CN-122018485-A - Electrical control cabinet fault early warning method and system based on deep learning

CN122018485ACN 122018485 ACN122018485 ACN 122018485ACN-122018485-A

Abstract

The application relates to the technical field of state monitoring and fault diagnosis of electrical equipment, and discloses a fault early warning method and a system of an electrical control cabinet based on deep learning, wherein physical sensing data and control logic data are acquired and time sequence alignment is carried out by the method; analyzing control data to identify a working condition mode, updating a correction coefficient representing the heat dissipation attenuation of equipment by using an actual temperature drop rate in a self-calibration mode, reconstructing an effective thermodynamic parameter through a depth parameter estimation network and the correction coefficient in a monitoring mode, reversely calculating real-time contact impedance of a monitoring node based on a thermal balance differential equation, monitoring an impedance evolution trend, and generating an early warning signal when the impedance rises monotonically and is decoupled from a load current variation trend. The application realizes accurate monitoring and fault diagnosis of contact impedance under complex thermal environment and equipment aging conditions by integrating the physical mechanism model and the data driving network.

Inventors

HE MAOJIE
CHEN YIDE
QIAN XIXIN
ZHUANG QIANQIAN

Assignees

青岛振海船舶设备有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. The electric control cabinet fault early warning method based on deep learning is characterized by comprising the following steps of: Acquiring physical sensing data and control logic data in an electrical control cabinet, and aligning the physical sensing data with the control logic data in time sequence according to a unified clock source; analyzing the control logic data to identify a current working condition mode of the system, wherein the working condition mode comprises a self-calibration mode and a monitoring mode; In the self-calibration mode, calculating the ratio of the actual temperature drop rate to the theoretical temperature drop rate of the monitoring node, and updating a heat dissipation correction coefficient, wherein the heat dissipation correction coefficient represents the attenuation degree of the actual heat dissipation capacity of the electrical control cabinet relative to a design value; Under the monitoring mode, the control logic data is input into a preset depth parameter estimation network, a nominal thermodynamic parameter is output, and the nominal thermodynamic parameter is corrected by utilizing the heat dissipation correction coefficient to obtain an effective thermodynamic parameter; constructing a thermal balance differential equation of the monitoring node, substituting the physical sensing data and the effective thermodynamic parameters into an inverse solution form of the thermal balance differential equation, and calculating real-time contact impedance of the monitoring node; and monitoring the time evolution trend of the real-time contact impedance, and generating a fault early warning signal when the real-time contact impedance shows monotonous rising and is decoupled from the change trend of the load current.
2. The deep learning-based electrical control cabinet fault early warning method and system according to claim 1, wherein the physical sensing data comprises a current value of a loop where a monitoring node is located, a node temperature value of the monitoring node and an environmental temperature value in the electrical control cabinet; The control logic data comprises a start-stop instruction of the cooling fan and a load instruction of adjacent equipment; the identifying system for the current working condition mode specifically comprises the following steps: If the control logic data indicate that the cooling fan is in an on state and the current value is lower than a preset zero drift threshold value, judging that the self-calibration mode is adopted; And if the current value is larger than the zero drift threshold value, judging the monitoring mode.
3. The deep learning-based electrical control cabinet fault early warning method and system according to claim 1, wherein the updating of the heat dissipation correction coefficient specifically comprises: in a cooling stage after the load current is cut off, acquiring the actual temperature drop rate of the monitoring node; Based on Newton's law of cooling, calculating a theoretical temperature drop rate by using a current node temperature value, an environment temperature value and a reference heat exchange coefficient in a design state; And determining the ratio of the actual temperature drop rate to the theoretical temperature drop rate as a current heat dissipation correction coefficient, and storing the coefficient into a parameter holder for calling the monitoring mode.
4. The deep learning-based electrical control cabinet fault early warning method and system according to claim 1, wherein the depth parameter estimation network is a pre-trained multi-layer neural network, and the input and output relationship is specifically as follows: The input feature vector of the depth parameter estimation network at least comprises: fan running state, ambient temperature value and comprehensive external thermal potential characteristics; the comprehensive external thermal potential characteristics are obtained by weighting calculation according to load instructions of adjacent equipment, geometric distances of the adjacent equipment relative to the monitoring nodes and rated powers of the adjacent equipment; The output of the depth parameter estimation network is the nominal thermodynamic parameter, and specifically comprises a nominal convective heat transfer coefficient and external interference heat flux; The nominal convective heat transfer coefficient characterizes the theoretical heat transfer capacity in an ideal clean environment.
5. The deep learning-based electrical control cabinet fault early warning method and system according to claim 4, wherein the correcting the nominal thermodynamic parameter by using the heat dissipation correction coefficient specifically comprises: Reading the heat dissipation correction coefficient updated in the latest self-calibration mode; multiplying the nominal convective heat transfer coefficient by the heat dissipation correction coefficient to obtain an effective convective heat transfer coefficient; Taking the external disturbance heat flux directly as an effective external heat flux; The effective thermodynamic parameter consists of the effective convective heat transfer coefficient and the effective external heat flux.
6. The deep learning-based electrical control cabinet fault early warning method and system according to claim 5, wherein the thermal equilibrium differential equation describes that the internal energy increment in unit time of the monitoring node is equal to the algebraic sum of joule heat generated by current, effective external heat flux and convective heat dissipation power; the real-time contact impedance of the monitoring node is calculated, and the following calculation is specifically executed: Calculating a thermal inertia term which is the product of the equivalent heat capacity of the monitoring node and the node temperature variable rate; calculating a convection heat radiation item which is the product of the effective convection heat exchange coefficient, the effective heat radiation area and the node temperature rise; adding the thermal inertia term and the convection heat dissipation term, and subtracting the effective external heat flux to obtain total heat generation power; Dividing the total generated thermal power by the square of the current value to obtain the real-time contact impedance.
7. The deep learning-based electrical control cabinet fault early warning method and system according to claim 6, wherein when calculating the real-time contact impedance of the monitoring node, the method further comprises the step of numerical stability processing: Regularization parameters are introduced into denominators of division operation to prevent calculation singularities from occurring when current crosses zero; introducing non-negative physical constraint, and when the calculated molecular term is smaller than zero, forcedly setting the real-time contact impedance to be zero; And setting a minimum calculated current threshold value, and only when the theoretical temperature rise rate generated by the current value is greater than the resolution of the temperature sensor, executing the updating of the real-time contact impedance, otherwise, maintaining the calculated value at the last moment.
8. The deep learning-based electrical control cabinet fault early warning method and system according to claim 1, wherein the early warning model decoupled from the variation trend of the load current specifically comprises: establishing a sliding window, and carrying out statistical analysis on the real-time contact impedance and the load current in the window; Calculating the ratio of the relative change rate of the real-time contact impedance to the relative change rate of the load current to obtain a coupling factor; When the coupling factor is smaller than a preset decoupling threshold, determining that the change of the real-time contact impedance is independent of the change of the load current, and confirming the physical effectiveness of impedance data; when the coupling factor is larger than or equal to the decoupling threshold, the impedance data is judged to be influenced by the model error, and the current early warning signal is shielded.
9. The deep learning-based electrical control cabinet fault early warning method and system according to claim 8, wherein the generating the fault early warning signal specifically comprises: Acquiring a reference contact impedance determined in a system initialization stage; calculating the current drift percentage of the real-time contact impedance relative to the reference contact impedance; when the drift percentage is in a first interval, generating a first-level early warning signal for prompting the inspection of the fastening state; And when the drift percentage is in a second interval, generating a secondary early warning signal for prompting shutdown maintenance, wherein the lower limit value of the second interval is larger than the upper limit value of the first interval.
10. The deep learning-based electrical control cabinet fault early warning system, according to any one of claims 1 to 9, characterized by comprising: the data synchronization module is used for acquiring physical sensing data and control logic data and performing time sequence alignment; The state mapping module is used for identifying that the system is in a self-calibration mode or a monitoring mode according to the control logic data; The parameter calibration module is used for updating the heat dissipation correction coefficient based on the actual temperature drop rate in the self-calibration mode; The parameter reconstruction module is used for generating effective thermodynamic parameters by utilizing a depth parameter estimation network and the heat dissipation correction coefficient in a monitoring mode; the impedance calculating module is used for calculating real-time contact impedance based on a physical inverse decoupling algorithm by utilizing the physical sensing data and the effective thermodynamic parameters; and the trend analysis module is used for monitoring the evolution trend of the real-time contact impedance and executing fault early warning.

Description

Electrical control cabinet fault early warning method and system based on deep learning Technical Field The invention relates to the technical field of electrical equipment state monitoring and fault diagnosis, in particular to an electrical control cabinet fault early warning method and system based on deep learning. Background The electric control cabinet is used as a distribution and control pivot of the electric power system, and a breaker, a contactor and various wiring terminals are densely distributed in the electric control cabinet. In the long-term operation process, the electric connection point is easily loosened or the contact surface is easily deteriorated due to the influence of factors such as mechanical vibration, thermal expansion and contraction, surface oxidation and the like, so that the contact resistance is abnormally increased. If such faults cannot be found in time, the continuous joule heat accumulation will cause insulation aging, equipment burnout and even electrical fire accidents. Currently, monitoring for electrical contact failures relies mainly on temperature detection means, such as wireless temperature sensors or infrared thermal imaging techniques. However, the node temperature is the result of the combined action of load current, contact resistance, ambient temperature and heat dissipation conditions, and there is a limitation in determining based on only a single temperature threshold. Under heavy load working condition, normal electrical connection can generate higher temperature rise, false alarm is easy to trigger, and under light load or forced air cooling condition, even if poor contact exists, the node temperature can not reach a preset threshold value, so that report missing is caused. In order to obtain a more intrinsic contact state, some prior art attempts to invert the contact resistance using the principle of thermal equilibrium. This method requires the establishment of an accurate thermodynamic model in which the convective heat transfer coefficient and the external heat flux are critical boundary parameters that determine the accuracy of the calculation. However, in a practical industrial scenario, the thermal environment of an electrical control cabinet is extremely complex and dynamically changing. On one hand, the start and stop of the cooling fan, dust accumulation and blockage of the dust screen and aging attenuation of the fan can cause time-varying convection heat exchange capacity; On the other hand, the operational heating of adjacent equipment within the cabinet can create radiant heat interference that is difficult to measure. The existing model method generally simplifies the thermodynamic parameters into fixed constants or simple empirical formulas, cannot track the change of heat dissipation conditions and the interference of external thermal fields in real time, so that the calculated contact impedance value contains a large number of error components caused by environmental factors, the actual physical state of a contact surface is difficult to accurately reflect, and the practical application effect of the technology under complex working conditions is limited. Disclosure of Invention Aiming at the defects of the prior art, the invention provides a deep learning-based electrical control cabinet fault early warning method and a deep learning-based electrical control cabinet fault early warning system, which solve the problems that the existing contact impedance monitoring technology is easy to suffer from environmental temperature fluctuation, heat dissipation condition change and load dynamic interference, so that the calculation accuracy is low and the false alarm rate is high. In order to achieve the purpose, the invention is realized by the following technical scheme that the fault early warning method of the electric control cabinet based on deep learning comprises the following steps: first, the system performs data acquisition and time sequence alignment to acquire physical sensing data and control logic data in the electrical control cabinet. The physical sensing data reflects the real-time running state of the electric loop, the control logic data reflects the active heat dissipation behavior of the system and the external heat interference source state, and the two heterogeneous data are aligned through a unified clock source, so that a data base with consistent space and time is provided for subsequent multi-source information fusion. Secondly, the system identifies the current working condition mode based on control logic data, the working condition mode is divided into a self-calibration mode and a monitoring mode, and the system carries out logic judgment according to the running state of the fan and the current level of the loop, so that the self-adaptive sensing of different running stages of the equipment is realized. In the self-calibration mode, the system executes on-line correction of thermodynamic parameters, calculate