid machine learning model, performing outlier detection processing on the sequence feature representation, and identifying abnormal feature points deviating from a normal mode in the sequence feature representation; performing context association analysis processing on the abnormal feature points, and judging whether the abnormal feature points are continuous abnormality or sporadic abnormality by combining the feature information of the adjacent equipment state recording units in the time-frequency domain feature set; If yes, carrying out quantitative evaluation processing on the abnormality degree of the abnormal feature points to generate an abnormal feature descriptor containing abnormal confidence coefficient; And carrying out association mapping processing on the abnormal feature descriptor and the timestamp information of the equipment state recording unit to generate an abnormal identification result of the equipment state recording unit.
5. The machine learning based building equipment anomaly identification method of claim 4, wherein the inputting the time-frequency domain feature set into the LSTM neural network of the hybrid machine learning model models the time-sequence dependencies of the time-frequency domain feature set to generate a sequence feature representation with timing-related information, comprising: Performing characteristic dimension alignment processing on the time-frequency domain characteristic set to unify the characteristic dimension quantity of different equipment state recording units; inputting the time-frequency domain feature set with aligned dimensions into an input layer of the LSTM neural network according to the time stamp sequence, and generating an initial feature input vector; performing history information memory processing on the initial characteristic input vector through a memory unit of the LSTM neural network, and reserving characteristic information of a state recording unit of the preamble equipment to obtain history information; Screening the historical information through a forgetting gate of the LSTM neural network, and filtering noise historical characteristic information to obtain screened historical information; and carrying out fusion processing on the screened historical information and the current characteristic input vector through an output gate of the LSTM neural network to generate a sequence characteristic representation containing time sequence associated information.
6. The machine learning-based building equipment anomaly identification method according to claim 1, wherein the performing information fusion processing on the dynamic early warning level parameter, the maintenance policy code and the current active time interval to generate a target early warning instruction including a time positioning identifier includes: distributing corresponding early warning identification symbols for the dynamic early warning grade parameters, wherein the early warning identification symbols and the early warning grade parameters are in positive correlation; The maintenance strategy codes are converted into maintenance operation description texts containing maintenance step descriptions; converting the current active time interval into a time range descriptor, wherein the time range descriptor comprises a starting time point and an ending time point; Carrying out structural combination processing on the early warning identification symbol, the maintenance operation description text and the time range descriptor to generate a target early warning instruction main body containing a time positioning identification; adding equipment identification information for the target early warning instruction main body, wherein the equipment identification information is used for uniquely identifying the target building equipment; and performing verification coding processing on the target early warning instruction main body and the equipment identification information to generate a target early warning instruction with an error verification function.
7. The machine learning based building equipment anomaly identification method of claim 6, wherein said transcoding the maintenance policy into a maintenance operation descriptive text containing a maintenance step description comprises: Acquiring a multi-section type coding structure of the maintenance strategy code, wherein the multi-section type coding structure consists of a maintenance action type identification section, an action object identification section and an operation condition identification section which are sequentially arranged; Invoking a pre-constructed maintenance action knowledge base, wherein the maintenance action knowledge base stores basic step description units corresponding to the maintenance action type identification sections one by one, equipment component positioning description units corresponding to the action object identification sections one by one and execution constraint description units corresponding to the operation condition identification sections one by one; Performing segmentation analysis processing on the multi-section type coding structure, and extracting coding values of the maintenance action type identification section, the action object identification section and the operation condition identification section; Based on the coding value of the maintenance action type identification section, matching and acquiring a corresponding basic step description unit from the maintenance action knowledge base, wherein the basic step description unit comprises standard action verbs and general operation direction information; Based on the coding value of the action object identification section, matching and acquiring a corresponding equipment component positioning description unit from the maintenance action knowledge base, wherein the equipment component positioning description unit comprises a component name and space position information; Based on the coding value of the operation condition identification section, matching and acquiring a corresponding execution constraint description unit from the maintenance action knowledge base, wherein the execution constraint description unit comprises environment parameter range requirements and safety notice information; performing semantic fusion processing on the standard action verb of the basic step description unit and the component name and space position information of the equipment component positioning description unit to generate an action description clause containing an operation object; performing condition association processing on the action description clause, the environment parameter range requirement of the execution constraint description unit and the safety notice information to generate a step constraint description clause containing operation preconditions; according to the arrangement sequence of the multi-section type coding structure, carrying out logic series connection processing on the action description clause and the step constraint description clause to generate a maintenance step description sequence with sequence relevance; And carrying out natural language fluency optimization processing on the maintenance step description sequence, adjusting the grammar structure of the sentence and the use of the connecting word, and generating a maintenance operation description text containing complete operation logic.
8. A building equipment anomaly identification system comprising a processor and a memory, wherein the memory stores a computer program which, when executed by the processor, causes the processor to perform the steps of the method of any one of claims 1 to 7.

Description

Building equipment abnormality identification method and system based on machine learning Technical Field The application belongs to the technical field of machine learning, and particularly relates to a building equipment abnormality identification method and system based on machine learning. Background In the existing building equipment management technology, a plurality of defects exist in detection and early warning means of equipment abnormality. The traditional technology only depends on data characteristic analysis of a single dimension, can not comprehensively grasp the running condition of equipment, and is easy to miss complicated and changeable abnormal conditions. Although some technologies can acquire equipment operation data, the depth mining of the data in the time dimension is lacking, and the occurrence time and type of the abnormality are difficult to accurately locate. The method is characterized in that the method is based on a simple model, and is low in detection accuracy and reliability, and false detection and missing detection are frequent in the face of complex equipment operation modes and diversified abnormal conditions. In addition, the prior art mostly fails to effectively integrate abnormal information and time distribution characteristics, so that generated early warning instructions lack accurate time positioning, so that management personnel are difficult to timely and purposefully maintain equipment, the influence of equipment faults on normal operation of a building is increased, and the service life and the operation efficiency of the equipment are reduced. Disclosure of Invention The application provides a building equipment abnormality identification method and system based on machine learning, which are used for realizing accurate detection and accurate early warning of building equipment abnormality, improving equipment management efficiency and ensuring stable operation of equipment. In a first aspect, an embodiment of the present application provides a building equipment anomaly identification method based on machine learning, which is applied to a building equipment anomaly identification system, and the method includes: acquiring a continuous operation data set of target building equipment, wherein the continuous operation data set comprises a multi-section equipment state recording unit with a timestamp mark; performing time-frequency domain feature extraction processing on the continuous operation data set to obtain a time-frequency domain feature set of the equipment state recording unit; Invoking a pre-built hybrid machine learning model to perform anomaly detection processing on the time-frequency domain feature set, and generating an anomaly identification result of the equipment state recording unit; Determining the abnormal type of the target building equipment and the distribution characteristic information of the abnormal type in the time dimension according to the abnormal identification result; And generating a target early warning instruction containing a time positioning identifier based on the abnormal type and the time distribution characteristic information, and sending the equipment early warning instruction to a target equipment management terminal. In a second aspect, an embodiment of the present application provides a building equipment anomaly identification system, including a processor and a memory, where the memory stores a computer program, and when the computer program is executed by the processor, causes the processor to execute the steps of the above method. In a third aspect, embodiments of the present application provide a computer readable storage medium comprising a computer program for causing a building equipment anomaly identification system to perform the steps of the above method when the computer program is run on the building equipment anomaly identification system. The embodiment of the application integrally constructs a complete and efficient building equipment abnormality identification and early warning system. The time-frequency domain feature extraction processing can deeply mine data features from two dimensions of a time domain and a frequency domain, overcomes the limitation of single-dimension analysis, more comprehensively characterizes the operation features of the equipment and greatly improves the capturing capability of abnormal features. And secondly, the pre-constructed hybrid machine learning model combines the advantages of various algorithms, and performs anomaly detection on the time-frequency domain feature set, so that the accuracy and reliability of anomaly detection are remarkably improved, and the false detection and omission detection conditions are effectively reduced. And determining the abnormal type and the time-space distribution characteristic information according to the abnormal identification result, so that a manager can clearly grasp the nature and the development rule of the abnormali