Intelligent remote management and control platform of unattended substation and real-time monitoring control method

Abstract

The application relates to the technical field of substation monitoring, and discloses an intelligent remote management and control platform of an unattended substation and a real-time monitoring and control method, wherein the method is characterized in that: a vibration footprint ring is established, vibration acceleration signals are collected, vibration sources are positioned and contribution calculation is performed; a vibration propagation attenuation model is constructed and multi-level vibration shadow stripping is performed, a device vibration characteristic vector and a time-varying matrix are generated; a dynamic energy distribution atlas is established to identify weak faults, predict fault trends, construct a device health state prediction matrix and perform an intelligent load adjustment strategy, and a system optimization control instruction sequence is generated; the application solves the feature separation problem under the mutual coupling interference of multiple device vibration signals, especially under the condition of low signal-to-noise ratio, through double decoupling in time domain and space, the weak fault characteristics of the device are accurately identified, potential faults are found in advance, the accuracy and timeliness of preventive maintenance are improved, and a strong guarantee is provided for the safe and stable operation of the unattended substation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of substation monitoring, and more particularly, to an intelligent remote management and control platform for unattended substations and a real-time monitoring and control method. BACKGROUND

[0002] In the technical field of substation monitoring, the stable operation of unattended substations, as key nodes of power systems, highly depends on precise remote monitoring and dynamic control technology. With the expansion of power network scale and the increase of equipment complexity, higher requirements are put forward for real-time perception of equipment status, fault early warning and self-adaptive adjustment in substations, and reliable management and control of equipment throughout the life cycle need to be realized through intelligent control strategies.

[0003] In the prior art, remote control of unattended substations has been explored. Chinese Patent Application No. CN106094916A discloses a humidity remote control system for unattended substations, which realizes remote adjustment of humidity through the cooperation of humidity monitoring devices, remote signal transceiver components and monitoring station equipment, without the need for on-site operation by duty personnel, thereby saving labor costs. Chinese Patent Application No. CN102122840A proposes a remote intelligent adjustment and control system for operating environment of unattended substations, which remotely controls and monitors the state of air conditioners, exhaust fans, lights and the like through integrated control equipment, thereby improving controllability and energy efficiency of environmental management.

[0004] However, the above-mentioned prior art focuses on the regulation of environmental parameters of substations and does not involve precise monitoring and analysis of vibration signals during equipment operation. In actual operation of unattended substations, multiple devices such as transformers, cooling fans and switching devices are often in working state at the same time, and their vibration signals are coupled with each other through media such as substation steel structure frame and foundation to form complex mixed signals. Traditional vibration monitoring methods can only collect overall vibration data and cannot separate the vibration characteristics of a single device under low signal-to-noise ratio conditions. When early-stage loosening, bearing wear and other weak faults occur in key devices such as circuit breaker mechanisms, the characteristic vibration signals will be masked by the normal vibration of surrounding devices. Due to the lack of effective signal decoupling and feature extraction means, the operation and maintenance system cannot identify faults early, and when the fault develops to a significant degree that can be detected, the best maintenance window has often been missed, which not only increases the risk of unplanned power outage, but also may cause a chain reaction due to fault propagation, threatening the overall operation safety of the substation. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the application provides an intelligent remote monitoring and control platform for an unattended substation and a real-time monitoring and control method, precise separation and feature extraction of multi-device coupled vibration signals are realized through construction of a vibration footprint circle, multi-level vibration shadow stripping and the like, and the problem of weak fault feature recognition under low signal-to-noise ratio is solved; the device fault trend can be predicted in advance and dynamic load adjustment is performed, and the control accuracy and operation safety of the unattended substation are significantly improved.

[0006] In order to achieve the above-mentioned purpose, the application provides the following technical scheme:

[0007] The intelligent remote real-time monitoring and control method for the unattended substation comprises the following steps:

[0008] A vibration footprint circle of each device in the substation is established, monitoring points are arranged on the boundary of the vibration footprint circle, and vibration acceleration signals in the starting stage of the device are collected; the vibration source is positioned and the vibration contribution degree of the vibration source is calculated based on the vibration acceleration signals;

[0009] According to the vibration source positioning result and the vibration contribution degree, a vibration propagation attenuation model is established and multi-level vibration shadow stripping is performed, and a device vibration feature vector and a vibration feature time-varying matrix are generated;

[0010] According to the device vibration feature vector and the vibration feature time-varying matrix, a dynamic energy distribution atlas is established; a weak fault is identified and a fault development trend is predicted according to the dynamic energy distribution atlas, and a fault development trend prediction curve is generated;

[0011] Based on the fault development trend prediction curve, a device health state prediction matrix is constructed, and an intelligent load adjustment strategy is executed, and a system optimization control instruction sequence is generated.

[0012] Further, the method for establishing the vibration footprint circle of each device in the substation comprises the following steps:

[0013] The spatial coordinate information, rated power and installation foundation stiffness of each device in the substation are obtained, the geometric center of each device is determined according to the spatial coordinate information of the device, and the maximum vibration influence radius of each device is calculated according to the rated power and installation foundation stiffness of the device; a circular area is formed on the horizontal plane with the geometric center of each device as the center and the maximum vibration influence radius as the radius, and the circular area is the vibration footprint circle of each device.

[0014] Further, the method for positioning the vibration source based on the vibration acceleration signals is that a device vibration identity library is established based on the vibration acceleration signals in the starting stage of the device and the vibration source is positioned through pattern matching.

[0015] Further, the method for establishing the device vibration identity library comprises the following steps:

[0016] performing time domain analysis on the vibration acceleration signals to extract vibration intensity time series of each monitoring point;

[0017] combining the vibration intensity time series of each monitoring point to form a start-up vibration fingerprint matrix of the equipment, performing normalization processing on the start-up vibration fingerprint matrix to obtain a normalized start-up vibration fingerprint matrix;

[0018] obtaining normalized start-up vibration fingerprint matrices under different load conditions to form a multi-condition vibration fingerprint set;

[0019] extracting vibration identity feature vectors of each equipment from the multi-condition vibration fingerprint set, and combining the vibration identity feature vectors of all equipment to establish an equipment vibration identity library.

[0020] Further, the vibration source refers to an equipment in a running state, i.e., a running equipment;

[0021] The method for locating the vibration source through pattern matching comprises:

[0022] real-time acquisition of mixed vibration signals of each monitoring point, construction of a real-time vibration observation matrix, elements of the real-time vibration observation matrix being vibration intensities of each monitoring point; and pattern matching of the real-time vibration observation matrix with the equipment vibration identity library to identify the vibration source.

[0023] Further, the method for identifying the vibration source by pattern matching of the real-time vibration observation matrix with the equipment vibration identity library is: calculating the similarity of the real-time vibration observation matrix with each vibration identity feature vector in the equipment vibration identity library, and determining the vibration source when the similarity exceeds a set similarity threshold.

[0024] Further, the calculation method of the vibration contribution degree of the vibration source is:

[0025] adding up the vibration intensities of all monitoring points of the vibration source i to obtain the total vibration intensity of the vibration source i; dividing the total vibration intensity of the vibration source i by the number of monitoring points of the vibration source i to obtain the average vibration intensity of the vibration source i, wherein i is an index variable of the vibration source;

[0026] calculating the sum of the average vibration intensities of all vibration sources;

[0027] the ratio of the average vibration intensity of the vibration source i to the sum of the average vibration intensities of all vibration sources is the vibration contribution degree of the vibration source i.

[0028] Further, the method for performing multi-level vibration shadow stripping comprises:

[0029] constructing a vibration propagation attenuation model according to the vibration source positioning result and the vibration contribution degree to generate a vibration shadow identification matrix;

[0030] identify the vibration source with the greatest impact on the target device based on the vibration shadow identification matrix, mark it as the maximum impact vibration source, and calculate the vibration shadow of the maximum impact vibration source at the position of the target device;

[0031] subtract the vibration shadow generated by the maximum impact vibration source from the mixed vibration signal of the target device, complete the first level of stripping, and obtain the first level of stripped signal;

[0032] On the basis of the first level of stripping, continue to identify the secondary vibration source and strip the vibration shadow of the secondary vibration source, and obtain the second level of stripped signal; repeat the stripping process until the total intensity of the vibration shadow of the remaining vibration source is lower than the set vibration intensity threshold, and obtain the pure vibration signal.

[0033] Further, the execution of the multi-level vibration shadow stripping further includes:

[0034] According to the pure vibration signal, calculate the stripping quality evaluation index Q, when the stripping quality evaluation index Q is greater than the preset stripping quality threshold q0, it is considered that the stripping is successful; otherwise, re-execute the multi-level vibration shadow stripping operation on the mixed vibration signal.

[0035] Further, the method for generating the device vibration feature vector and the vibration feature time-varying matrix includes:

[0036] Perform frequency domain analysis on the pure vibration signal, calculate the power spectral density, identify the fundamental frequency, the frequency of each harmonic and the power of each harmonic from the power spectral density, and construct the device vibration feature vector; arrange the device vibration feature vectors obtained in time sequence in turn to construct the vibration feature time-varying matrix.

[0037] Further, the method for identifying weak faults according to the dynamic energy distribution map includes:

[0038] Construct a vibration energy gradient vector based on the dynamic energy distribution map, draw an energy evolution trajectory of the target device according to the vibration energy gradient vector, and calculate the trajectory deviation degree of the energy evolution trajectory; when the trajectory deviation degree exceeds the preset deviation threshold, it is considered that a weak fault is identified.

[0039] The intelligent remote management and control platform of the unattended substation is used to realize the intelligent remote real-time monitoring and control method of the unattended substation, and the intelligent remote management and control platform of the unattended substation includes:

[0040] The vibration source positioning module is used to establish the vibration footprint ring of each device in the substation, arrange monitoring points on the boundary of the vibration footprint ring, collect vibration acceleration signals in the starting stage of the device, position the vibration source based on the vibration acceleration signals and calculate the vibration contribution degree of the vibration source.

[0041] Vibration shadow stripping module: for establishing a vibration propagation attenuation model and performing multi-level vibration shadow stripping according to the vibration source positioning result and the vibration contribution degree, generating a device vibration feature vector and a vibration feature time-varying matrix;

[0042] Fault prediction module: for establishing a dynamic energy distribution atlas according to the device vibration feature vector and the vibration feature time-varying matrix; identifying weak faults and predicting fault development trends according to the dynamic energy distribution atlas, and generating a fault development trend prediction curve;

[0043] Intelligent regulation module: based on the fault development trend prediction curve, constructing a device health state prediction matrix, and executing an intelligent load regulation strategy, generating a system optimization control instruction sequence.

[0044] Compared with the prior art, the present application has the following beneficial effects:

[0045] The present application is based on the spatial layout of the vibration footprint ring and the optimization of the monitoring points, physically isolates the device vibration propagation path, combines vibration source positioning and contribution quantification analysis, effectively suppresses the spatial aliasing effect of multi-device vibration signals; through multi-level vibration shadow stripping technology, the time domain signal is decomposed layer by layer, the background noise interference is dynamically removed by using the vibration propagation attenuation model, and the weak fault features are enhanced and displayed in the time-frequency domain; the construction of the dynamic energy distribution atlas realizes the visual mapping of the device vibration state, combined with the fault development trend prediction curve and the health state matrix, a closed-loop control system from feature extraction to fault diagnosis to operation and maintenance decision is formed. Through the synergistic effect of the above technical solutions, the present application significantly improves the feature separation accuracy of the vibration signal under the condition of low signal-to-noise ratio without relying on additional hardware, which helps to solve the technical problem that early weak faults are easily covered by background noise in traditional methods, and provides a precise, efficient and intelligent solution for the state monitoring and risk prevention and control of key devices in unattended substations. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creating any creative labor.

[0047] Figure 1 The method flowchart of the intelligent remote real-time monitoring and control method of the unattended substation in the present application;

[0048] Figure 2 The principle flowchart of positioning the vibration source and calculating the vibration contribution degree of the vibration source provided by the embodiment of the present application;

[0049] Figure 3 A schematic diagram of a monitoring point layout method for large equipment and small equipment in a substation is provided for an embodiment of the present application.

[0050] Figure 4 A principle flowchart of obtaining a pure vibration signal is provided for an embodiment of the present application.

[0051] Figure 5 A functional module diagram of an intelligent remote management and control platform for an unattended substation is provided in the present application.

[0052] The reference signs are as follows: 1 - substation; 2 - geometric center of large equipment; 3 - geometric center of small equipment; 4 - vibration footprint ring of large equipment; 5 - vibration footprint ring of small equipment; 6 - monitoring point. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0054] Embodiment 1

[0055] Please refer to Figure 1 The present embodiment provides an intelligent remote real-time monitoring and control method for an unattended substation, which includes the following steps:

[0056] Step S10, a vibration footprint ring is established for each device in the substation, monitoring points are arranged on the boundary of the vibration footprint ring, and vibration acceleration signals in the starting stage of the device are collected; vibration sources are located and vibration contribution degrees of the vibration sources are calculated based on the vibration acceleration signals.

[0057] Further, step S10 includes the following steps:

[0058] Step S11, a three-dimensional vibration coordinate system is established with the center point of the substation as the origin, and a vibration footprint ring is constructed for each device in the substation in the three-dimensional vibration coordinate system.

[0059] A three-dimensional vibration coordinate system is constructed with the center point of the substation as the origin O, the position and direction of the main transformer are set as the positive direction of the Z axis, and the direction of the control room is set as the positive direction of the X axis. The Y axis direction is determined by the right-hand rule. The establishment of the three-dimensional vibration coordinate system makes the spatial positions of all devices in the substation have a unified reference, avoiding the position confusion caused by the traditional local coordinate system. In the three-dimensional vibration coordinate system, the method for constructing the vibration footprint ring of each device in the substation includes: obtaining the spatial coordinate information, the rated power and the installation foundation stiffness of each device in the substation, determining the geometric center of each device according to the spatial coordinate information of the device, and calculating the maximum vibration influence radius of each device according to the rated power and the installation foundation stiffness of the device; forming a circular ring area on the horizontal plane with the geometric center of each device as the center and the maximum vibration influence radius as the radius, and the circular ring area is the vibration footprint ring of each device. The spatial coordinate information of the device is obtained by a laser range finder. The calculation formula of the maximum vibration influence radius is , wherein, is the maximum vibration influence radius of the device i*, i* is the index of each device in the substation; k1 is the vibration propagation coefficient, which is calibrated by vibration propagation experiments on the same type of device in the same type of substation, and the value range is 0.02 to 0.05. Specifically, at least 10 groups of vibration propagation experiments are conducted on the same type of device in the same type of substation by controlling the environmental temperature and the device load, and the linear relationship between the maximum vibration influence radius and is obtained by least squares fitting, and the goodness of fit is . is the rated power of the device i*; is the installation foundation stiffness of the device i*, which is a quantitative index of the deformation resistance of the foundation and is determined by the synergistic effect of concrete and steel: the concrete part, the elastic modulus is determined according to its strength grade, and the inertia moment of the concrete part is calculated combined with the foundation section shape, and the product of the two is the contribution of the concrete stiffness; the steel part, the elastic modulus is determined according to the type, and the inertia moment of the steel part is calculated combined with the total cross-sectional area and the distance to the neutral axis, and the product of the two is the contribution of the steel stiffness; the installation foundation stiffness is the sum of the contributions of the above two parts divided by the effective stress height of the foundation, such as thickness, which comprehensively reflects the constraint characteristics of the foundation to the vibration propagation of the device, and provides parameters for calculating the maximum vibration influence radius .

[0060] For example, the inertia moment of the concrete part is calculated combined with the foundation section shape: if the foundation of a device is a rectangular section with a length of 4m and a width of 2m, the calculation logic of the inertia moment is based on the geometric characteristics of the rectangular section, and the formula "inertia moment = length x width 3 / 12" is used. The specific calculation is: 4m x (2m) 3÷12; the moment of inertia of the concrete part in the formula is directly determined by the length and width of the rectangular cross section, which reflects the influence of the cross section shape on the moment of inertia. Under the same area, the moment of inertia will change accordingly if the cross section size distribution is different (for example, the width is larger).

[0061] For example, the moment of inertia of the steel bar part is calculated in combination with the total cross-sectional area and the distance to the neutral axis: if the total cross-sectional area of the steel bar of a certain foundation configuration is 0.04 m2, and all the steel bars are symmetrically distributed on both sides of the neutral axis of the cross section, and the distance from the neutral axis of a single steel bar is 1 m, the calculation logic of the moment of inertia is based on the distribution characteristics of the steel bar, and the formula "moment of inertia = total cross-sectional area x distance to neutral axis squared" is used. The specific calculation is: 0.04 m 2 ×(1m) 2 =0.04 m 4 The result is determined by the total cross-sectional area of the steel bar and the distance to the neutral axis, which reflects the influence of the steel bar arrangement on the moment of inertia. Under the same cross-sectional area, the farther away from the neutral axis, the greater the moment of inertia.

[0062] Step S12, arranging monitoring points on the boundary of the vibration footprint circle, collecting vibration acceleration signals in the starting stage of the equipment;

[0063] Step S13, establishing an equipment vibration identity library based on the vibration acceleration signals in the starting stage of the equipment and positioning the vibration source through pattern matching and calculating the vibration contribution degree of the vibration source, wherein the vibration source refers to the equipment in the running state.

[0064] Please refer to Figure 2 Further, step S13 includes:

[0065] Step S131, performing time domain analysis on the vibration acceleration signals to extract the vibration intensity time series of each monitoring point;

[0066] Step S132, combining the vibration intensity time series of each monitoring point to form a starting vibration fingerprint matrix of the equipment, and performing normalization processing on the starting vibration fingerprint matrix to obtain a normalized starting vibration fingerprint matrix;

[0067] Step S133, obtaining normalized starting vibration fingerprint matrices under different load conditions to form a multi-condition vibration fingerprint set;

[0068] Step S134, extracting vibration identity feature vectors of each equipment from the multi-condition vibration fingerprint set, and combining the vibration identity feature vectors of all equipment to establish an equipment vibration identity library;

[0069] Step S135, collecting mixed vibration signals of each monitoring point in real time, constructing a real-time vibration observation matrix, and performing pattern matching between the real-time vibration observation matrix and the equipment vibration identity library to identify the vibration source;

[0070] Step S136, according to the vibration intensity distribution of the identified vibration source within the vibration footprint circle, the vibration contribution degree of each vibration source is calculated.

[0071] N monitoring points are arranged at equal angle intervals θ on the boundary of the vibration footprint circle, the equal angle interval θ is 360 degrees divided by N, N is the number of monitoring points, N is determined according to the size of the equipment, the number of monitoring points of large equipment is more than that of small equipment, for example, N=12 for large equipment and N=8 for small equipment. Figure 3 A large equipment and a small equipment running in a certain substation 1 are shown, and a vibration footprint circle 4 of the large equipment with the geometric center 2 of the large equipment as the center and a vibration footprint circle 5 of the small equipment with the geometric center 3 of the small equipment as the center, 12 monitoring points 6 are arranged at equal angle intervals of 30° on the boundary of the vibration footprint circle of the large equipment, and 8 monitoring points are arranged at equal angle intervals of 45° on the boundary of the vibration footprint circle of the small equipment. This arrangement makes the monitoring points evenly cover the key path of vibration propagation, ensuring the capture of the spatial distribution characteristics of vibration. According to the center point coordinates and the maximum vibration influence radius of the vibration footprint circle, the position coordinates of each monitoring point can be accurately obtained. In the traditional technology, the monitoring points are arranged randomly, which cannot reflect the spatial attenuation law of vibration, while in the present application, the vibration propagation can be quantitatively analyzed through the standardized coordinate system and the circle structure, providing accurate spatial coordinate reference for subsequent positioning; if this step is missing, the vibration signal will be unable to distinguish the source due to the lack of spatial reference, resulting in positioning failure.

[0072] When the device receives the start instruction, the N monitoring points on the boundary of the vibration footprint circle synchronously record the vibration acceleration signals, and the collection time needs to be greater than the longest time required for the device to run from static to stable. The root mean square value of each vibration acceleration signal is calculated in the preset time window (such as 0.1s) to obtain the vibration intensity time sequence, and the vibration acceleration signals of the N monitoring points obtain N vibration intensity time sequences. The N vibration intensity time sequences are combined into a start vibration fingerprint matrix, with rows corresponding to monitoring points and columns corresponding to time, and the matrix element value is the vibration intensity value. The start vibration fingerprint matrix is normalized: each element is divided by the average vibration intensity of the device in stable operation to obtain a normalized start vibration fingerprint matrix, so as to eliminate the influence of absolute intensity difference on the fingerprint; the average vibration intensity of the device in stable operation is the average vibration intensity in 30 minutes of stable operation of the device. The normalized start vibration fingerprint matrix under different load conditions is obtained to form a multi-condition vibration fingerprint set; different load conditions include no load, light load, full load, etc. The vibration acceleration signal of the device during the start-up process is determined by the mechanical structure and has uniqueness, such as the significant difference between the start-up characteristics of the transformer core vibration and the winding vibration. Multi-condition acquisition can cover different operating states of the device and avoid recognition errors caused by incomplete fingerprints in a single condition. The traditional method only collects steady-state vibration and cannot capture the unique characteristics in the start-up stage, while the fingerprint constructed by the dynamic process acquisition in this step has higher distinguishability; if steps S12 and S13 are missing, the vibration signals of different devices will be similar in characteristics and cannot be distinguished, and the subsequent positioning will lose the basis.

[0073] The vibration identity feature vector is extracted from the multi-condition vibration fingerprint set, including the start time constant, the vibration rising slope, the steady-state vibration mode, and the working condition sensitivity. The start time constant is the time for the vibration intensity to rise to a preset proportion of the steady-state value from the initial value, wherein the preset proportion can be determined according to the device type or experimental data, for example, it can be selected in the range of 50% to 70%, the initial value refers to the vibration intensity reference value before the device starts, i.e. the device is in a static state, reflecting the environmental background vibration or the basic vibration when the device is not running; the steady-state value refers to the constant vibration intensity value after the device completes the start-up and enters the stable operation state. The vibration rising slope is the average slope of the rising section; the steady-state vibration mode is the normalized distribution of the steady-state intensity of each monitoring point, reflecting the spatial attenuation characteristics of the vibration; the working condition sensitivity is the average value of the cosine similarity of the normalized start vibration fingerprint matrix under different working conditions, and the smaller the value, the greater the influence of the working condition on the vibration. The vibration identity feature vectors of all devices are combined into a device vibration identity library.

[0074] In real-time operation, the mixed vibration signals of each monitoring point are collected to construct a real-time vibration observation matrix, the elements of the real-time vibration observation matrix are the vibration intensities of each monitoring point, the similarity between the real-time vibration observation matrix and each vibration identity characteristic vector in the device vibration identity library is calculated, the cosine similarity calculation is adopted, the value range is 0 to 1, when the similarity exceeds the set similarity threshold, it is determined that it is a vibration source, the vibration source refers to the device in the running state; the similarity threshold is determined by the similarity distribution of the normal vibration source in the historical data, for example, the 95% quantile is taken. Let i be the index variable of the vibration source, the vibration contribution degree C i The calculation of (t) is as follows: first, add the vibration intensities of all monitoring points of the vibration source i to obtain the total vibration intensity of the vibration source i; then, divide the total vibration intensity of the vibration source i by the number of monitoring points of the vibration source i, for example, 12 for large devices and 8 for small devices, to obtain the average vibration intensity of the vibration source i. At the same time, the sum of the average vibration intensities of all vibration sources is calculated: the average vibration intensity of each vibration source is calculated according to the above method, and then the average vibration intensities are added to obtain the sum of the average vibration intensities of all vibration sources. The ratio of the average vibration intensity of the vibration source i to the sum of the average vibration intensities of all vibration sources is the vibration contribution degree C i (t) of the vibration source i. i is the index variable of the vibration source in the transformer substation. The vibration contribution degree C i (t) reflects the proportion of the influence of the vibration source on the overall vibration field. The traditional method cannot identify the specific vibration source when multiple devices are running at the same time, but the present application realizes accurate identification through pattern matching, and the vibration contribution quantifies the influence weight of each device; if step S13 is missing, the subsequent signal separation will be unable to target interference due to the inability to determine the vibration source.

[0075] Step S10 solves the problem that the signal source cannot be accurately positioned due to the spatial position aliasing of multiple vibration sources by combining spatial decoupling and dynamic fingerprinting, and realizes accurate positioning of multiple vibration sources. The cooperation of the three-dimensional vibration coordinate system and the vibration footprint ring makes the vibration propagation have a quantifiable spatial reference, solving the problem of missing spatial reference in traditional positioning; the multi-working-condition vibration fingerprint contains unique characteristics of the device under different loads, compared with the single-working-condition fingerprint, the identification accuracy of the vibration source under complex operating conditions is improved, for example, the starting vibration slope of the transformer is significantly different between light load and full load, the multi-working-condition fingerprint can capture this difference and avoid misjudgment; the working condition sensitivity in the vibration identity characteristic vector can distinguish the sensitivity of the vibration source to the load, providing a basis for subsequent load adjustment, which is not involved in the traditional method; the symmetrical arrangement of the monitoring points makes the spatial distribution characteristics of the vibration more obvious, improving the accuracy of the vibration contribution degree calculation. Step S10 provides clear vibration source information for the signal separation of step S20, if S10 is missing, the subsequent signal separation will be blind due to unknown source, cannot realize effective decoupling, and the fault detection capability of the overall scheme will be greatly reduced.

[0076] Step S20, according to the vibration source positioning result and the vibration contribution degree, a vibration propagation attenuation model is established and multi-level vibration shadow stripping is performed to generate a device vibration feature vector and a vibration feature time-varying matrix;

[0077] Step S20 is based on the time-domain feature dynamic stripping method of the vibration shadow. By constructing a vibration propagation attenuation model, performing multi-level stripping operation and generating a device vibration feature vector, the problem of single device feature extraction caused by time-domain overlap of vibration signals in the same frequency band is solved, and pure vibration features are provided for subsequent weak fault signal recognition.

[0078] Further, step S20 includes:

[0079] Step S21, according to the vibration source positioning result and the vibration contribution degree, a vibration propagation attenuation model is constructed to generate a vibration shadow identification matrix;

[0080] The vibration shadow refers to the secondary vibration signal formed by the vibration of a vibration source propagating through the steel structure frame of the substation to another vibration source position, the intensity of which attenuates with the propagation distance and has a time delay. The core parameters of the vibration propagation attenuation model include the attenuation coefficient a and the propagation velocity v: the attenuation coefficient a is obtained by experiment calibration. The specific method is to select typical steel structure components in the substation, such as beams and columns, and arrange vibration sensors at different distances to collect the attenuation data of vibration signals with known intensity in the propagation process. The least square method is used to fit the "vibration intensity-distance" curve to obtain the value of a. The propagation velocity v is calculated according to the elastic modulus E and the density p of the steel material, and the formula is v equal to the square root of the quotient of E divided by p, where E is obtained from the steel material quality parameter table, and p is the density of the steel material. This formula conforms to the classical calculation method of elastic wave propagation velocity in solid, and its physical meaning is that the wave velocity is proportional to the square root of the elastic modulus of the material and inversely proportional to the square root of the density. For any two vibration sources i and j, where j is the index variable of the vibration source, i≠j, the vibration shadow intensity of the vibration generated by the vibration source i propagating to the vibration source j is The calculation formula is wherein, is the original vibration intensity of the vibration source i, is the spatial distance between the two vibration sources i and j, e represents the natural constant, which is calculated by the three-dimensional vibration coordinate system in step S11; the time delay is equal to divided by , is the time delay required for the vibration to propagate from the vibration source i to the vibration source j. The vibration shadow identification matrix S is a K×K order matrix, K is the total number of vibration sources, and the element S ijThe intensity attenuation ratio and time delay of the vibration shadow generated by vibration source i at vibration source j. The intensity attenuation ratio of the vibration shadow generated by vibration source i at vibration source j refers to the ratio of the intensity of the vibration shadow generated by vibration source i at vibration source j to the original vibration intensity of vibration source i, that is , for quantifying the degree of intensity attenuation of vibration in the propagation process. The conventional technology does not consider the propagation characteristics of vibration in solid medium, and all vibration signals are regarded as the vibration of the vibration source itself, while this step provides a theoretical basis for distinguishing the vibration of the vibration source itself and the interference vibration by quantifying the attenuation and delay characteristics of the vibration shadow; if this step is missing, the subsequent stripping operation will be blind due to the inability to determine the characteristic parameters of the interference source, and precise separation cannot be achieved.

[0081] Step S22, performing a multi-level vibration shadow stripping operation on the mixed vibration signal based on the vibration shadow identification matrix to obtain the pure vibration signal of each vibration source;

[0082] Please refer to Figure 4 , further, step S22 includes:

[0083] Step S221, identifying the vibration source with the greatest impact on the target device based on the vibration shadow identification matrix, and marking it as the maximum impact vibration source, and calculating the vibration shadow generated by the maximum impact vibration source at the target device position;

[0084] Step S222, subtracting the vibration shadow generated by the maximum impact vibration source from the mixed vibration signal of the vibration source to complete the first level stripping, and obtaining the first level stripped signal;

[0085] Step S223, continuing to identify and strip the vibration shadow of the secondary vibration source based on the first level stripping to obtain the second level stripped signal; repeat the stripping process until the total intensity of the vibration shadow of the remaining vibration source is lower than the set vibration intensity threshold, and obtain the pure vibration signal;

[0086] Step S224, calculating the stripping quality evaluation index Q according to the pure vibration signal, and considering that the stripping is successful when the stripping quality evaluation index Q is greater than the preset stripping quality threshold q0; otherwise, re-performing the multi-level vibration shadow stripping operation on the mixed vibration signal.

[0087] The method for performing a multi-level vibration shadow stripping operation on the mixed vibration signal based on the vibration shadow identification matrix includes: setting vibration source g as the target device, and the mixed vibration signal V g,mixed (t) of the target device at time t is the self-vibration signal V g,self(t) and other vibration sources. The first level of stripping first identifies the vibration source that has the greatest impact on the target device: traverse the vibration shadow identification matrix to identify the elements related to the target device, select the vibration source with the largest intensity attenuation ratio as the maximum impact vibration source; let the maximum impact vibration source be vibration source q, calculate the vibration shadow formed by the vibration generated by the maximum impact vibration source and propagated to the target device position according to the vibration shadow identification matrix , and then subtract g,mixed from V , to obtain the signal after the first level of stripping , where is the vibration shadow intensity of the maximum impact vibration source at the target device, is the original vibration intensity of the maximum impact vibration source, is the time delay required for the vibration to propagate from the maximum impact vibration source to the target device position; is the vibration signal of the maximum impact vibration source generated at

[0088] The second level of stripping is repeated for The above process is repeated to identify the secondary vibration source. The method for identifying the secondary vibration source is: traverse the vibration shadow identification matrix to identify the elements related to the target device, exclude the maximum impact vibration source that has been stripped, and select the vibration source with the largest intensity attenuation ratio from the remaining elements as the secondary vibration source. Calculate the vibration shadow formed by the vibration generated by the secondary vibration source and propagated to the target device and strip it until the total intensity of the vibration shadow of the remaining vibration sources is lower than the set vibration intensity threshold. The vibration intensity threshold is determined based on a large number of experiments, which can ensure the stripping effect and avoid excessive stripping that leads to loss of useful signals, for example, it can be set to 5% of the target device's own vibration intensity. The total intensity of the remaining vibration shadow refers to the sum of the intensities of the vibration shadows that have not been stripped and are propagated to the target device position by other vibration sources after multiple vibration shadow stripping operations. If the total intensity of the vibration shadow of the remaining vibration sources is greater than or equal to the set vibration intensity threshold, the next level of stripping is continued. The signal obtained after multiple stripping processes is taken as the pure vibration signal.

[0089] The calculation method of the stripping quality evaluation index Q is the autocorrelation coefficient of the pure vibration signal obtained after stripping and the mixed vibration signal V g,mixed ​The ratio of the absolute values of the cross-correlation coefficients of (t) is used to measure the internal consistency and periodicity of the pure vibration signal itself. If the pure vibration signal retains the inherent mode of the target device's own vibration, such as stable fundamental frequency vibration, the autocorrelation coefficient will show a high value, reflecting that the pure vibration signal has not lost its core characteristics due to the stripping operation. The commonly used calculation method of autocorrelation coefficient is sliding window method, that is, by setting a fixed length time window, such as 0.5 seconds to 2 seconds, according to the characteristic period of the target device vibration, the similarity of the pure vibration signal and its different lag time versions within the time window is calculated, so as to quantify the self-similarity of the pure vibration signal in time. The cross-correlation coefficient refers to the correlation coefficient between the pure vibration signal obtained after stripping and the mixed vibration signal V g,mixed The correlation coefficient between (t) is used to measure the correlation degree of the pure vibration signal after stripping and the target device related vibration component in the mixed vibration signal, so as to avoid losing the effective vibration information of the target device due to excessive stripping. The calculation method of cross-correlation coefficient is also sliding window method, by calculating the similarity of the pure vibration signal and the mixed vibration signal of the target device within the same time window, the retention degree of the target component in the mixed vibration signal by the pure vibration signal after stripping is reflected. When Q is greater than the stripping quality threshold q0, it is determined that the stripping is successful, which means that the pure vibration signal after stripping not only maintains the inherent characteristics of the target device's own vibration, but also is closely related to the target component in the mixed vibration signal of the target device, that is, the interference is effectively removed and the useful information is retained, and the stripping quality meets the standard. Otherwise, adjust the vibration intensity threshold, and perform the multi-level vibration shadow stripping operation on the mixed vibration signal of the target device again. q0 is determined by the Q value distribution of the normal stripping success cases, usually taking 0.8. The traditional method uses a fixed threshold to filter the interference at one time, which cannot handle complex multi-source interference, while this step can gradually eliminate vibration shadows of different intensities by multi-level progressive stripping and quality evaluation, which is suitable for the scene of dense arrangement of multiple devices in the substation. If this step is missing, the interference components in the mixed vibration signal will mask the abnormal vibration characteristics of the target device itself, resulting in failure of subsequent fault detection.

[0090] In step S23, the device vibration feature vector and the vibration feature time-varying matrix are generated according to the pure vibration signal after stripping.

[0091] The method for generating the equipment vibration feature vector and the vibration feature time-varying matrix according to the stripped pure vibration signal comprises: performing frequency domain analysis on the stripped pure vibration signal to calculate the power spectral density; identifying the fundamental frequency, the frequency of each harmonic and the power of each harmonic from the power spectral density to construct the equipment vibration feature vector; arranging the continuously acquired equipment vibration feature vectors in sequence according to the time sequence to construct the vibration feature time-varying matrix. The row index of the vibration feature time-varying matrix corresponds to the time node, and the column index corresponds to each element of the equipment vibration feature vector. Step S23 converts the time domain vibration signal into a frequency domain feature, which facilitates capturing the frequency shift or harmonic power change caused by the target equipment fault, for example, bearing wear will cause the fundamental frequency power to drop and the high-frequency harmonic power to rise; if this step is missing, the vibration feature of the target equipment will remain at the original signal level, and it is impossible to extract quantifiable indicators for fault diagnosis, and the energy pool analysis of step S30 will also lose the data basis.

[0092] Step S20 realizes effective separation of mixed vibration signals in the same frequency band through quantitative modeling and dynamic stripping of the vibration shadow. The vibration propagation attenuation model cooperates with the three-dimensional vibration coordinate system of step S11, so that the spatial distribution of the vibration shadow can be accurately calculated by distance. Compared with the traditional empirical interference elimination method, the separation accuracy is improved; the multi-stage stripping strategy combines with the vibration contribution degree sorting to preferentially process strong interference sources, so that the stripping efficiency is improved, and the separation time-consuming in the dense equipment area is reduced compared with the traditional method; the harmonic change law recorded by the vibration feature time-varying matrix can not only reflect the fault of the target equipment itself, but also discover the hidden correlation fault between devices through the harmonic phase difference, such as the resonance fault caused by two fans sharing a base, which cannot be realized by single device monitoring. If step S20 is missing, step S30 will not be able to identify weak fault features due to the input signal containing a large amount of interference components, and the fault detection capability of the whole scheme will return to the traditional level, which cannot meet the demand of early fault warning of unattended substations.

[0093] Step S30, according to the equipment vibration feature vector and the vibration feature time-varying matrix, establishes a dynamic energy distribution atlas; identifies weak faults and predicts fault development trend according to the dynamic energy distribution atlas, and generates a fault development trend prediction curve;

[0094] Step S30 is a weak fault signal amplification identification method based on vibration energy pool, which solves the problem that weak vibration signals generated by early faults are covered by normal operation vibrations by constructing a dynamic energy distribution atlas, analyzing energy evolution trajectory, implementing resonance amplification and predicting fault trend, and provides accurate fault information and maintenance basis for subsequent remote intelligent control.

[0095] Further, step S30 comprises:

[0096] Step S31, constructing an energy pool three-dimensional space according to the equipment vibration feature vector and the vibration feature time-varying matrix, and forming a dynamic energy distribution atlas;

[0097] The three dimensions of the energy pool three-dimensional space are a frequency axis, an energy axis and a time axis. The frequency axis is divided based on the fundamental frequency and the harmonic frequency range extracted in the equipment vibration feature vector, covers the frequency interval of the possible fault features of the target equipment, and ensures that the relevant fundamental frequency and each harmonic of the target equipment can be captured. The energy axis is obtained by integrating the power spectral density of each frequency component in the equipment vibration feature vector. The energy value of a certain frequency point is the product of the power spectral density of the frequency point in the corresponding time window and the frequency interval. The time axis is based on the time nodes of the vibration feature time-varying matrix and is divided by using the sliding window technology. The window length is matched with the response speed of the equipment, and a certain overlap degree is set between the windows to avoid information loss. The "frequency-energy-time" parameters corresponding to each element in the vibration feature time-varying matrix are mapped to the coordinate points in the three-dimensional space, and a dynamic energy distribution atlas changing over time is formed by continuous sampling. Under normal operating conditions, the energy of the equipment presents a stable aggregation mode in the three-dimensional space, while early faults will cause a persistent small deviation of the energy of a specific frequency point. The existing two-dimensional spectrum analysis can only reflect the frequency-energy relationship at a certain moment, and cannot capture the evolution law of the energy over time, resulting in the slow change characteristics of early faults being hidden. This step realizes the continuity tracking of the vibration energy in the time dimension by constructing a three-dimensional space, and provides a multi-dimensional data basis for subsequent weak feature extraction. If this step is missing, the subsequent analysis will lack a reference benchmark in the time dimension, and cannot distinguish between the instantaneous fluctuations of normal vibration and the trend changes caused by faults, resulting in the early fault characteristics being submerged.

[0098] Step S32, constructing a vibration energy gradient vector based on the dynamic energy distribution atlas, drawing an energy evolution trajectory of the target equipment according to the vibration energy gradient vector, and calculating a trajectory deviation degree of the energy evolution trajectory; when the trajectory deviation degree exceeds a preset deviation threshold, it is considered that a weak fault is identified;

[0099] The construction process of the vibration energy gradient vector is as follows: in the dynamic energy distribution map, the energy difference of adjacent time windows at the same frequency point is calculated to obtain the energy change rate in the time dimension; at the same time, the energy difference of adjacent frequency points in the same time window is calculated to obtain the energy change rate in the frequency dimension. The change rates in the above two dimensions are combined to form the vibration energy gradient vector, the modulus of the vibration energy gradient vector reflects the intensity of the energy change, and the direction reflects the dominant dimension of the energy change, which is the time dimension or the frequency dimension. The energy evolution trajectory is drawn in sequence with time, and the direction of the vibration energy gradient vector is used to assist in determining the trajectory node: when the direction of the vibration energy gradient vector is biased to the frequency dimension, the frequency point with the highest energy in the time window (dominant frequency) is selected as the node frequency coordinate according to the frequency change; when the direction of the vibration energy gradient vector is biased to the time dimension, the time coordinate and the corresponding energy value of the node are determined according to the time change; these nodes are connected in time sequence to form a continuous path in the three-dimensional space, that is, the energy evolution trajectory. The calculation of the trajectory deviation degree adopts the dynamic reference method: 100 groups of energy evolution trajectories of the target device in normal operation are selected as the reference evolution trajectory set, the reference feature vector of the reference evolution trajectory is extracted through principal component analysis, and the cosine distance between the energy evolution trajectory obtained in real time and the reference feature vector is calculated. The distance is the trajectory deviation degree. The deviation threshold is determined by the deviation degree distribution of the normal operation data, and the upper limit value of the 95% confidence interval is taken. When the target device has an early fault, such as contactor contact wear, the energy of a specific harmonic frequency will continue to rise over time, at this time, the time dimension change rate of the vibration energy gradient vector increases, and the trajectory deviation degree gradually exceeds the deviation threshold, realizing early identification of the fault. The existing abnormality detection method based on a fixed threshold value is difficult to adapt to the dynamic characteristics of the device vibration, and false positives or false negatives often occur. This step captures the energy change trend through the vibration energy gradient vector, and combines the deviation degree calculation of the dynamic reference, so that the detection sensitivity of the weak fault is improved. If this step is missing, only the absolute value of the energy is used for judgment, which will cause the fault signal to be overwhelmed by the background noise in a low signal-to-noise ratio environment, and early warning cannot be realized.

[0100] Step S33, after identifying the weak fault, an energy resonance amplification strategy is implemented, the amplification multiple is recorded, and the fault type and severity are determined;

[0101] The core of the resonant amplification strategy is to adjust the operating parameters of the target equipment to bring the fault characteristic frequency close to the natural frequency of the target equipment, thereby achieving natural amplification of weak signals. The identification process for the fault characteristic frequency ff is as follows: The fundamental frequency and harmonic distribution in the vibration characteristic vector of the target equipment (where the trajectory deviation exceeds the allowable threshold) are compared with the fundamental frequency and harmonic distribution in the vibration characteristic vector of the target equipment under normal operating conditions. Newly added or abnormally increasing frequency components are identified as the fault characteristic frequency. The adjustment of the equipment's natural frequency fn is achieved by changing operating parameters. For example, adjusting the speed of a fan changes the natural frequency of the blade vibration, and adjusting the load rate of a transformer changes the natural frequency of the magnetostriction of the iron core. During the adjustment process, the change in the equipment's natural frequency is monitored in real time until it approaches the fault characteristic frequency.

[0102] Frequency proximity The method used to quantify the proximity of the fault characteristic frequency ff to the equipment's natural frequency fn is as follows: when the fault characteristic frequency ff is within the range of 50% to 200% of the equipment's natural frequency fn (i.e., 0.5ff ≤ ff ≤ 2fn), ;when When it exceeds the above range, A value of 0 indicates that the difference between the two frequencies is too large, and no resonance effect occurs. When this is the case, it means that the difference between the two frequencies is less than 10% of the natural frequency fn (i.e., At this point, the resonance effect becomes significant, and the amplitude of the fault vibration signal gradually increases with the resonance phenomenon. The fault vibration signal refers to the portion of the pure vibration signal obtained after multi-level vibration shadow stripping of the target equipment after a minor fault occurs, which contains the fault characteristic frequency ff. If the two frequencies are not within the effective range for resonance, then resonance amplification is unnecessary, avoiding ineffective adjustment. During resonance amplification, the change in vibration amplitude A(t) is monitored in real time. When A(t) exceeds the safety threshold A... safe If this occurs, immediately stop adjusting the parameters to avoid damaging the equipment. Safety threshold A safe The determination is based on 80% of the rated maximum vibration amplitude provided by the equipment manufacturer to avoid excessive resonance that could damage the equipment.

[0103] The vibration signals before and after resonance amplification are recorded, the amplification multiple M' is calculated, M' is the ratio of the vibration energy of the vibration signal at resonance stability to the vibration energy of the vibration signal before amplification, and reflects the degree of amplification of the vibration signal. According to the amplification multiple M' and the fault characteristic frequency ff, the abnormal mode library is inquired to determine the fault type and severity. This step realizes signal amplification through the physical resonance principle, compared with traditional filtering or gain amplification, it can retain the fault characteristics while avoiding noise amplification, which is its unique advantage; if this step is missing, early fault signals with amplitude lower than noise level will not be effectively identified.

[0104] In step S34, a fault development rate model is established based on the fault type, severity and amplification multiple; according to the fault development rate model, a fault development trend prediction curve and a remaining safe operation time are generated.

[0105] The fault development rate model is: fault development rate v'= basic rate v0 x energy influence coefficient x temperature influence coefficient x time influence coefficient. The basic rate v0 is the average development rate of the type of fault, which is obtained by statistical analysis of historical fault data; the energy influence coefficient is the ratio of the current vibration energy to the initial fault energy, and the higher the energy, the faster the fault development; the initial fault energy refers to the vibration energy when the target device is first identified as having a weak fault, that is, the vibration energy corresponding to the moment when the trajectory deviation degree of the energy evolution trajectory in step S32 first exceeds the preset deviation threshold, indicating that a weak fault has been identified. The temperature influence coefficient is determined according to the deviation of the environmental temperature from the reference temperature (25℃), and the coefficient increases by 0.1 for every 10℃ increase in temperature, reflecting the accelerating effect of high temperature on mechanical wear; the time influence coefficient is the ratio of the running time to the design life of the target device, and the larger the ratio, the larger the coefficient, reflecting the promoting effect of aging on faults. The drawing process of the fault development trend curve is: taking the current time as the starting point, predicting the future vibration energy trajectory with time based on the fault development rate v', when the predicted energy reaches the device failure threshold E critical , the corresponding time is the remaining safe operation time T safe , E critical is determined according to the device damage threshold, for example, when the vibration energy of the transformer core exceeds 10J, it is determined as failure.

[0106] It is worth noting that steps S10-S30 focus on a "single target device" as the analysis object, which is the technical basis for realizing step S40 "control at each device system level": because the vibration signals of multiple devices in the substation are mutually coupled through media such as steel structures and foundations, if all devices are directly analyzed synchronously, the fault source cannot be accurately locked due to signal aliasing. Therefore, S10-S30 uses the technical path of "vibration footprint ring positioning → multi-level vibration shadow stripping → dynamic energy distribution map analysis" to first focus on a single target device, strip the vibration shadow interference of other devices, extract the pure vibration signal of the target device, and then generate a fault development trend prediction curve; this process is not only for a fixed device, but also for each device in the substation as a "target device" to repeat S10-S30, and finally form a set of fault development trend prediction curves covering all devices.

[0107] Step S40, based on the fault development trend prediction curve, constructs a device health state prediction matrix, and executes an intelligent load adjustment strategy to generate a system optimization control instruction sequence.

[0108] Step S40, based on the fault development trend prediction curve, constructs a device health state prediction matrix, and executes an intelligent load adjustment strategy to generate a system optimization control instruction sequence. This step solves the problem of missing the best maintenance opportunity caused by the response lag of the remote monitoring system by converting the fault warning information into an executable system control strategy, and realizes closed-loop management from fault identification to active maintenance.

[0109] Step S41, based on the fault development trend prediction curve, constructs a device health state prediction matrix;

[0110] The predicted vibration energy values at each time node in the future are extracted from the fault development trend prediction curves of each target device; the health index of each target device at each time node is calculated according to the predicted vibration energy value; the health indices of all target devices are arranged according to the device and time node to construct a device health state prediction matrix. The health index of each target device at each time node is calculated using a normalization method, and the formula is health index , wherein the reference energy E0 is the average vibration energy of the target device during normal operation, which is taken from the energy pool data in the normal state in step S31; is the critical energy of the predicted vibration energy value of the target device at a future time node t'; E critical is the device failure threshold, i.e. the vibration energy when the target device fails, and the health index has a value range of 0 to 1, and the closer the value is to 1, the better the health state of the target device, and the closer the value is to 0, the closer to failure. The structure of the device health state prediction matrix is an MxN1 order matrix, M is the total number of target devices to be included in health control, N1 is the total number of future time nodes, such as one node per hour in the future 72 hours, then N1=72, and the matrix element Hi''j' represents the health index of the ith target device at the jth time node, i is the index of the target device to be included in health management, and j is the index of the future time node.

[0111] The device health state prediction matrix converts the scattered fault trend data into a structured global health view. In the prior art, single-device independent evaluation is often used, and the mutual influence between devices is ignored, resulting in one-sided maintenance decisions. This step integrates the health states of multiple devices and multiple time points in the form of a matrix, providing a data basis for subsequent system-level optimization. If this step is missing, subsequent load adjustment will fall into the local optimization error due to the lack of global health reference, and the maintenance needs of multiple devices cannot be balanced. The calculation of the health index combines the difference between the critical energy and the baseline energy as the normalized baseline, which not only preserves the absolute characteristics of energy changes, but also eliminates the influence of different energy magnitudes of different devices through relative values, enabling horizontal comparison of health states of different types of devices. This quantification method, in cooperation with the fault development trend prediction curve, enables the time evolution of the health state to be directly traced back to energy changes, and the precision is significantly improved compared to traditional qualitative descriptions such as "good" and "abnormal".

[0112] In step S42, a device correlation influence matrix is constructed, and a three-level early warning response mechanism is designed based on the device health state prediction matrix, the device correlation influence matrix, and the remaining safe operation time. The three-level early warning response mechanism includes a first-level early warning, a second-level early warning, and a third-level early warning.

[0113] The element r i''j'' in the device correlation influence matrix represents the correlation degree between the target device i and the target device j, j is the index of the target device to be included in health management, and i ≠ j. The construction of the device correlation influence matrix is based on the physical connection relationship, functional dependency relationship, and vibration propagation characteristics of the target device. The specific method is as follows: the device spatial position relationship of the target device is obtained through a three-dimensional vibration coordinate system and quantified as a physical correlation coefficient. The physical correlation coefficient is determined according to the distance between the target devices, for example, 0.8 when the distance is less than 5 meters, 0.4 when the distance is between 5 and 10 meters, and 0.1 when the distance is greater than 10 meters. The functional dependency relationship is determined by combining the electrical wiring diagram of the substation, such as the cooling fan providing heat dissipation support for the transformer, and quantified as a functional correlation coefficient. The functional correlation coefficient is determined according to the dependency degree, for example, 0.9 if the target device j stops running and causes the operating efficiency of the target device i to decrease by more than 30%, 0.5 if the decrease is between 10% and 30%, and 0.1 if the decrease is less than 10%. The intensity attenuation ratio of the vibration shadow of the target device i at the position of the target device j is defined as the vibration interference coefficient of the target device i to the target device j. The physical correlation coefficient, the functional correlation coefficient, and the vibration interference coefficient are weighted and fused to obtain the correlation degree r i''j'' . The correlation degree r i''j''The value of the association degree between the target device i" and the target device j" ranges from 0 to 1, and the greater the value, the closer the association between the target device i" and the target device j".

[0114] The triggering condition of the first-level early warning is that the device health index of the target device to be evaluated is between H1 and 1, the remaining safe operation time T safe is greater than T1 hours, and the maximum association degree between the target device to be evaluated and all other target devices is less than r1. At this time, since the target device to be evaluated is in a good health state and the fault develops slowly, and is weakly associated with other target devices, it will not cause a chain effect, so only the monitoring frequency needs to be enhanced.

[0115] The triggering condition of the second-level early warning includes three situations: one is that the device health index of the target device to be evaluated is between H2 and H1, the remaining safe operation time T safe is between T2-T1 hours, and the association degree with at least one other target device is between r1 and r2; two is that the device health index of the target device to be evaluated is between H1 and 1, but the remaining safe operation time T safe is between T2-T1 hours and the association degree with at least one other target device is between r1 and r2; three is that the device health index of the target device to be evaluated is between H2 and H1, the remaining safe operation time T safe is greater than T1 hours, but the association degree with at least one other target device is between r3 and r2. At this time, the target device to be evaluated is in a medium health state or the fault develops quickly, and is moderately associated with other target devices, and may affect other target devices through vibration coupling or functional dependence if not intervened, so intelligent load adjustment needs to be performed, and the devices with an association degree greater than r3 are enhanced in monitoring.

[0116] The triggering condition of the third-level early warning includes three situations: one is that the device health index of the target device to be evaluated is less than H2, or the remaining safe operation time T safe is less than T2 hours; two is that the device health index of the target device to be evaluated is between H2 and H1, but the association degree with at least one other target device is greater than r2; three is that the target device to be evaluated has a device with a health index less than H3 among the other target devices with an association degree greater than r2 with the target device to be evaluated. At this time, the target device to be evaluated is in a high fault risk state, or the strong association with other target devices may cause a chain failure, so an emergency shutdown preparation needs to be triggered immediately, a maintenance instruction is sent synchronously, and load limitation is performed on all other target devices with an association degree greater than r3 with the target device to be evaluated. The target device to be evaluated that triggers the second-level early warning and the third-level early warning is marked as a fault device.

[0117] H1 is a first health threshold, H2 is a second health threshold, H3 is a third health threshold, T1 is a first remaining time threshold, T2 is a second remaining time threshold, r1 is a first correlation threshold, r2 is a second correlation threshold, and r3 is a third correlation threshold. H1 corresponds to the 95th percentile of the normal equipment health index distribution, such as 0.7, H2 corresponds to the 5th percentile of the fault equipment health index distribution, such as 0.3; H3 is between H2 and H1, such as 0.5, which belongs to the medium health level. T1 corresponds to the average time from the early fault that can be identified to the impact on operation, such as 72 hours, and T2 corresponds to the critical time point of the rapid development stage of the fault, such as 24 hours; r1 is the critical value of the fault conduction probability between devices below 5%, such as 0.3, and r2 is the critical value of the fault conduction probability above 80%, such as 0.7. r3 is between r1 and r2, such as 0.5, which belongs to the medium correlation, indicating that there is a certain possibility of fault conduction between the target devices, and the fault conduction probability is between 5% and 80%.

[0118] The three-level early warning response mechanism solves the "isolated judgment" defect caused by the traditional early warning relying on the single device state, and expands the single device early warning to system-level risk assessment through the device correlation influence matrix. The correlation calculation can be traced back to the device space position, ensuring that the influence of physical distance on correlation can be quantified; the vibration interference coefficient is directly related to the vibration propagation attenuation model, so that implicit vibration coupling is considered in early warning; without the fusion of the device correlation influence matrix, there may be over-warning or insufficient warning, such as ignoring the early failure of high-correlation devices leading to cascading failure, and after fusion, the early warning response is more in line with the actual risk of the system, which improves the accuracy of early warning and reduces the incidence of cascading failure.

[0119] Step S43, when triggering the secondary early warning or the tertiary early warning, an intelligent load adjustment strategy is executed based on the device correlation influence matrix;

[0120] According to the early warning level, the range of devices to be adjusted is determined: in the secondary early warning, the devices to be adjusted include the fault device and other target devices with high correlation, and the state change of such devices may have a significant impact on the fault device due to the close correlation; in the tertiary early warning, the range of devices to be adjusted is expanded to other target devices with a correlation with the fault device at a medium level or above to cope with the risk of possible cascading impact. The core goal of the load adjustment amount is to slow down the deterioration of the system health status, while ensuring that the actual load of all devices does not exceed 90% of the rated load to avoid secondary failures caused by overload.

[0121] The load adjustment amount of the faulty device needs to be determined in combination with the current health index, rated load and adjustment coefficient of the device, and the load adjustment amount = rated load x (1-current health index) x adjustment coefficient. Among them, the adjustment coefficient is not a fixed value, but changes dynamically with the degree of association of the device. The higher the degree of association, the smaller the value of the adjustment coefficient. This is because the mutual influence between devices with high degree of association is more significant, and excessive adjustment may damage the stable operation of associated devices. For example, when the degree of association is between 0.3 and 0.5, the adjustment coefficient is 0.6 to 0.8; when the degree of association is between 0.5 and 0.7, the adjustment coefficient is 0.4 to 0.6; and when the degree of association is greater than 0.7, the adjustment coefficient is 0.2 to 0.4. The load adjustment amount of other target devices associated with the faulty device is allocated according to the degree of association between the target device and the faulty device shown in the device association influence matrix. The higher the degree of association of the device, the greater the proportion of load adjustment amount it bears, ensuring that the total adjustment amount is conserved in the system, reducing the operating pressure of the faulty device to slow down the development of the fault, and maintaining the overall stability of the system through the coordinated adjustment of other associated target devices. The change of the vibration energy of the device needs to be monitored during the adjustment process, and the effectiveness of the adjustment is judged by comparing the deterioration speed of the health state of the faulty device before and after the adjustment. If the degree of slowing down is not as expected, the load adjustment amount needs to be recalculated. This process is coordinated with the results of the vibration shadow stripping in step S20, which can avoid the intensification of vibration coupling between devices due to load changes and solve the problem of ignoring system coordination in traditional single-device adjustment.

[0122] Step S44, generating an optimized control instruction sequence according to the load adjustment result.

[0123] The calculated fault device and the load adjustment amount of the other target devices associated therewith are converted into specific control parameters executable by the device, which vary depending on the type of the fault device or the other target devices associated therewith, such as the opening angle of the circuit breaker, the rotating speed of the fan, the tap position of the transformer, etc. These control parameters are arranged in time sequence and priority to form an instruction sequence, and the priority is determined by considering the current health index and the degree of association of the fault device and the other target devices associated therewith. The worse the health status of the device, the more urgent the adjustment is required. The target devices associated with the fault device have a higher degree of association, and the adjustment operation can simultaneously improve the operating state of the fault device and itself, so the adjustment priority of these two types of devices is improved and the sorting in the instruction sequence is earlier. The instruction sequence needs to include time markers, device identifiers, specific control parameters and execution thresholds, wherein the execution threshold is a prerequisite for instruction execution, which is determined based on the vibration range of safe operation of the device to ensure that the instruction is only executed when the state of the device allows. After the instruction is generated, it is transmitted to the remote control module through the substation industrial Ethernet, and a closed-loop feedback mechanism is established: immediately after the execution of each instruction, the vibration data of the fault device and the other target devices associated therewith are re-collected, and the health state prediction matrix and the fault development trend prediction curve are updated. If the feedback shows that the improvement degree of the health state of the fault device does not meet the expectation, i.e., the slowing down of the deterioration speed of the health state does not reach the set standard, the instruction correction process is triggered, and the control parameters are adjusted to avoid the operating state that may exacerbate the vibration, in combination with the real-time energy change law reflected by the dynamic energy distribution map in step S30. This process solves the problem of poor execution effect caused by the lack of dynamic adjustment of traditional control instructions, ensures that the control strategy always matches the actual state of the fault device and the other target devices associated therewith, and makes the achievements of early fault identification and trend prediction into actual maintenance actions. If this step is missing, the load adjustment strategy cannot be implemented, and the practical value of the whole scheme will be greatly reduced.

[0124] The device health state prediction matrix in step S40 realizes horizontal and vertical comparability of health states of different devices and different time nodes through multi-dimensional health quantification, and solves the one-sidedness of traditional single-device evaluation; the device correlation influence matrix not only considers physical and functional correlations, but also captures implicit correlations through vibration propagation characteristics, for example, discovering that two devices that are not directly electrically connected but share a concrete base cause fault conduction due to vibration coupling, and the identification of such implicit correlations makes the maintenance strategy more systematic; the intelligent load adjustment strategy combines with the coordinated adjustment of related devices, not only slows down the development of single-device faults, but also optimizes the overall operation efficiency of the system; step S40 upgrades the preventive maintenance of the unattended substation from "after-the-fact remedy" to "accurate prediction + active intervention", significantly reducing the risk of unplanned power outages, wherein the identification of implicit correlations realizes system-level correlation fault perception that is not involved in traditional single-device control, and this perception capability is formed through the deep cooperation of steps S40, S10, S20, and S30, and the technical fusion across steps enables the system to have global perception and adaptive adjustment capabilities, which cannot be achieved by individual steps.

[0125] Embodiment 2

[0126] This embodiment is based on embodiment 1 and provides an intelligent remote management and control platform for an unattended substation, as shown in Figure 5 , which includes:

[0127] A vibration source positioning module is configured to establish a vibration footprint ring for each device in the substation, arrange monitoring points on the boundary of the vibration footprint ring, collect vibration acceleration signals during the startup of the device, position the vibration source based on the vibration acceleration signals, and calculate the vibration contribution degree of the vibration source.

[0128] A vibration shadow stripping module is configured to establish a vibration propagation attenuation model and perform multi-level vibration shadow stripping based on the vibration source positioning results and the vibration contribution degree, and generate a device vibration feature vector and a vibration feature time-varying matrix.

[0129] A fault prediction module is configured to establish a dynamic energy distribution map based on the device vibration feature vector and the vibration feature time-varying matrix, identify weak faults and predict the fault development trend based on the dynamic energy distribution map, and generate a fault development trend prediction curve.

[0130] An intelligent control module is configured to construct a device health state prediction matrix based on the fault development trend prediction curve, and execute an intelligent load adjustment strategy to generate a system optimization control instruction sequence.

[0131] In the vibration source positioning module, the method of positioning the vibration source based on the vibration acceleration signals and calculating the vibration contribution degree of the vibration source includes:

[0132] Step S131, time domain analysis is performed on the vibration acceleration signal to extract a vibration intensity time sequence of each monitoring point;

[0133] Step S132, the vibration intensity time sequences of the monitoring points are combined to form a start-up vibration fingerprint matrix of the device, and the start-up vibration fingerprint matrix is normalized to obtain a normalized start-up vibration fingerprint matrix;

[0134] Step S133, the normalized start-up vibration fingerprint matrices under different load conditions are obtained to form a multi-condition vibration fingerprint set;

[0135] Step S134, vibration identity feature vectors of each device are extracted from the multi-condition vibration fingerprint set, and the vibration identity feature vectors of all devices are combined to establish a device vibration identity library;

[0136] Step S135, mixed vibration signals of each monitoring point are collected in real time to construct a real-time vibration observation matrix, and the real-time vibration observation matrix is matched with the device vibration identity library to identify a vibration source;

[0137] Step S136, vibration contribution degrees of each vibration source are calculated according to vibration intensity distributions of the identified vibration sources in the vibration footprint ring.

[0138] In the vibration shadow stripping module, the method for performing multi-stage vibration shadow stripping includes:

[0139] Step S221, a vibration source with the greatest impact on the target device is identified based on the vibration shadow identification matrix, and is marked as a greatest-impact vibration source, and a vibration shadow generated by the greatest-impact vibration source at the position of the target device is calculated;

[0140] Step S222, the vibration shadow generated by the greatest-impact vibration source is subtracted from the mixed vibration signal of the target device to complete first-stage stripping, and a first-stage stripped signal is obtained;

[0141] Step S223, on the basis of the first-stage stripping, secondary vibration sources are continuously identified and vibration shadows of the secondary vibration sources are stripped to obtain a second-stage stripped signal; the stripping process is repeated until a total intensity of vibration shadows of remaining vibration sources is lower than a set vibration intensity threshold, and a pure vibration signal is obtained;

[0142] Step S224, a stripping quality evaluation index Q is calculated according to the pure vibration signal, and when the stripping quality evaluation index Q is greater than a preset stripping quality threshold q0, it is considered that the stripping is successful; otherwise, the multi-stage vibration shadow stripping operation is re-performed on the mixed vibration signal.

[0143] The intelligent remote management and control platform of the unattended substation and the real-time monitoring and control method of the present application can be implemented in many ways. For example, the intelligent remote management and control platform of the unattended substation and the real-time monitoring and control method of the present application can be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is only for illustration, and the steps of the method of the present application are not limited to the above specifically described order, unless otherwise specifically stated.

[0144] In addition, the part of the above technical solution provided in the embodiments of the present application that is consistent with the implementation principle of the corresponding technical solution in the prior art is not described in detail to avoid excessive repetition.

[0145] The specific embodiments described above further illustrate the objects, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for intelligent remote real-time monitoring and control of unattended substations, characterized in that, The method includes: Establish a vibration footprint ring for each piece of equipment in the substation, arrange monitoring points at the boundary of the vibration footprint ring, and collect vibration acceleration signals during the equipment startup phase; locate the vibration source based on the vibration acceleration signals and calculate the vibration contribution of the vibration source; The method for establishing a vibration footprint ring for each piece of equipment in the substation includes: acquiring the spatial coordinate information, rated power, and installation foundation stiffness of each piece of equipment in the substation; determining the geometric center of each piece of equipment based on the spatial coordinate information; calculating the maximum vibration influence radius of each piece of equipment based on the rated power and installation foundation stiffness; forming a ring-shaped area on a horizontal plane with the geometric center of each piece of equipment as the center and the maximum vibration influence radius as the radius, wherein the ring-shaped area is the vibration footprint ring for each piece of equipment. Based on the vibration source location results and vibration contribution, a vibration propagation attenuation model is established and multi-level vibration shadow stripping is performed to generate equipment vibration feature vectors and time-varying vibration feature matrices. Based on the equipment vibration feature vector and the time-varying matrix of vibration features, a dynamic energy distribution map is established; based on the dynamic energy distribution map, minor faults are identified and fault development trends are predicted, and a fault development trend prediction curve is generated. Based on the fault development trend prediction curve, a device health status prediction matrix is ​​constructed, and an intelligent load adjustment strategy is executed to generate a system optimization control command sequence.

2. The intelligent remote real-time monitoring and control method for unattended substations according to claim 1, characterized in that, The method for locating vibration sources based on vibration acceleration signals is as follows: a vibration identity database of the equipment is established based on the vibration acceleration signals during the equipment startup phase, and the vibration source is located by pattern matching.

3. The intelligent remote real-time monitoring and control method for unattended substations according to claim 2, characterized in that, The method for establishing the equipment vibration identity database includes: Time-domain analysis was performed on the vibration acceleration signal to extract the vibration intensity time series of each monitoring point; The vibration intensity time series of each monitoring point are combined to form the equipment start-up vibration fingerprint matrix. The start-up vibration fingerprint matrix is ​​then normalized to obtain the normalized start-up vibration fingerprint matrix. Obtain the normalized startup vibration fingerprint matrix under different load conditions to form a multi-condition vibration fingerprint set; The vibration identity feature vectors of each device are extracted from the multi-condition vibration fingerprint set, and the vibration identity feature vectors of all devices are combined to establish a device vibration identity database.

4. The intelligent remote real-time monitoring and control method for unattended substations according to claim 3, characterized in that, The vibration source refers to equipment that is in operation. The method for locating vibration sources through pattern matching includes: Real-time acquisition of mixed vibration signals from various monitoring points is used to construct a real-time vibration observation matrix, where each element represents the vibration intensity at each monitoring point. The real-time vibration observation matrix is ​​then matched with a vibration identity database to identify vibration sources.

5. The intelligent remote real-time monitoring and control method for unattended substations according to claim 4, characterized in that, The method for identifying vibration sources by pattern matching between the real-time vibration observation matrix and the equipment vibration identity database is as follows: calculate the similarity between the real-time vibration observation matrix and each vibration identity feature vector in the equipment vibration identity database, and determine the source as a vibration source when the similarity exceeds the set similarity threshold.

6. The intelligent remote real-time monitoring and control method for unattended substations according to claim 5, characterized in that, The method for calculating the vibration contribution of the vibration source is as follows: The total vibration intensity of vibration source i is obtained by summing the vibration intensities of all monitoring points of vibration source i; the average vibration intensity of vibration source i is obtained by dividing the total vibration intensity of vibration source i by the number of monitoring points of vibration source i, where i is the index variable of vibration source. Calculate the sum of the average vibration intensities of all vibration sources; The ratio of the average vibration intensity of vibration source i to the sum of the average vibration intensities of all vibration sources is the vibration contribution of vibration source i.

7. The intelligent remote real-time monitoring and control method for unattended substations according to claim 6, characterized in that, The method for performing multi-level vibration shadow stripping includes: Based on the vibration source location results and vibration contribution, a vibration propagation attenuation model is constructed to generate a vibration shadow recognition matrix. Based on the vibration shadow recognition matrix, identify the vibration source that has the greatest impact on the target equipment, mark it as the vibration source with the greatest impact, and calculate the vibration shadow generated by the vibration source with the greatest impact at the location of the target equipment. The vibration shadow generated by the vibration source with the greatest influence is subtracted from the mixed vibration signal of the target equipment to complete the first stage of stripping, and the signal after the first stage of stripping is obtained. Based on the first-level stripping, secondary vibration sources are further identified and their vibration shadows are stripped away to obtain the signal after the second-level stripping. The stripping process is repeated until the total intensity of the vibration shadows of the remaining vibration sources is lower than the set vibration intensity threshold, thus obtaining a pure vibration signal.

8. The intelligent remote real-time monitoring and control method for unattended substations according to claim 7, characterized in that, The multi-level vibration shadow stripping process also includes: Based on the pure vibration signal, the peeling quality evaluation index Q is calculated. When the peeling quality evaluation index Q is greater than the preset peeling quality threshold q0, the peeling is considered successful; otherwise, the multi-level vibration shadow peeling operation is re-executed on the mixed vibration signal.

9. The intelligent remote real-time monitoring and control method for unattended substations according to claim 8, characterized in that, The method for generating the device vibration feature vector and the time-varying matrix of vibration features includes: Frequency domain analysis is performed on the pure vibration signal to calculate the power spectral density; the fundamental frequency, the frequency of each harmonic, and the power of each harmonic are identified from the power spectral density to construct the equipment vibration feature vector; the continuously acquired equipment vibration feature vectors are arranged sequentially in time order to construct the vibration feature time-varying matrix.

10. The intelligent remote real-time monitoring and control method for unattended substations according to claim 9, characterized in that, The method for identifying weak faults based on dynamic energy distribution maps includes: A vibration energy gradient vector is constructed based on the dynamic energy distribution map. The energy evolution trajectory of the target device is drawn according to the vibration energy gradient vector, and the trajectory deviation of the energy evolution trajectory is calculated. When the trajectory deviation exceeds the preset deviation threshold, a minor fault is considered to have been identified.

11. An intelligent remote control platform for unattended substations, used to implement the intelligent remote real-time monitoring and control method for unattended substations as described in any one of claims 1-10, characterized in that, The intelligent remote control platform for the unattended substation includes: Vibration source location module: used to establish a vibration footprint ring for each piece of equipment in the substation, arrange monitoring points at the boundary of the vibration footprint ring, collect vibration acceleration signals during the equipment startup phase; locate the vibration source based on the vibration acceleration signals and calculate the vibration contribution of the vibration source; Vibration Shadow Stripping Module: Used to establish a vibration propagation attenuation model based on the vibration source location results and vibration contribution, and perform multi-level vibration shadow stripping to generate equipment vibration feature vectors and vibration feature time-varying matrices; Fault prediction module: used to establish a dynamic energy distribution map based on the equipment vibration feature vector and the time-varying matrix of vibration features; to identify minor faults based on the dynamic energy distribution map and predict the fault development trend, and generate a fault development trend prediction curve; Intelligent control module: Constructs equipment health status prediction matrix based on fault development trend prediction curve, executes intelligent load adjustment strategy, and generates system optimization control command sequence.

Images