Fault monitoring system for electromechanical installation engineering

Abstract

The invention belongs to the technical field of electrical equipment fault monitoring, and discloses a fault monitoring system for electromechanical installation engineering, which comprises a high-impedance fault identification module, extracts nonlinear harmonic features through current signals, constructs a frequency domain-time domain joint feature set in combination with space-time magnetic distortion in an electromechanical installation process, and performs fault identification on the nonlinear harmonic features through the high-impedance fault identification module. Non-linear characteristic response identification of the high-impedance fault is carried out, and a high-impedance fault characterization result is output; the transient fault capturing module is used for positioning a risk time window according to a high-impedance fault characterization result, and identifying the starting and ending boundaries of an electrical disturbance behavior in combination with a double-threshold mutation mechanism; synchronously extracting corresponding operation condition parameters, constructing a short-time process behavior model taking electrical disturbance as a source, and outputting a transient disturbance sequence; and the sensitivity of fault identification, the accuracy of positioning and the real-time performance of diagnosis are integrally improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical equipment fault monitoring, more specifically, the present application relates to a fault monitoring system for electromechanical installation engineering. BACKGROUND

[0002] The patent with the patent publication number CN116663875A discloses an intelligent installation engineering monitoring system and method based on big data, which includes an installation database establishment module, a to-be-analyzed installation link determination module, a target installation link determination module, a to-be-analyzed traceability set generation module, an influence data extraction module, an influence link determination module, and a warning response module. The installation database establishment module is used for digitally recording and storing the installation process. The to-be-analyzed installation link determination module analyzes and extracts the to-be-analyzed installation link recorded in the installation database. The target installation link determination module is used for analyzing and extracting the target installation link recorded in the installation database based on the installation database. The to-be-analyzed traceability set generation module is used to determine the to-be-analyzed sequence of the installation process to which the to-be-analyzed installation link belongs. The present application improves the efficiency of workers in analyzing the overall installation engineering, and makes the workers' handling of abnormalities more targeted.

[0003] The existing fault monitoring system for electromechanical installation engineering mainly has the following problems: The existing electromechanical installation fault monitoring technology generally has the defects of "single signal analysis, static model positioning, centralized processing, and low communication efficiency". The recognition ability for low current level and non-linear high impedance faults (such as poor grounding and contact arc) is weak, resulting in high missed detection rate and high false alarm rate. The duration of transient faults (such as discharge flashover and contact jump) is extremely short. The sampling rate of conventional monitoring systems is insufficient, and there is a lack of event triggering mechanism, which cannot preserve the original disturbance waveform, making it difficult to reproduce. In the multi-device parallel structure, the electrical signals are coupled and disturbed, and the fault source may be incorrectly classified into adjacent branches. Traditional monitoring methods cannot achieve accurate positioning.

[0004] In view of this, the present application provides a fault monitoring system for electromechanical installation engineering to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a fault monitoring system for electromechanical installation engineering, comprising: A high impedance fault recognition module extracts non-linear harmonic features from current signals, combines the space-time magnetic distortion in the electromechanical installation process, constructs a frequency-time domain joint feature set, and performs non-linear feature response recognition of high impedance faults to output high impedance fault representation results. The transient fault capturing module locates a risk time window according to the high-impedance fault characterization result, identifies the start and end boundaries of the electrical disturbance behavior in combination with a double-threshold mutation mechanism, synchronously extracts corresponding operating condition parameters, constructs a short-time process behavior model with the electrical disturbance as a source, and outputs a transient disturbance sequence. The fault source positioning module fuses the high-impedance fault characterization result and the transient disturbance sequence, constructs a dynamic fingerprint response matrix in a multi-device parallel environment, introduces a signal blind source separation mechanism to eliminate signal coupling interference between parallel devices, and in combination with a preset device characteristic spectrum library, identifies a fault attribution branch and generates source positioning path information. The multi-modal analysis module constructs a distributed fault data collaborative processing architecture, performs local inference on the edge side, generates a structured fault event, and obtains a time-consistent fault analysis and diagnosis result. The coordinated control module interfaces a management terminal and transmits the fault analysis and diagnosis result based on lightweight communication and intelligent publishing technology.

[0006] Preferably, the method for obtaining the nonlinear harmonic feature comprises: The current signal of the end device in the electromechanical installation process is sampled in real time to obtain a sampled current signal set; the sampled current signal set is stored in a current data buffer after anti-aliasing filtering, and is sliced according to a preset time window; Based on the sampled current signal in each time window, a bispectrum atlas is constructed, and a frequency domain statistical feature is extracted; by quantifying the complexity of the current signal, the sampled current signal set in each time window is discretized into a one-dimensional time sequence, and the embedding dimension and the tolerance threshold are predefined; by constructing a subsequence of the one-dimensional time sequence and calculating the maximum distance between all subsequences, the proportion of similar subsequences is counted under the predefined tolerance threshold, and a sample entropy value is calculated; the sample entropy values in all time windows are collected and integrated to obtain sample entropy value data; the frequency domain statistical feature and the sample entropy value data are integrated to obtain the nonlinear harmonic feature.

[0007] Preferably, the method for constructing the frequency domain-time domain joint feature set comprises: A magnetic field sensor is arranged at a key node of the end device in the electromechanical installation, and the intensity data and direction data of the magnetic field during the operation of the device are sampled in real time; the key node of the end device includes a bus connection point, a switch device interface, a circuit breaker interface, a cable joint, a terminal connection point and a grounding device; The collected intensity data and direction data are subjected to time domain analysis to extract magnetic field time sequence data; the nonlinear harmonic feature and the magnetic field time sequence data are time step aligned with the sampling time stamp as an index, and are spliced and combined under the same time window to form a unified frequency domain-time domain joint feature set.

[0008] Preferably, the method for obtaining the high-impedance fault characterization result comprises: According to the frequency-time domain combined feature set, a high-impedance fault occurrence probability value is calculated; when the high-impedance fault occurrence probability value exceeds a preset high-impedance fault occurrence probability threshold value, it is determined that a high-impedance fault occurs, and a structured high-impedance fault characterization result is generated; the high-impedance fault characterization result comprises the high-impedance fault occurrence probability value and a fault start time stamp.

[0009] Preferably, the method for positioning the risk time window comprises: based on the fault start time stamp in the high-impedance fault characterization result, a preset time width is extended forward and backward to set the risk time window.

[0010] Preferably, the method for identifying the start and end boundaries of the electrical disturbance behavior comprises: Within the risk time window, electrical waveform data is obtained by continuous sampling, and after denoising filtering, it is used to determine the electrical disturbance boundary; by jointly considering the waveform instantaneous jump amplitude and the current change rate, the occurrence time and duration of the electrical disturbance behavior are determined, and then the start point and end point of the electrical disturbance behavior are identified; When the electrical disturbance boundary determination value is 1, it indicates that the time point is identified as an electrical disturbance boundary candidate point, all time points satisfying the electrical disturbance boundary determination value of 1 are constructed into an electrical disturbance boundary candidate set, all identified electrical disturbance boundary candidate points are clustered and processed, adjacent electrical disturbance boundary candidate point segments are merged, and an electrical disturbance boundary candidate point segment interval is obtained. For each electrical disturbance boundary candidate point segment interval, the time point at which the first electrical disturbance boundary determination value of the electrical disturbance boundary candidate point segment is 1 is defined as the start time point, and the time point at which the last electrical disturbance boundary determination value of the electrical disturbance boundary candidate point segment is 1 is defined as the end time point, and the electrical disturbance boundary candidate point segment interval is output as an effective electrical disturbance behavior start and end time interval; all effective electrical disturbance behavior start and end time intervals are collected to form the complete start and end boundaries of the electrical disturbance behavior.

[0011] Preferably, the method for obtaining the transient disturbance sequence comprises: After the start and end boundaries of the electrical disturbance behavior are identified, based on the start time point and the end time point of the electrical disturbance behavior, a target time window is constructed with the start time point and the end time point as the center and extending a preset buffer time T forward and backward; the operating condition parameters within the target time window are extracted; The electrical waveform data and the operating condition parameters are timestamped and processed to construct a short-time process behavior model with the electrical disturbance as the source, and an electrical disturbance driven transient disturbance sequence is generated.

[0012] Preferably, the method for constructing the dynamic fingerprint response matrix comprises: Based on the high-impedance fault characterization result and the transient disturbance sequence, and based on the device-fault feature mapping relationship, the running state and the corresponding topological position of each terminal device in the current parallel branch are determined in a target time window through the SCADA platform and the preset device topology database, the local response signal of each terminal device is extracted, and a device disturbance response vector is formed; and based on the device disturbance response vector, a dynamic fingerprint response matrix for a multi-device parallel operation scene is constructed.

[0013] Preferably, the method of identifying the fault attribution branch and generating source positioning path information comprises: A blind source separation mechanism is introduced to eliminate the signal coupling interference between different terminal devices in the parallel environment, and the blind source separation mechanism comprises constructing the local response signal of each terminal device in the dynamic fingerprint response matrix into a mixed signal matrix; An independent component analysis algorithm is used to separate the source signals of the mixed signal matrix, m potential fault source signals with independent statistical characteristics are obtained; based on the similarity between each potential fault source signal obtained by separation and a preset device feature spectrum library; the target terminal device to which the potential fault source signal belongs is determined, and the source positioning path information is generated in combination with the topological structure of the terminal device provided by the preset device topology database.

[0014] Preferably, the fault analysis and diagnosis result comprises a fault type, a fault event ID, a fault start and end timestamp, a fault position positioning, a fault duration, a structured fault event identification, and a fault handling suggestion.

[0015] Compared with the prior art, the present application has the following beneficial effects: The present application jointly considers the current instantaneous jump amplitude and the current rate of change, can identify disturbance behaviors with large amplitude or fast change at the same time, takes into account both amplitude anomaly and derivative mutation characteristic modes, and effectively improves the overall perception ability of the system to different types of disturbance events. The double-threshold mutation identification mechanism is particularly suitable for detecting weak disturbances and short-time nonlinear changes caused by high-impedance faults, significantly improves the detection sensitivity and accuracy of high-impedance faults, and makes up for the deficiency of the traditional single-threshold method in the high-impedance scene that is prone to miss detection; the disturbance boundary judgment value is used to construct a disturbance candidate point set, and clustering processing and disturbance segment merging are performed based on time proximity, effectively solving the problems of scattered fragments, unclear start and end in the original disturbance detection result, improving the structural integrity and logical continuity of the disturbance behavior, enhancing the timeliness and accuracy of transient fault capture, and reducing the misjudgment rate and fragmentation risk.

[0016] By setting the disturbance interval threshold and the minimum duration threshold, the ability to automatically exclude pseudo-disturbances, short-term noise and non-fault fluctuations is further enhanced, thereby improving the diagnostic stability and robustness of the system in a strong interference environment, and achieving high-precision fault source positioning in a multi-device parallel system; boundary identification is performed based on the real dynamic response process of the current waveform, with strong adaptability, and the identification range can be dynamically adjusted according to different equipment types, operating conditions and fault evolution characteristics; the use of disturbance boundary discrimination values ​​can achieve precise positioning of the starting and ending points of the disturbance, providing more timely and spatially resolved basic data support for disturbance source modeling, fault attribution analysis and response control. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic structural diagram of a fault monitoring system for an electromechanical installation project of the present invention; Figure 2 A schematic flow chart of a fault monitoring method for an electromechanical installation project according to the present invention; Figure 3 This is a flow chart of the method for identifying the start and end boundaries of electrical disturbance behavior provided by the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] Example 1 See also Figure 1 and Figure 3 As shown, this embodiment 1 further illustrates a fault monitoring system for a mechanical and electrical installation project proposed by the present invention, including: With the widespread application of electromechanical installation engineering in fields such as power generation, rail transit, and industrial automation, electrical system structures are becoming increasingly complex, and the number of parallel systems between devices is increasing. This places higher demands on the real-time, accurate, and discriminative capabilities of fault monitoring. Especially in high-voltage power supply systems, large-scale load switching scenarios, and multi-terminal parallel operation environments, electrical faults often exhibit atypical characteristics such as short disturbances, low energy, and weak responses. Existing monitoring methods are no longer able to meet the actual needs of these projects.

[0020] The existing electromechanical installation engineering fault monitoring system mainly adopts an amplitude threshold detection method based on current mutation, and the judgment basis is usually whether the amplitude of the current value at a moment relative to the previous moment exceeds the set threshold. However, in typical scenarios such as high-impedance faults or slight contact failure, the amplitude of the current change is small, but the change rate (i.e. the derivative of the current) may suddenly increase. The traditional recognition mechanism based on a single amplitude criterion cannot accurately capture such transient characteristics, and is prone to cause missed detection, delayed reporting or even complete failure to identify problems. For example, when the disturbance current does not significantly exceed the limit but its derivative presents a sharp peak feature, the system will incorrectly judge that there is no abnormality, missing the key signs in the early stage of the fault.

[0021] During the operation of the electromechanical system, normal operations such as periodic start-stop of equipment, load switching, motor start / stop, etc. may cause short-time current waveform disturbances. Such disturbances are not caused by faults, but their transient characteristics are often similar to fault signal patterns. The existing methods lack a comprehensive discrimination mechanism that integrates derivative, duration, disturbance pattern and other multi-dimensional parameters, often misjudging such operational disturbances as abnormalities, causing a large number of false positives and reducing the system's practicality and stability.

[0022] The existing system generally uses a fixed time window clipping method to capture fault segments in the disturbance behavior extraction process, making it difficult to adaptively identify the true start and end points of the disturbance. The boundaries of the disturbance segment and the actual dynamic behavior are often inconsistent, seriously affecting the accuracy of subsequent behavior modeling, feature analysis and fault tracing. At the same time, for noise disturbances or environmental interference signals with extremely short duration and non-obvious derivative changes, the existing system also lacks an effective disturbance segment validity judgment mechanism, which can easily misclassify invalid disturbances into the analysis process, causing resource waste and diagnostic performance degradation.

[0023] In particular, in complex power distribution systems, multiple devices often operate in parallel, and the current signals of multiple branches are coupled and superimposed, resulting in signals collected at a single measurement point often being a mixture of multiple source signals. The existing technology generally does not introduce effective signal blind source separation or multi-channel decoupling algorithms, making it difficult to distinguish the response characteristics of different devices from the mixed signals, and the fault attribution judgment is ambiguous, which seriously restricts the accurate positioning and deployment of subsequent maintenance strategies.

[0024] To effectively solve the above problems, the present application proposes a fault monitoring system for electromechanical installation engineering, comprising: A high-impedance fault recognition module extracts nonlinear harmonic features from the current signal, combines the spatiotemporal magnetic distortion in the electromechanical installation process, constructs a frequency-time domain joint feature set, and performs nonlinear feature response recognition of high-impedance faults, outputting high-impedance fault representation results; The transient fault capturing module locates a risk time window according to the high-impedance fault characterization result, identifies the start and end boundaries of the electrical disturbance behavior in combination with a double-threshold mutation mechanism; synchronously extracts corresponding operating condition parameters, constructs a short-time process behavior model with the electrical disturbance as the source, and outputs a transient disturbance sequence; The fault source positioning module fuses the high-impedance fault characterization result and the transient disturbance sequence, constructs a dynamic fingerprint response matrix in a multi-device parallel environment, introduces a signal blind source separation mechanism to eliminate signal coupling interference between parallel devices, and in combination with a preset device characteristic spectrum library, identifies a fault attribution branch and generates source positioning path information; The multi-modal analysis module constructs a distributed fault data collaborative processing architecture, and performs local inference on the edge side to generate a structured fault event and obtain a time-consistent fault analysis and diagnosis result; The coordinated control module interfaces a management terminal and transmits the fault analysis and diagnosis result based on lightweight communication and intelligent publishing technology.

[0025] The method for acquiring the nonlinear harmonic feature comprises the following steps: The high-impedance fault is characterized by weak amplitude and strong nonlinear current disturbance, and the frequency spectrum often contains odd high-frequency harmonics, non-integer sub-harmonics and background noise coupling distortion. Therefore, a frequency domain analysis method is used to extract nonlinear harmonic features from the collected current signal, and through nonlinear mode analysis, early identification of high-impedance type weak arc faults is realized, including the following steps: By deploying a wideband response type current sensor (such as a Hall sensor or a Rogowski coil), the current signal of the end device during the mechanical and electrical installation process is sampled in real time, and a set of sampled current signals is obtained; the sampling frequency needs to meet the Nyquist criterion to ensure that the high-frequency harmonic components are covered; after anti-aliasing filtering, the set of sampled current signals is stored in a current data buffer, and is sliced according to a preset time window for subsequent frequency domain analysis; In the mechanical and electrical installation engineering, the "end device" refers to the final energy-using device or control device located at the load end or execution end of the system, directly connected to the energy-using, energy-using, and control objects. These devices are one of the most frequent fault occurrence, most scattered management, and most complex operation parts of the entire mechanical and electrical system. According to different types of mechanical and electrical systems, the end devices are slightly different, for example: Electrical end devices, suitable for power supply and distribution systems and automation systems: electric motors, frequency converters, soft starters (such as fan, pump, and compressor drive motors); end control boxes, power distribution boxes; sockets, circuit breakers, contactors, relays; end luminaires of lighting systems, dimming controllers; HVAC system end devices: fan coil units (FCU), VAV end boxes; electric air valves, electric water valves, temperature controllers; air purification end devices, humidifiers, air supply outlets; On the basis of obtaining the denoised current signal, a time-frequency joint analysis method is used to extract nonlinear harmonic features: The sampling current signal in each time window is respectively subjected to Fourier transform, and third-order cumulant calculation is performed in combination with bispectrum analysis to construct a bispectrum atlas to extract frequency domain statistical features, including low-order odd harmonics (3rd, 5th, 7th, etc.) and their change rates, spectral energy density, harmonic distortion coefficient, bispectrum amplitude and bispectrum phase, so as to capture the nonlinear disturbance mode caused by arc or poor contact; the bispectrum atlas quantifies the phase coupling strength between each order of harmonics, and when high impedance fault occurs, the nonlinear coupling between frequencies is enhanced due to arc discharge, which is manifested as the appearance and amplitude increase of the characteristic peak in the bispectrum atlas, which can be used as an important feature to distinguish between normal operation and nonlinear fault state; High impedance faults will introduce randomness (such as arc discharge), which will cause significant changes in the entropy value of the current signal. In order to enhance the recognition ability of nonlinear harmonic structure, the approximate entropy is introduced to quantify the complexity of the current signal. The sampling current signal in each time window is respectively discretized into a one-dimensional time series, and the embedding dimension and tolerance threshold are predefined as the matching judgment basis for similar subsequences; by constructing the subsequence of the one-dimensional time series and calculating the maximum distance between all subsequences (if the distance is not greater than the tolerance threshold, it is considered that the two subsequences are similar), the proportion of similar subsequences is counted under the predefined tolerance threshold, and the sample entropy value is calculated; the sample entropy values in all time windows are collected and integrated to obtain the sample entropy value data; the frequency domain statistical features and the sample entropy value data are integrated to obtain the nonlinear harmonic features.

[0026] The method for constructing the frequency-time domain joint feature set includes: Magnetic field sensors are arranged at key nodes of the end device of the mechanical and electrical installation to sample the intensity data and direction data of the magnetic field in real time during the operation of the device to obtain the dynamic changes of the magnetic field in space and time; the key nodes of the end device include bus connection points, switch device interfaces, circuit breaker interfaces, cable joints, terminal connection points and grounding devices; Time domain analysis is performed on the collected intensity data and direction data to extract magnetic field time series data (mean, variance, peak value and instantaneous change rate of intensity data and direction data) reflecting the transient disturbance and abnormal fluctuation of the magnetic field; the nonlinear harmonic features and the magnetic field time series data are time step aligned, and are spliced and combined under the same time window to form a joint feature vector; the joint feature vector is standardized to form a unified frequency-time domain joint feature set.

[0027] The method for obtaining the high impedance fault characterization result includes: The obtained frequency-time domain combined feature set is input into a pre-trained convolutional neural network, a convolutional layer is used to perform convolution operation on the frequency-time domain combined feature set, a pooling layer is used to perform down-sampling on the convolution result, and a Softmax function is used to output a high impedance fault occurrence probability value; when the high impedance fault occurrence probability value exceeds a preset high impedance fault occurrence probability threshold, it is determined that a high impedance fault occurs, and a structured high impedance fault representation result is generated; the high impedance fault representation result includes the high impedance fault occurrence probability value and a fault start timestamp.

[0028] The positioning method of the risk time window comprises: Based on the fault start timestamp in the high impedance fault representation result, a preset time width (such as ±100 ms) is extended forward and backward to set a risk time window to cover the time period before and after the fault occurs.

[0029] The method for identifying the start and end boundaries of the electrical disturbance behavior comprises: Within the risk time window, the waveform collector is triggered to continuously sample the electrical waveform to obtain continuous electrical waveform data, which is saved in real time through a ring buffer structure to ensure the integrity of the electrical waveform before and after sampling; the electrical waveform data refers to the original waveform sequence of the electrical quantity varying with time collected by the high-speed ADC (waveform collector) within the risk time window; according to the common end load type and fault feature detection requirements in mechanical and electrical installation engineering, the electrical waveform data includes current waveform data (three-phase current waveform, zero-sequence current waveform and differential mode current) and voltage waveform data (three-phase voltage waveform, phase voltage and line voltage and voltage zero-sequence component); The electrical waveform data is subjected to denoising filtering, and a double-threshold mutation recognition function is constructed to identify the electrical disturbance boundary of the denoising filtered electrical waveform data; by jointly considering the waveform instantaneous jump amplitude and the current change rate, the occurrence time and duration of the electrical disturbance behavior are determined, and then the start and end points of the electrical disturbance behavior are identified; The double-threshold mutation recognition function is: ; wherein, represents the electrical disturbance boundary discrimination value; represents a preset instantaneous jump amplitude threshold; represents a preset current change rate threshold; represents the current signal at time point ; represents the current signal at time point ; represents the derivative of the current signal with respect to time, i.e. the current change rate; It should be noted that, refers to other cases except : at time point the absolute value of the current signal at time point is less than or equal to a preset instantaneous jump amplitude threshold, and the current rate of change is greater than a preset current rate of change threshold; at time point the absolute value of the current signal at time point is greater than a preset instantaneous jump amplitude threshold, and the current rate of change is less than or equal to a preset current rate of change threshold; at time point the absolute value of the current signal at time point is less than or equal to a preset instantaneous jump amplitude threshold, and the current rate of change is less than or equal to a preset current rate of change threshold; When the electrical disturbance boundary discrimination value is 1, it indicates that the time point is identified as an electrical disturbance boundary candidate point. All time points satisfying the electrical disturbance boundary discrimination value of 1 are constructed into an electrical disturbance boundary candidate set. All identified electrical disturbance boundary candidate points are clustered and processed, adjacent electrical disturbance boundary candidate point segments are merged, and an electrical disturbance boundary candidate point segment interval is obtained. It should be noted that: if the time interval between two adjacent electrical disturbance boundary candidate point segments is less than a preset time interval threshold, it is determined that it is a continuous performance of the same electrical disturbance behavior, and the two adjacent electrical disturbance boundary candidate point segments are merged into one complete electrical disturbance boundary candidate point segment interval. If the duration of any one electrical disturbance boundary candidate point segment is less than a preset minimum electrical disturbance duration threshold, it is determined to be noise and is not included in the electrical disturbance behavior recognition. For each electrical disturbance boundary candidate point segment interval, the time point at which the first electrical disturbance boundary discrimination value of the electrical disturbance boundary candidate point segment is 1 is defined as the starting time point, and the time point at which the last electrical disturbance boundary discrimination value of the electrical disturbance boundary candidate point segment is 1 is defined as the ending time point. The electrical disturbance boundary candidate point segment interval is output as an effective electrical disturbance behavior start and end time interval. All effective electrical disturbance behavior start and end time intervals are collected to form the complete start and end boundaries of the electrical disturbance behavior.

[0030] Solve the following problems existing in the prior art: In electromechanical installation engineering, especially in high-voltage or complex power distribution system scenarios, the existing electrical fault monitoring means have the problem that most methods rely on a single threshold, such as a fixed amplitude mutation threshold, which cannot identify transient disturbances with small amplitude changes but large derivatives, is prone to miss detection or only detects amplitude overrun events, and ignores the dynamic evolution characteristics of the disturbance process. Without a multi-dimensional judgment mechanism, normal load switching, motor start-stop and other short-time non-fault fluctuations often occur during system operation, which are easily misjudged as disturbances. Most existing methods use fixed time window clipping to extract disturbance segments, which cannot adaptively identify the true start and end points of the disturbance, seriously affecting the subsequent modeling and positioning accuracy. There is a lack of disturbance segment effectiveness screening mechanism, and disturbances with short duration or insufficient derivative changes cannot be excluded, resulting in waste of analysis resources.

[0031] The beneficial effects of the prior art are: By jointly considering the current instantaneous jump amplitude and the current change rate, disturbances with large amplitude or fast change can be identified; it is especially suitable for detecting weak disturbances and short-time nonlinear changes caused by high-impedance faults, improving the detection sensitivity of the system to slight disturbances and marginal abnormalities; the disturbance boundary discriminant value is used to construct a disturbance candidate point set, and clustering and candidate segment merging are performed based on time proximity, reducing the misjudgment rate; the introduction of interval threshold and minimum duration threshold between disturbance segments can effectively exclude false disturbances or short-time noise segments, improving the diagnosis reliability; the start and end boundaries are adaptively identified according to the actual waveform dynamic behavior; the disturbance start and end time are accurately located using the disturbance boundary discriminant value, improving the timeliness of subsequent fault modeling and response control.

[0032] The method for obtaining the transient disturbance sequence comprises: After identifying the start and end boundaries of the electrical disturbance behavior, based on the start time point and the end time point of the electrical disturbance behavior, a closed interval is constructed with the start time point and the end time point as the center and extending a preset buffer time T forward and backward, to construct a target time window for ensuring complete capture of the disturbance precursor and the disturbance response; the data interface of the SCADA platform is called to extract the operating condition parameters within the corresponding target time window; the operating condition parameters include real-time voltage, current, active power, reactive power, load change information, switch operation state, automatic control instruction record and equipment operation identification; The electrical waveform data and operating condition parameters are timestamped and aligned to construct a short-time process behavior model with the electrical disturbance as the source, which is used to describe the dynamic correlation characteristics between waveform changes before and after the disturbance and the operating condition response; based on the short-time process behavior model, an electrical disturbance driven transient disturbance sequence is generated, which expresses the current signal value and the corresponding operating condition state vector at each time in the electrical disturbance process in the form of a time sequence.

[0033] The method for constructing a dynamic fingerprint response matrix comprises the following steps: Based on the high-impedance fault characterization result and the transient disturbance sequence, and based on the device-fault feature mapping relationship, the running state and the corresponding topological position of each terminal device in the current parallel branch are determined in a target time window through the SCADA platform and the preset device topology database, the local response signals (including the current change amplitude, the electrical disturbance response delay, the waveform mutation slope, and the frequency spectrum offset amplitude) of each terminal device are extracted, and a device disturbance response vector is formed; and based on the device disturbance response vector, a dynamic fingerprint response matrix for a multi-device parallel operation scene is constructed.

[0034] The method for identifying a fault attribution branch and generating source positioning path information comprises the following steps: Since multiple devices are operated in parallel, electrical disturbances can cause signal crosstalk on the common bus, and the observed local response signals need to be decoupled, and therefore the application adopts a blind source separation algorithm (such as independent component analysis ICA, non-negative matrix factorization NMF, or a decoupling network based on contrastive learning) to separate the coupled signals; The blind source separation mechanism is introduced to eliminate the signal coupling interference between different terminal devices in the parallel environment, and the blind source separation mechanism comprises constructing the local response signals of each terminal device in the dynamic fingerprint response matrix into a mixed signal matrix; An independent component analysis algorithm is used to separate the source signals from the mixed signal matrix, and m potential fault source signals with independent statistical characteristics are obtained; for each potential fault source signal obtained by separation, the waveform feature, the frequency spectrum feature, the response time delay, and the disturbance amplitude are extracted, and the similarity score between the potential fault source signal and the preset device feature spectrum library is calculated; the preset device feature spectrum library comprises the response frequency, the waveform feature, the phase characteristic, and the delay model of different devices under a preset disturbance condition; Based on the similarity score, the target terminal device to which the potential fault source signal belongs is determined, and the source positioning path information is generated in combination with the topological structure of the terminal device provided by the preset device topology database; the source positioning path information refers to: in a multi-device parallel electromechanical installation system, when a fault is located, the system generates a complete path information which is structured and can be used for maintenance guidance, and the content covers the physical / logical connection links between the system backbone and the specific devices where the fault occurs; and the path information specifically comprises the following three levels of information: The target terminal device refers to a specific device in the parallel structure which is confirmed to have a fault, and has the following attributes: device type, device unique number, installation position (such as power distribution cabinet number, unit position, field label), running state label (such as “abnormal”, “high-impedance alarm”), and associated control logic (such as whether the device participates in the SCADA closed-loop control); Fault attribution branch refers to the distribution circuit or control circuit path where the target end device is located, used to describe the electrical path of the device branching from the main busbar / supply node, including the branch number (such as BR-02), the upper-level device or feeder (such as MCB-F02), the position level in the branch electrical topology, the branch typical characteristics (such as bus current direction, branch capacitance, cable length), and whether it is a branch with parallel cross-coupling characteristics; Corresponding topology path node: refers to all intermediate connection nodes from the main power supply node to the target end device along the actual wiring topology of the electromechanical device, used for path visualization and maintenance guidance, such as busbar number, relay node, branch switch, distribution box number, cable number, sensor node, ground node, and spatial coordinates of each node.

[0035] Fault analysis and diagnosis results include fault type, fault event ID, fault start and end timestamp, fault location positioning, fault duration, structured fault event identification, and fault handling suggestions.

[0036] The preset high-impedance fault occurrence probability threshold is set by the staff based on historical data analysis results, and the historical analysis process includes the system collecting multiple data points of high-impedance fault occurrence probability and calculating the average value as a reference. It can be adjusted by the staff during system operation according to the actual situation; similarly, set the preset instantaneous jump amplitude threshold and the preset current change rate threshold.

[0037] In this embodiment, by jointly considering the current instantaneous jump amplitude and the current change rate, the disturbance behavior with large amplitude or fast change can be recognized at the same time, both amplitude anomaly and derivative mutation are considered, and the overall perception ability of the system to different types of disturbance events is effectively improved. This double-threshold mutation recognition mechanism is particularly suitable for detecting weak disturbances and short-time nonlinear changes caused by high-impedance faults, significantly improving the detection sensitivity and accuracy of high-impedance faults, and making up for the shortcomings of traditional single-threshold methods in high-impedance scenarios. The disturbance boundary discriminant value is used to construct the disturbance candidate point set, and based on the time proximity, clustering processing and disturbance segment merging are carried out, effectively solving the problems of scattered fragments, unclear start and end in the original disturbance detection results, improving the structural integrity and logical continuity of the disturbance behavior, enhancing the timeliness and accuracy of transient fault capture, and reducing the false positive rate and fragmentation risk.

[0038] By setting the disturbance section interval threshold and the minimum duration threshold, the automatic exclusion ability of false disturbances, short-time noises and non-fault fluctuations is further enhanced, thereby improving the diagnostic stability and robustness of the system in a strong interference environment, and realizing high-precision fault source positioning in a multi-device parallel system; According to the real dynamic response process of the current waveform, the boundary is identified, which has strong adaptability and can dynamically adjust the identification range according to different device types, operating states and fault evolution characteristics; The disturbance boundary discriminant value can be used to realize accurate positioning of the starting point and ending point of the disturbance, and provide more time-effective and spatially-resolved basic data support for disturbance source modeling, fault attribution analysis and response control.

[0039] Embodiment 2 Please refer to Figure 2 As shown in the drawings, some parts of the embodiment are not described in detail in the description of Embodiment 1, and a fault monitoring method for mechanical and electrical installation engineering is provided, comprising: S1, extracting nonlinear harmonic features from current signals, combining with the space-time magnetic distortion in the mechanical and electrical installation process, constructing a frequency-time domain combined feature set, and identifying nonlinear feature responses of high impedance faults, and outputting high impedance fault characterization results; S2, positioning the risk time window according to the high impedance fault characterization results, identifying the start and end boundaries of the electrical disturbance behavior combined with the double threshold mutation mechanism; Synchronously extracting corresponding operating condition parameters, constructing a short-time process behavior model with the electrical disturbance as the source, and outputting the transient disturbance sequence; S3, by fusing the high impedance fault characterization results and the transient disturbance sequence, constructing a dynamic fingerprint response matrix in a multi-device parallel environment, introducing a signal blind source separation mechanism to eliminate the signal coupling interference between parallel devices, and combining a pre-set device characteristic spectrum library to identify the fault attribution branch and generate source positioning path information; S4, constructing a distributed fault data collaborative processing architecture, and performing local inference on the edge side to generate structured fault events and obtain time sequence consistent fault analysis and diagnosis results; S5, based on lightweight communication and intelligent publishing technology, connecting management terminals and transmitting fault analysis and diagnosis results.

[0040] Since the electronic device introduced in the embodiment is the electronic device used in the fault monitoring system for mechanical and electrical installation engineering in the embodiment of the application, based on the fault monitoring system for mechanical and electrical installation engineering in the embodiment of the application, those skilled in the art can understand the specific implementation of the electronic device of the embodiment and its various forms, so the electronic device how to realize the method in the embodiment of the application is not introduced in detail. As long as the electronic device used in the fault monitoring system for mechanical and electrical installation engineering in the embodiment of the application is implemented, it belongs to the scope of protection of the present application.

[0041] The above formulas are all dimensionless numerical calculations, the formulas are obtained by collecting a large amount of data to simulate the recent real situation, and the preset parameters and threshold values in the formulas are set by a person skilled in the art according to the actual situation.

[0042] The above is only the preferred embodiment of the present application, the protection scope of the present application is not limited to the above-mentioned examples, any technical solution belonging to the idea of the present application is within the protection scope of the present application. It should be pointed out that, for ordinary technical operators in the technical field, some improvements and decorations without departing from the principle of the present application should also be considered as the protection scope of the present application.

Claims

1. A fault monitoring system for electromechanical installation engineering, characterized in that: include: The high-impedance fault identification module extracts nonlinear harmonic features from current signals, combines them with the spatiotemporal magnetic distortion during electromechanical installation, constructs a frequency-domain-time-domain joint feature set, identifies the nonlinear characteristic response of high-impedance faults, and outputs high-impedance fault characterization results. The transient fault capture module locates the risk time window based on the high-impedance fault characterization results and uses a dual-threshold mutation mechanism to identify the start and end boundaries of electrical disturbance behavior. It also simultaneously extracts corresponding operating condition parameters, constructs a short-term process behavior model with electrical disturbances as the source, and outputs a transient disturbance sequence. The fault source location module integrates high-impedance fault characterization results with transient disturbance sequences to construct a dynamic fingerprint response matrix for multiple devices in parallel. It introduces a blind source separation mechanism to eliminate signal coupling interference between parallel devices. Combined with a preset device feature library, it identifies the fault-attributing branch and generates source location path information. The multimodal analysis module builds a distributed fault data collaborative processing architecture and performs local reasoning on the edge to generate structured fault events and obtain time-consistent fault analysis and diagnosis results. The coordination control module, based on lightweight communication and intelligent publishing technology, connects to the management terminal and transmits fault analysis and diagnosis results.

2. A fault monitoring system for electromechanical installation engineering according to claim 1, characterized in that: The method for acquiring the nonlinear harmonic characteristics includes: The current signal of the terminal equipment during the electromechanical installation process is sampled in real time to obtain a sampled current signal set; the sampled current signal set is stored in the current data buffer after anti-aliasing filtering, and sliced ​​according to the preset time window; Based on the sampled current signal in each time window, a bispectral map is constructed to extract frequency domain statistical features. By quantifying the complexity of the current signal, the sampled current signal set in each time window is discretized into a one-dimensional time series with a predefined embedding dimension and tolerance threshold. By constructing subsequences of the one-dimensional time series and calculating the maximum distance between all subsequences, the proportion of similar subsequences is counted under the predefined tolerance threshold to calculate the sample entropy value. The sample entropy values ​​in all time windows are collected and integrated to obtain sample entropy data. The frequency domain statistical features are integrated with the sample entropy data to obtain nonlinear harmonic features.

3. A fault monitoring system for electromechanical installation engineering according to claim 2, characterized in that: The method for constructing a frequency domain-time domain joint feature set includes: Magnetic field sensors are placed at key nodes of electromechanical installation terminal equipment to sample the intensity and direction data of the magnetic field in real time during equipment operation. Key nodes of terminal equipment include busbar connection points, switchgear interfaces, circuit breaker interfaces, cable connectors, terminal connection points, and grounding devices. Time domain analysis is performed on the collected intensity data and direction data to extract magnetic field time series data. Using the sampling timestamp as the index, the nonlinear harmonic features and magnetic field time series data are time-step aligned and spliced ​​and combined in the same time window to form a unified frequency domain-time domain joint feature set.

4. A fault monitoring system for electromechanical installation engineering according to claim 3, characterized in that: The method for obtaining the high impedance fault characterization result includes: The high-impedance fault occurrence probability value is calculated and output based on the frequency domain-time domain joint feature set. When the high-impedance fault occurrence probability value exceeds the preset high-impedance fault occurrence probability threshold, it is determined that a high-impedance fault has occurred, and a structured high-impedance fault characterization result is generated. The high-impedance fault characterization result includes the high-impedance fault occurrence probability value and the fault start timestamp.

5. A fault monitoring system for electromechanical installation engineering according to claim 4, characterized in that: The method for locating the risk time window includes: based on the fault start timestamp in the high-impedance fault characterization result, extending the preset time width forward and backward to set the risk time window.

6. A fault monitoring system for electromechanical installation engineering according to claim 5, characterized in that: The method for identifying the start and end boundaries of electrical disturbance behavior includes: Electrical waveform data is acquired through continuous sampling within the risk time window and, after denoising and filtering, is used to identify the electrical disturbance boundary. By jointly considering the waveform instantaneous jump amplitude and the current change rate, the occurrence time and duration of the electrical disturbance behavior are determined, thereby identifying the starting and end points of the electrical disturbance behavior. When the electrical disturbance boundary discrimination value is 1, it indicates that the time point is identified as an electrical disturbance boundary candidate point. All time points that meet the electrical disturbance boundary discrimination value of 1 are used to construct an electrical disturbance boundary candidate set. All identified electrical disturbance boundary candidate points are clustered, and adjacent electrical disturbance boundary candidate point segments are merged to obtain the electrical disturbance boundary candidate point segment interval. For each electrical disturbance boundary candidate point segment interval, the time point when the first electrical disturbance boundary discrimination value of the electrical disturbance boundary candidate point segment is 1 is defined as the starting time point, and the time point when the last electrical disturbance boundary discrimination value of the electrical disturbance boundary candidate point segment is 1 is defined as the ending time point, and the electrical disturbance boundary candidate point segment interval is output as a valid start and end time interval of the electrical disturbance behavior; all valid start and end time intervals of the electrical disturbance behavior are collected to form the complete start and end boundaries of the electrical disturbance behavior.

7. A fault monitoring system for electromechanical installation engineering according to claim 6, characterized in that: The method for obtaining the transient disturbance sequence includes: After identifying the start and end boundaries of the electrical disturbance behavior, a target time window is constructed based on the start and end time points of the electrical disturbance behavior, with the start and end time points as the center and a closed interval of a preset buffer time T extended forward and backward. The operating condition parameters within the corresponding target time window are extracted. The electrical waveform data and operating condition parameters are timestamp-aligned to construct a short-term process behavior model with electrical disturbance as the source, and generate a transient disturbance sequence driven by the electrical disturbance.

8. A fault monitoring system for electromechanical installation engineering according to claim 7, characterized in that: The method for constructing the dynamic fingerprint response matrix includes: Based on the high-impedance fault characterization results and transient disturbance sequences, and on the basis of the device-fault feature mapping relationship, the operating status and corresponding topological position of each terminal device in the current parallel branch are determined through the SCADA platform and the preset device topology database within the target time window, and the local response signal of each terminal device is extracted to form the device disturbance response vector. Based on the device disturbance response vector, a dynamic fingerprint response matrix is ​​constructed for the multi-device parallel operation scenario.

9. A fault monitoring system for electromechanical installation engineering according to claim 8, characterized in that: The method for identifying the fault-attributing branch and generating source location path information includes: A blind source separation mechanism is introduced to eliminate signal coupling interference between different terminal devices in a parallel environment. The blind source separation mechanism involves constructing the local response signal of each terminal device in the dynamic fingerprint response matrix into a mixed signal matrix. An independent component analysis algorithm is used to separate the source signals of the mixed signal matrix to obtain m potential fault source signals with independent statistical characteristics. Based on the similarity between each separated potential fault source signal and the preset device feature spectrum library, the target terminal device to which the potential fault source signal belongs is determined, and the source location path information is generated in combination with the terminal device topology structure provided by the preset device topology database.

10. A fault monitoring system for electromechanical installation engineering according to claim 9, characterized in that: The fault analysis and diagnosis results include the fault type, fault event ID, fault start and end timestamps, fault location, fault duration, structured fault event identifier, and fault handling suggestions.

Images