A real-time monitoring and fault prediction method and system for an electric energy metering box

Abstract

The application discloses a real-time monitoring and fault prediction method and system for an electric energy metering box, and the method comprises the following steps: acquiring original current waveform data and original temperature data of a terminal, performing time-frequency domain analysis and processing to obtain a waveform distortion sequence and a temperature rise sequence; calculating a time delay amount between the waveform distortion sequence and the temperature rise sequence, performing feature splicing on the time delay amount and the two sequences to obtain a signal lag feature sequence; performing feature fusion on the signal lag feature sequence to obtain a correlation feature matrix, determining a degradation state evaluation value of the terminal through the correlation feature matrix; extracting hidden danger features of the terminal, determining a hidden danger stage to which the degradation state evaluation value belongs, and generating a graded early warning instruction according to the matching of the hidden danger stage and a warning level. The application can associate asynchronous time sequence features of current waveform distortion and temperature change, and realize dynamic perception and graded early warning of the degradation process of the terminal.

Description

Technical Field

[0001] This invention relates to the field of online monitoring technology for power equipment, and in particular to a method and system for real-time monitoring and fault prediction of power metering boxes. Background Technology

[0002] Electricity metering is a crucial foundation for power system operation management and line loss analysis. Currently, the operation and maintenance of electricity metering devices typically involves regular manual inspections combined with post-event power outages for repair. When the contact resistance of the terminals increases due to long-term load fluctuations, mechanical vibrations, or thermal expansion and contraction, it is often only discovered passively through power outages or user complaints after metering inaccuracies, local overheating, or even burnout have occurred.

[0003] When terminal blocks deteriorate over a long period, causing the formation of an oxide film or intermittent micro-arcs at the contact interface, the contact resistance exhibits a non-linear change characteristic. The current waveform flowing through the terminal will become distorted, and the terminal surface temperature will rise abnormally. However, these two physical signals do not change in sync. Because changes in electrical characteristics are immediately apparent with the load, while temperature rise is constrained by the heat conduction time constant and has a time lag of minutes, the distortion of the current waveform is usually observed before the abnormal temperature rise. This asynchronous relationship between waveform distortion and temperature rise means that neither single current monitoring data nor temperature monitoring data can accurately reflect the actual degree of change in terminal contact resistance. The same distortion value may correspond to initial slight loosening or severe deterioration nearing burnout. Maintenance personnel lack effective multi-source information correlation methods to distinguish the differences in characteristics between different deterioration stages. In application scenarios oriented towards fault prediction and health management, the false alarm rate and false negative rate of anomaly detection are difficult to control.

[0004] In summary, existing technologies have the technical problem of failing to effectively correlate the asynchronous timing characteristics of current waveform distortion and temperature changes in order to accurately identify the deterioration stage of terminals. Summary of the Invention

[0005] This invention provides a method and system for real-time monitoring and fault prediction of an electricity metering box, in order to solve the technical problem in the prior art that it is impossible to effectively correlate the asynchronous timing characteristics of current waveform distortion and temperature change in order to accurately identify the deterioration stage of the wiring terminals.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for real-time monitoring and fault prediction of an electricity metering box, comprising: Acquire the raw current waveform data and raw temperature data of the terminals in the power metering device; Time-frequency domain analysis is performed on the original current waveform data and the original temperature data to obtain waveform distortion sequence and temperature rise sequence. Calculate the time delay between the waveform distortion sequence and the temperature rise sequence, and perform feature concatenation of the time delay with the waveform distortion sequence and the temperature rise sequence to obtain the signal hysteresis feature sequence; The signal hysteresis feature sequence is fused to obtain an associated feature matrix, and the degradation status is evaluated based on the associated feature matrix to calculate the degradation status evaluation value of the terminal block. The terminal block is subjected to hazard feature extraction to obtain hazard feature vector. Based on the hazard feature vector, the stage division boundary is determined and a fuzzy membership function is constructed. The degradation state evaluation value is calculated and judged according to the fuzzy membership function to determine the hazard stage. Based on the stage of the potential hazard, interval mapping is performed in a pre-established set of judgment intervals to obtain the target judgment interval. Based on the target judgment interval, the corresponding early warning level is matched to generate a graded early warning instruction for the terminal block.

[0007] In a second aspect, the present invention provides a real-time monitoring and fault prediction system for an electricity metering box, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect above.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention performs time-frequency domain analysis on the original current waveform data and the original temperature data to obtain a waveform distortion sequence and a temperature rise sequence, and calculates the time delay between the two sequences. The time delay is then concatenated with the two sequences to obtain a signal hysteresis feature sequence. Therefore, this invention can quantify the asynchronous temporal relationship between current waveform distortion and temperature change, linking the instantaneous response of electrical characteristics to load and the hysteresis response of temperature rise constrained by the heat conduction time constant to the same feature expression. Ultimately, this avoids the problem that single current monitoring data or temperature monitoring data cannot accurately reflect the actual degree of change in contact resistance, improving the accuracy of sensing the degradation process of the terminals.

[0009] (2) This invention obtains an associated feature matrix by performing feature fusion on the signal hysteresis feature sequence, and then performs feature extraction and fusion calculation on the associated feature matrix to determine the degradation status assessment value of the terminal block. Therefore, this invention can dynamically quantify the degree of terminal degradation by comprehensively considering multi-dimensional features such as waveform distortion, temperature gradient change, and thermal response delay, distinguishing between initial slight loosening and severe degradation nearing burnout corresponding to the same distortion value. Ultimately, the degradation status assessment value can dynamically reflect the actual cumulative degree of degradation at the terminal block contact interface, providing a reliable state quantity basis for determining potential problems.

[0010] (3) This invention extracts hidden danger features from the terminal block to obtain a hidden danger feature vector. Based on the hidden danger feature vector, it determines the stage division boundary and constructs a fuzzy membership function. It calculates and judges the deterioration state evaluation value to determine the hidden danger stage. Then, it performs interval mapping in a pre-established judgment interval set according to the hidden danger stage to match the warning level and generate a graded warning instruction for the terminal block. Thus, this invention can distinguish the differences between different deterioration stages based on hidden danger features such as the deterioration evolution rate and the frequency of local overheating, and match the corresponding warning level and scheduling priority for different stages. Finally, in the application scenario of fault prediction and health management, it reduces the false alarm rate and false negative rate of anomaly detection and realizes dynamic perception and graded warning of the terminal block deterioration process. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the real-time monitoring and fault prediction method for an electricity metering box provided in the first embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 The first embodiment of the present invention provides a method for real-time monitoring and fault prediction of an electricity metering box, including the following steps: S1, acquire the raw current waveform data and raw temperature data of the terminals in the power metering device; S2, perform time-frequency domain analysis on the original current waveform data and the original temperature data to obtain waveform distortion sequence and temperature rise sequence; S3, calculate the time delay between the waveform distortion sequence and the temperature rise sequence, and perform feature concatenation of the time delay with the waveform distortion sequence and the temperature rise sequence to obtain the signal hysteresis feature sequence; S4, perform feature fusion on the signal hysteresis feature sequence to obtain the correlation feature matrix, and perform degradation status assessment based on the correlation feature matrix to calculate the degradation status assessment value of the terminal block. S5, extract the hidden danger features of the wiring terminal to obtain the hidden danger feature vector, determine the stage division boundary based on the hidden danger feature vector and construct the fuzzy membership function, calculate and judge the deterioration state evaluation value according to the fuzzy membership function, and determine the hidden danger stage; S6. Based on the stage of the potential hazard, perform interval mapping in the pre-established set of judgment intervals to obtain the target judgment interval, match the corresponding early warning level according to the target judgment interval, and generate a graded early warning instruction for the terminal block.

[0014] In step S1, the raw current waveform data and raw temperature data of the terminals in the power metering device are acquired, including: The analog current signal at the terminal is acquired, and the analog current signal is converted from analog to digital to obtain a digital current sequence as the original current waveform data. The resistance change of the terminal is obtained, and the resistance change is converted into a voltage drop value by a bridge circuit. The voltage drop value is then converted into raw temperature data according to a pre-calibrated temperature-resistance mapping relationship.

[0015] In one implementation, after acquiring the raw current waveform data and raw temperature data of the terminals in the energy metering device, the method further includes: The original current waveform data and the original temperature data are timestamp aligned according to the synchronization clock signal to obtain aligned current waveform data and aligned temperature data. The aligned current waveform data and the aligned temperature data are used as the original current waveform data and the original temperature data in the time-frequency domain analysis and processing step. If the time stamps of the aligned current waveform data and the aligned temperature data are not continuous, data interpolation or retransmission request processing is performed.

[0016] In one implementation, this embodiment acquires an analog current signal at the terminal block using a current sensing unit located at the terminal block. The bandwidth of the current sensing unit needs to cover the current waveform distortion caused by nonlinearity of the contact interface and intermittent micro-arcs. For example, the current sensing unit can be a Hall current sensor or a Rogowski coil. Subsequently, the analog current signal is converted from analog to digital to obtain a digital current sequence, which is used as the original current waveform data. The sampling frequency of the analog-to-digital conversion is determined based on the highest frequency of the harmonic component to be extracted. For example, the sampling frequency is not lower than 50 kHz, so that the subsequent frequency domain harmonic extraction process can completely preserve the current waveform distortion caused by nonlinearity of the contact interface and intermittent micro-arcs.

[0017] In one implementation, the resistance change of the terminal is obtained, and the resistance change is converted into a voltage drop value using a Wheatstone bridge. The voltage drop value is then converted into raw temperature data according to a pre-calibrated temperature-resistance mapping relationship. Specifically, a temperature sensing element is mounted at a predetermined detection position on the terminal. The resistance of the temperature sensing element changes with the temperature of the terminal. For example, a thermistor can be used. The resistance change is the deviation of the temperature sensing element's resistance value from a reference resistance value at the current sampling moment. The resistance change is input into a Wheatstone bridge conversion circuit to obtain a voltage drop value proportional to the resistance change. The temperature-resistance mapping relationship is pre-established through calibration experiments, recording the correspondence between voltage drop values ​​and temperature detection values ​​under different temperature conditions, and storing this information in the form of a lookup table or fitted curve. By inputting the voltage drop value into the temperature-resistance mapping relationship, the raw temperature data at the current sampling moment can be retrieved.

[0018] It should be noted that the synchronization clock signal is provided by a clock synchronization unit, which is exemplarily a second-level synchronization pulse signal or a network time signal. The acquisition channel corresponding to the current sensing unit and the acquisition channel corresponding to the temperature sensing element are both assigned a sampling timestamp based on the synchronization clock signal, so that the original current waveform data and the original temperature data are under a unified time reference.

[0019] In one implementation, the original current waveform data and the original temperature data are timestamped according to a synchronous clock signal to obtain aligned current waveform data and aligned temperature data, which are then used as the original current waveform data and original temperature data in the time-frequency domain analysis processing step. Specifically, based on the clock beat of the synchronous clock signal, current sampling points and temperature sampling points within the same time beat are merged under the same time mark to form timestamped data. Subsequently, it is determined whether the timestamped data of the timestamped data is continuous, that is, whether the interval between adjacent timestamped data is equal to a preset sampling interval. If the timestamped data is continuous, the timestamped data is encapsulated into a target data frame; if the timestamped data of the aligned current waveform data and the aligned temperature data are not continuous, it is determined that there is a data gap in the current time period, and data interpolation processing is performed or a retransmission request is sent to the acquisition side to reacquire the data of the missing time period, so as to avoid the data of the missing time period from participating in subsequent calculations and introducing errors. For example, when the missing duration is less than a preset duration threshold, linear interpolation is performed to complete the missing period; when the missing duration is greater than or equal to the preset duration threshold, a retransmission request is sent to the acquisition side to reacquire the original data of the missing period.

[0020] It should be noted that the target data frame includes a frame header identifier and a checksum. The frame header identifier is used to identify the start position and data type of the data frame, enabling the receiving side to correctly segment each frame of data from the continuous data stream. The checksum is calculated based on all fields in the data frame except for the checksum field, and is exemplarily a cyclic redundancy check (CRC) code. It is used by the receiving side to verify whether bit errors have occurred in the data frame during transmission. Data frames that fail the checksum are discarded and retransmitted, thereby ensuring the integrity and reliability of the original current waveform data and original temperature data participating in subsequent processing. The original current waveform data and original temperature data processed in subsequent steps S2 to S6 are all extracted from the target data frame, thereby ensuring that the processed objects in each step are consistent with the timestamp-aligned and encapsulated data.

[0021] In step S2, the original current waveform data and the original temperature data are subjected to time-frequency domain analysis to obtain a waveform distortion sequence and a temperature rise sequence, including: Frequency domain harmonic extraction is performed on the original current waveform data to obtain a waveform distortion sequence; The original temperature data is subjected to a first-order forward difference operation to obtain a temperature gradient vector; the temperature gradient vector is then subjected to a moving average filtering process to obtain a temperature rise sequence.

[0022] In one implementation, this embodiment performs frequency domain harmonic extraction on the original current waveform data to obtain a waveform distortion sequence. Specifically, a window function is applied to the original current waveform data, followed by a Fast Fourier Transform to obtain the current spectrum. The fundamental wave and all harmonic components are extracted from the current spectrum, and the total harmonic distortion rate of the current at each sampling time is calculated. The total harmonic distortion rates of the current at each sampling time are arranged in chronological order to form the waveform distortion sequence. The deeper the nonlinearity of the contact interface and the more frequent the intermittent micro-arcs, the more obvious the current waveform distortion, and the larger the value of the waveform distortion sequence. Therefore, the waveform distortion sequence can characterize the strength of the nonlinearity of the contact interface and the intermittent micro-arcs.

[0023] In one implementation, a first-order forward differencing operation is performed on the raw temperature data at a preset time step to obtain a temperature gradient vector. Specifically, temperature sampling points are extracted from the raw temperature data according to the preset time step, exemplarily, the preset time step is 1 second; for each temperature sampling moment, the difference between the temperature sampling value at that moment and the temperature sampling value at the next time step is calculated as the temperature gradient at that moment, and all temperature gradients are arranged in chronological order to form the temperature gradient vector. Subsequently, the temperature gradient vector is subjected to moving average filtering processing, exemplarily, the moving average window is 10 sampling points to suppress gradient jitter caused by temperature measurement noise, and the filtering results are arranged in chronological order to obtain a temperature rise sequence representing the temperature rise trend.

[0024] In step S3, the time delay between the waveform distortion sequence and the temperature rise sequence is calculated, and the time delay is concatenated with the waveform distortion sequence and the temperature rise sequence to obtain a signal hysteresis feature sequence, including: Perform cross-correlation operation on the waveform distortion sequence and the temperature rise sequence to obtain a cross-correlation function sequence; extract the offset corresponding to the peak value of the cross-correlation function sequence as the time delay; Align the waveform distortion sequence with the temperature rise sequence using the time delay as the center; The time delay is combined with the distortion amplitude feature in the waveform distortion sequence and the temperature gradient feature in the temperature rise sequence to output a signal hysteresis feature sequence.

[0025] In one implementation, this embodiment performs a cross-correlation operation on the waveform distortion sequence and the temperature rise sequence to obtain a cross-correlation function sequence. Before performing the cross-correlation operation, the waveform distortion sequence is aggregated and downsampled according to the preset time step, so that it is under the same time reference as the temperature rise sequence, to ensure that the offset extracted by the cross-correlation function can be directly converted into a time delay amount according to the preset time step. Specifically, the normalized cross-correlation function values ​​of the waveform distortion sequence and the temperature rise sequence under different offsets are calculated one by one within the preset offset range. The cross-correlation function values ​​under all offsets are arranged in order of offset to form the cross-correlation function sequence. Subsequently, the offset corresponding to the peak value of the cross-correlation function sequence is extracted as the time delay amount, and the time delay amount is multiplied by the preset time step to obtain the time delay amount expressed in time units. The time delay amount is constrained by the thermal conduction time constant of the terminal block and is usually in the minute range, exemplarily tens of seconds to several minutes.

[0026] It should be noted that the time delay is the same as the thermal response delay. Changes in electrical characteristics are immediately apparent with load, while temperature rise is delayed due to the constraint of the heat conduction time constant. The thermal response delay quantifies the time interval by which waveform distortion precedes temperature rise. Different degrees of terminal degradation alter the heat conduction path and heat dissipation conditions at the contact interface, thus changing the thermal response delay. Therefore, the time delay itself is an effective characteristic of the degradation state.

[0027] In one implementation, the waveform distortion sequence and the temperature rise sequence are aligned with the thermal response delay as the center. Specifically, the temperature rise sequence is shifted forward along the time axis by the number of sampling points corresponding to the thermal response delay, so that the distortion features in the waveform distortion sequence and the corresponding temperature rise features in the temperature rise sequence are aligned on the same time reference, eliminating the temporal misalignment between the two sequences and providing a temporally consistent data foundation for subsequent feature fusion. The aligned waveform distortion sequence and temperature rise sequence then participate in the feature splicing in step S3, providing temporally consistent distortion amplitude features and temperature gradient features for the signal hysteresis feature sequence.

[0028] In one implementation, based on the waveform distortion sequence and the temperature rise sequence, the time delay is concatenated with the distortion amplitude feature in the waveform distortion sequence and the temperature gradient feature in the temperature rise sequence to output a signal lag feature sequence. Specifically, the distortion amplitude feature exemplarily includes the mean amplitude and peak amplitude of the waveform distortion sequence within the current statistical window, and the temperature gradient feature exemplarily includes the mean gradient and peak gradient of the temperature rise sequence within the current statistical window. The distortion amplitude feature, the temperature gradient feature, and the time delay are concatenated into a feature vector according to a preset dimensional order, and updated window by window as the statistical window slides, forming the signal lag feature sequence arranged in time order. The signal lag feature sequence simultaneously includes the immediate response of the electrical characteristics, the temperature rise lag response, and the time lag relationship between the two.

[0029] In step S4, feature fusion is performed on the signal hysteresis feature sequence to obtain an associated feature matrix, and a degradation status assessment is performed based on the associated feature matrix to calculate the degradation status assessment value of the terminal block.

[0030] The feature fusion of the signal hysteresis feature sequence to obtain the associated feature matrix includes: Extract the current amplitude sequence from the original current waveform data and construct a thermal balance mapping relationship based on the preset terminal thermal time constant; Substitute the temperature gradient and the current amplitude sequence in the signal hysteresis feature sequence into the thermal balance mapping relationship, and perform inversion calculation to obtain the change in contact resistance. The change in contact resistance is added as a new feature dimension and concatenated into the signal hysteresis feature sequence to obtain an association feature matrix reflecting the change in contact resistance.

[0031] In one implementation, this embodiment performs point-by-point feature concatenation of the signal hysteresis feature sequence according to time steps to obtain a multidimensional input sequence. Specifically, for each time step, the feature vector of that time step in the signal hysteresis feature sequence, the measured temperature rise value of that time step, and the current amplitude are concatenated, so that each time step of the multidimensional input sequence contains multiple dimensions such as distortion amplitude features, temperature gradient features, time delay, measured temperature rise value, and current amplitude, forming a multidimensional input sequence arranged by time steps; wherein, the measured temperature rise value is determined by the temperature rise of the temperature value at the corresponding sampling time in the original temperature data relative to the ambient temperature.

[0032] In one implementation, a thermal balance mapping relationship is constructed based on the terminal thermal time constant, and the change in contact resistance is inverted from the measured temperature rise, temperature gradient, and current amplitude in the multidimensional input sequence. Specifically, the terminal thermal time constant is pre-determined based on the material, structure, and installation conditions of the wiring terminal, and is denoted as... This represents the time characteristic parameter of the terminal temperature stabilizing after a power step change.

[0033] In this embodiment, This represents the change in contact resistance to be inverted, expressed as... The equivalent thermal conductance of a terminal is expressed as its reciprocal, which is the equivalent thermal resistance of the terminal. This indicates the current terminal temperature rise (relative to ambient temperature), in... This represents the temperature gradient at the current moment, i.e. ,by This represents the current amplitude. Based on the first-order lumped-parameter thermal system model, the transient heat balance equation is: ,in The equivalent heat capacity of the terminal is used. This is based on the thermal time constant. and temperature gradient Substitute and organize According to Joule's law, the Joule heating power caused by contact resistance satisfies Solving the above equations simultaneously yields the inversion formula. This formula is the specific expression of the heat balance mapping relationship.

[0034] Therefore, the temperature rise at each time step in the multidimensional input sequence is... Temperature gradient With current amplitude By inputting this mapping relationship, the change in contact resistance at each time step can be retrieved. When the current amplitude is 0, the retrieval calculation for that time step is skipped, and the change in contact resistance uses the value from the previous time step. From this formula, it can be seen that when the system approaches thermal equilibrium, i.e. As the expression approaches zero, it automatically degenerates into a steady-state inversion formula. When in a transient state of heating, i.e. When greater than 0, The item on the current temperature rise Dynamic corrections are performed to prevent the inversion results from being distorted by thermal hysteresis. The change in contact resistance with temperature rise... and temperature gradient It increases with the increase of the magnitude of the current, and decreases with the increase of the square of the current amplitude.

[0035] It should be noted that since the temperature data and current data have been aligned with the thermal response time delay in step S3, the inverted contact resistance change is under the same time series reference as the electrical characteristics at the same time step, thus avoiding the influence of phase deviation caused by thermal inertia on the inversion results.

[0036] In one implementation, the change in contact resistance is added as a new feature dimension and concatenated into the signal hysteresis feature sequence to obtain an association feature matrix reflecting the change in contact resistance. Specifically, the change in contact resistance at each time step is added as a new feature dimension and concatenated column-wise with the distortion amplitude feature, temperature gradient feature, and time delay of the corresponding time step in the signal hysteresis feature sequence to form the association feature matrix. Each row of the association feature matrix corresponds to a time step, and each column corresponds to the distortion amplitude feature, temperature gradient feature, time delay, and change in contact resistance, thereby completely preserving the correlation between electrical features, thermal features, and time hysteresis features in matrix form.

[0037] The degradation status assessment is performed based on the correlation feature matrix, and the degradation status assessment value of the wiring terminal is calculated, including: Feature extraction is performed on the associated feature matrix to obtain a multidimensional state feature vector containing a temperature gradient distribution sub-vector and an electrical feature distribution sub-vector; Based on the temperature gradient distribution sub-vector and the electrical characteristic distribution sub-vector in the multidimensional state feature vector, a weighted summation is performed using the corresponding preset weight coefficients, and the weighted summation result is normalized to obtain the initial deterioration state evaluation value. Therefore, this embodiment obtains a multi-dimensional state feature vector by extracting features from the associated feature matrix, and dynamically quantifies the degradation state evaluation value of the terminal block based on weighted mapping and smoothing processing.

[0038] In one implementation, this embodiment extracts features from the associated feature matrix to obtain a multidimensional state feature vector, which includes a temperature gradient distribution sub-vector and an electrical feature distribution sub-vector. Specifically, the temperature gradient distribution sub-vector is composed of the time distribution statistical features of thermal feature-related dimensions such as temperature gradient and contact resistance change in the associated feature matrix, exemplarily including the mean, peak value, and variance of the temperature gradient; the electrical feature distribution sub-vector is composed of the time distribution statistical features of electrical feature-related dimensions such as current amplitude, waveform distortion amplitude, and time delay in the associated feature matrix, exemplarily including the mean, peak value, and variance of current amplitude, waveform distortion amplitude, and time delay; the temperature gradient distribution sub-vector and the electrical feature distribution sub-vector are concatenated to form the multidimensional state feature vector.

[0039] In one implementation, the multidimensional state feature vectors are weighted and summed according to preset weight coefficients. Specifically, the preset weight coefficients are determined based on the absolute value of the Pearson correlation coefficient between each feature dimension and the measured degree of degradation; the feature dimension with a stronger correlation is assigned a larger weight. For example, the weights corresponding to each dimension of the temperature gradient distribution sub-vector and each dimension of the electrical feature distribution sub-vector are 0.6 and 0.4, respectively, and the weights of each feature dimension within the same sub-vector are normalized according to their correlation coefficient with the degree of degradation. The weighted summation result is normalized to obtain an initial degradation state evaluation value ranging from 0 to 1, where a larger value indicates a deeper degree of degradation of the terminal block.

[0040] In one implementation, the multidimensional state feature vector encapsulates the distortion amplitude feature, temperature gradient feature, time delay amount, and contact resistance change amount of each time step. The above feature dimensions together constitute a comprehensive characterization of the terminal deterioration state. After weighted mapping and smoothing, a dynamically quantified deterioration state evaluation value can be obtained.

[0041] In one implementation, when the initial degradation state assessment value is greater than a preset degradation alarm threshold, the initial degradation state assessment value is smoothed by median filtering to determine the degradation state assessment value of the terminal block. Specifically, the preset degradation alarm threshold is determined by statistically analyzing the sequence of initial degradation state assessment values ​​calculated under historical normal operating conditions. Specifically, historical normal time-series data covering typical operating conditions are collected, the initial degradation state assessment values ​​for each time window are calculated using the aforementioned method, their probability distribution is statistically analyzed, and the 95th percentile is taken as the preset degradation alarm threshold. It should be noted that the preset degradation alarm threshold is only used as a trigger condition for whether to perform median filtering and smoothing on the initial degradation state assessment value. This is used to suppress jumps in assessment values ​​caused by occasional factors such as load changes, and does not participate in the determination of the hidden danger stage. When the initial degradation state assessment value exceeds the preset degradation alarm threshold, it indicates that the terminal has entered a degradation state that requires special attention. At this time, median filtering is performed on the initial degradation state assessment values ​​at multiple consecutive assessment times. For example, the median filtering window is 5 consecutive assessment times, and the filtering result is output as the degradation state assessment value of the terminal. When the initial degradation state assessment value does not exceed the preset degradation alarm threshold, the initial degradation state assessment value is directly determined as the degradation state assessment value of the terminal.

[0042] When the initial degradation status assessment value is greater than the preset degradation alarm threshold, the initial degradation status assessment value is smoothed by median filtering to determine the degradation status assessment value of the terminal block.

[0043] In step S5, the hidden danger features are extracted from the wiring terminal to obtain the hidden danger feature vector. Based on the hidden danger feature vector, the stage division boundary is determined and a fuzzy membership function is constructed. The deterioration state evaluation value is calculated and judged according to the fuzzy membership function to determine the hidden danger stage.

[0044] Specifically, the potential hazard features of the wiring terminals are extracted to obtain a potential hazard feature vector, including: Extract multiple degradation status evaluation values ​​of the terminal block arranged in chronological order within a historical preset statistical period to form an evaluation value time series; perform linear regression calculation on the evaluation value time series to obtain the slope of the evaluation value time series as the degradation evolution rate; The local overheating frequency is obtained by counting the number of times that the sampling points of the terminal block exceed the preset overheating threshold in the original temperature data within the preset statistical time period. A hazard feature vector is constructed based on the degradation evolution rate and the frequency of local overheating.

[0045] The process of determining the stage division boundary based on the hazard feature vector includes: Obtain the hazard feature vectors and the deterioration state evaluation values ​​corresponding to each hazard feature vector at multiple sampling times within a preset historical period, and construct a two-dimensional feature sample set; Using the degradation state assessment value as the horizontal axis and the feature values ​​of each dimension of the hidden danger feature vector as the vertical axis, cluster analysis is performed on the two-dimensional feature sample set to obtain multiple clusters; Extract the projection interval of each cluster on the horizontal axis, and use the boundary value between the projection intervals of adjacent clusters as the stage division boundary between each hidden danger stage.

[0046] A fuzzy membership function is constructed using the stage division boundary. The deterioration state assessment value is input into the fuzzy membership function for mapping calculation. The hidden danger stage to which the deterioration state assessment value belongs is determined according to the maximum membership rule.

[0047] In one implementation, this embodiment extracts time-series data of the degradation status assessment values ​​within a preset statistical period. For example, the preset statistical period is the past 24 hours. A linear regression calculation is performed on the time-series data with time as the independent variable. The slope of the regression is the degradation evolution rate. A positive degradation evolution rate indicates that the degree of degradation deepens over time; a larger value indicates faster degradation evolution.

[0048] In one implementation, the frequency of local overheating is obtained by counting the number of sampling points in the original temperature data that exceed a preset overheating threshold within a preset statistical period. Specifically, the preset overheating threshold is set according to the material of the terminal block and its corresponding upper limit of allowable temperature. Different terminal materials have different upper limits of allowable temperature. For example, the upper limit of allowable temperature for copper terminal blocks is higher than that for aluminum terminal blocks. Therefore, corresponding overheating thresholds are preset according to the terminal material. For example, 90°C can be set for copper terminal blocks, and 75°C can be set for aluminum or copper-aluminum transition terminal blocks as the upper limit of allowable temperature. The preset overheating threshold is determined based on the principle of not exceeding the corresponding upper limit of allowable temperature. Within the same time window corresponding to the preset statistical period, the temperature of each sampling point of the original temperature data is compared with the preset overheating threshold one by one. Sampling points that exceed the preset overheating threshold are recorded as sampling points that exceed the limit. The number of occurrences of sampling points that exceed the limit within the time window is counted to obtain the frequency of local overheating. The frequency of local overheating and the rate of degradation evolution together constitute the hazard characteristics characterizing the development trend of the hazard.

[0049] In one implementation, a hazard feature vector is constructed based on the degradation evolution rate and the frequency of local overheating, and the stage division boundary is determined based on the hazard feature vector. Specifically, the degradation evolution rate and the frequency of local overheating are arranged in a preset order to form the hazard feature vector. Subsequently, the hazard feature vector and the degradation state assessment value corresponding to each hazard feature vector are obtained at multiple sampling times within a preset historical period to construct a two-dimensional feature sample set. The preset historical period is exemplarily the most recent 30 natural days, which covers a variety of typical load conditions, and the sample size is sufficient to support the statistical stability of the clustering results. The two-dimensional feature sample set is clustered with the degradation state assessment value on the horizontal axis and the feature values ​​of each dimension of the hazard feature vector on the vertical axis. K-means clustering is used for example, and the number of clusters is set to 3 to obtain multiple clusters. The projection interval of each cluster on the horizontal axis is extracted, and the intersection value between the projection intervals of adjacent clusters is used as the stage division boundary between each hazard stage. For example, the stage division boundary divides the degree of hidden danger into three stages: early deterioration stage, accelerated deterioration stage, and near-failure stage, which correspond to initial slight loosening, rapid deterioration, and severe deterioration nearing burnout, respectively.

[0050] In one implementation, a fuzzy membership function is constructed using the stage division boundary. The deterioration state assessment value is input into the fuzzy membership function for mapping calculation. The hazard stage to which the deterioration state assessment value belongs is determined according to the maximum membership rule. Specifically, a fuzzy membership function corresponding to each hazard stage is constructed using the stage division boundary as a node, exemplarily a triangular membership function. The deterioration state assessment value is input into the fuzzy membership function to calculate its membership degree to each hazard stage. The membership degrees of each hazard stage are compared, and the hazard stage with the highest membership degree is determined as the hazard stage to which the deterioration state assessment value belongs. Since the deterioration state assessment value is a continuous quantity, using a fuzzy membership function for mapping calculation can avoid small fluctuations in the assessment value near the stage boundary causing the stage determination result to jump back and forth, thus improving the stability of the hazard stage determination.

[0051] In step S6, interval mapping is performed in a pre-established set of judgment intervals according to the stage of the hidden danger to obtain the target judgment interval. The corresponding warning level is matched according to the target judgment interval to generate a graded warning instruction for the wiring terminal.

[0052] Specifically, based on the target determination interval, a corresponding early warning level is matched to generate a graded early warning instruction for the wiring terminal, including: Extract the warning level and interval boundary values ​​corresponding to the target determination interval; wherein, the warning level is determined by the standard hazard level parameter associated with the target determination interval, and the standard hazard level parameter is pre-calibrated based on the hazard level classification of historically confirmed fault cases; Read the terminal identification code of the terminal block, and generate an initial warning command after encoding. The initial warning instruction is pushed into the instruction issuance queue. The queue priority is determined according to the warning level. After scheduling is performed according to the queue priority, the hierarchical warning instruction for the terminal block is output.

[0053] In one implementation, this embodiment performs interval mapping within a pre-established set of judgment intervals based on the identified hazard stage to obtain a target judgment interval. Specifically, the set of judgment intervals is pre-established based on historically confirmed fault cases and corresponding hazard stage annotation data. For each sample whose hazard stage is confirmed on-site, the numerical range of its deterioration state assessment value and the standard hazard level parameter are extracted. The interval boundary values ​​and the standard hazard level parameter are used as elements of the judgment interval, and a one-to-one mapping relationship between the judgment interval and the hazard stage is established. The standard hazard level parameter is determined with reference to the power industry's classification conventions for electrical connection overheating and poor contact faults. By querying the mapping relationship based on the identified hazard stage, the target judgment interval corresponding to the hazard stage can be obtained.

[0054] In one implementation, the target judgment interval is mapped from the hidden danger stage to a pre-established set of judgment intervals. Each judgment interval is pre-associated with a corresponding warning level and standard hazard level parameter. Specifically, the standard hazard level parameter is pre-calibrated based on the hazard level classification of historically confirmed fault cases, and the reference values ​​of the interval boundary values ​​and standard hazard level parameters under each warning level are recorded. The standard hazard level parameter associated with the target judgment interval directly gives its corresponding warning level, so the warning level can be directly determined based on the target judgment interval.

[0055] In one implementation, based on the warning level associated with the target determination interval, the terminal identification code of the wiring terminal is read, and an initial warning command is generated after encoding processing. For example, the warning level is divided into three levels: attention level, alarm level, and emergency level, where the emergency level corresponds to the near-failure stage, the alarm level corresponds to the accelerated degradation stage, and the attention level corresponds to the early degradation stage. The terminal identification code is exemplarily composed of the energy metering box number and the terminal number, used to uniquely identify the wiring terminal with potential hazards; the encoding processing involves encapsulating the warning level, terminal identification code, degradation status assessment value, and warning generation timestamp into an initial warning command according to a preset message format.

[0056] In one implementation, the initial warning instruction is pushed into an instruction issuance queue. The queue priority is determined based on the warning level, and scheduling is performed according to the queue priority before outputting tiered warning instructions for the terminal block. Specifically, different warning levels correspond to different queue priorities. For example, the emergency level warning instruction has the highest queue priority, followed by the alarm level, and the attention level has the lowest. The instruction issuance queue is scheduled and output in descending order of queue priority. The output tiered warning instructions are pushed to the operation and maintenance monitoring platform, allowing operation and maintenance personnel to take differentiated handling measures according to the warning level. Thus, serious potential hazards nearing failure can be addressed first, while early minor hazards are alerted with lower priority, preventing warning resources from being filled by low-level warnings.

[0057] In summary, this invention performs time-frequency domain analysis on the original current waveform data and original temperature data to obtain waveform distortion sequences and temperature rise sequences. It then calculates the time delay between the two sequences to obtain a signal hysteresis feature sequence. Feature fusion is then performed to obtain a correlation feature matrix to determine the degradation status assessment value of the terminal block. Subsequently, hidden danger features are extracted from the terminal block. Based on the hidden danger feature vector, stage division boundaries are determined and a fuzzy membership function is constructed. The degradation status assessment value is calculated and judged to determine the hidden danger stage, and interval mapping is performed to match the warning level. This achieves effective correlation between the asynchronous time-series features of the terminal block's current waveform distortion and temperature change, accurately identifying the degradation stage of the terminal block and implementing graded warnings.

[0058] The second embodiment of the present invention provides a real-time monitoring and fault prediction system for an electricity metering box, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method described in the above method embodiment.

[0059] It should be noted that the real-time monitoring and fault prediction system for an electricity metering box provided in this embodiment of the invention is used to execute all the process steps of the real-time monitoring and fault prediction method for an electricity metering box in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0060] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0061] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for real-time monitoring and fault prediction of an electricity metering box, characterized in that, include: Acquire the raw current waveform data and raw temperature data of the terminals in the power metering device; Time-frequency domain analysis is performed on the original current waveform data and the original temperature data to obtain waveform distortion sequence and temperature rise sequence. Calculate the time delay between the waveform distortion sequence and the temperature rise sequence, and perform feature concatenation of the time delay with the waveform distortion sequence and the temperature rise sequence to obtain the signal hysteresis feature sequence; The signal hysteresis feature sequence is fused to obtain an associated feature matrix, and the degradation status is evaluated based on the associated feature matrix to calculate the degradation status evaluation value of the terminal block. The terminal block is subjected to hazard feature extraction to obtain hazard feature vector. Based on the hazard feature vector, the stage division boundary is determined and a fuzzy membership function is constructed. The degradation state evaluation value is calculated and judged according to the fuzzy membership function to determine the hazard stage. Based on the stage of the potential hazard, interval mapping is performed in a pre-established set of judgment intervals to obtain the target judgment interval. Based on the target judgment interval, the corresponding early warning level is matched to generate a graded early warning instruction for the terminal block.

2. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, After acquiring the raw current waveform data and raw temperature data of the terminals in the power metering device, the method further includes: The original current waveform data and the original temperature data are timestamp aligned according to the synchronization clock signal to obtain aligned current waveform data and aligned temperature data. The aligned current waveform data and the aligned temperature data are used as the original current waveform data and the original temperature data in the time-frequency domain analysis and processing step. If the time stamps of the aligned current waveform data and the aligned temperature data are not continuous, data interpolation or retransmission request processing is performed.

3. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, The step of performing time-frequency domain analysis on the original current waveform data and the original temperature data to obtain a waveform distortion sequence and a temperature rise sequence includes: Frequency domain harmonic extraction is performed on the original current waveform data to obtain a waveform distortion sequence; Perform a first-order forward difference operation on the original temperature data to obtain the temperature gradient vector; The temperature gradient vector is subjected to moving average filtering to obtain the temperature rise sequence.

4. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, The calculation of the time delay between the waveform distortion sequence and the temperature rise sequence includes: Perform cross-correlation operation on the waveform distortion sequence and the temperature rise sequence to obtain a cross-correlation function sequence; The offset corresponding to the peak value of the cross-correlation function sequence is extracted as the time delay.

5. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, The feature fusion of the signal hysteresis feature sequence to obtain the associated feature matrix includes: Extract the current amplitude sequence from the original current waveform data and construct a thermal balance mapping relationship based on the preset terminal thermal time constant; Substitute the temperature gradient and the current amplitude sequence in the signal hysteresis feature sequence into the thermal balance mapping relationship, and perform inversion calculation to obtain the change in contact resistance. The change in contact resistance is added as a new feature dimension and concatenated into the signal hysteresis feature sequence to obtain an association feature matrix reflecting the change in contact resistance.

6. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, The step of assessing the degradation status based on the correlation feature matrix and calculating the degradation status assessment value of the terminal block includes: Feature extraction is performed on the associated feature matrix to obtain a multidimensional state feature vector containing a temperature gradient distribution sub-vector and an electrical feature distribution sub-vector; The degradation status assessment value of the terminal is determined by weighting and summing the temperature gradient distribution sub-vector and the electrical characteristic distribution sub-vector in the multidimensional state feature vector, combined with the corresponding preset weight coefficients.

7. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, The step of extracting potential hazard features from the wiring terminals to obtain a potential hazard feature vector includes: Extract multiple degradation status evaluation values ​​of the terminal block arranged in chronological order within a historical preset statistical time period to form an evaluation value time series; Linear regression calculation is performed on the time series of the evaluation values ​​to obtain the slope of the change in the time series of the evaluation values ​​as the rate of degradation evolution; The local overheating frequency is obtained by counting the number of times that the sampling points of the terminal block exceed the preset overheating threshold in the original temperature data within the preset statistical time period. A hazard feature vector is constructed based on the degradation evolution rate and the frequency of local overheating.

8. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, The step of determining the stage division boundary based on the hidden danger feature vector includes: Obtain the hazard feature vectors and the deterioration state evaluation values ​​corresponding to each hazard feature vector at multiple sampling times within a preset historical period, and construct a two-dimensional feature sample set; Using the degradation state assessment value as the horizontal axis and the feature values ​​of each dimension of the hidden danger feature vector as the vertical axis, cluster analysis is performed on the two-dimensional feature sample set to obtain multiple clusters; Extract the projection interval of each cluster on the horizontal axis, and use the boundary value between the projection intervals of adjacent clusters as the stage division boundary between each hidden danger stage.

9. The real-time monitoring and fault prediction method for an electricity metering box according to claim 1, characterized in that, The step of generating a tiered warning instruction for the terminal block by matching the corresponding warning level to the target determination interval includes: The target determination range is matched with the pre-defined standard hazard level to obtain the corresponding warning level; Read the terminal identification code of the terminal block, and generate an initial warning command after encoding. The initial warning instruction is pushed into the instruction issuance queue, and the queue priority of the initial warning instruction in the instruction issuance queue is determined according to the warning level. After scheduling is performed according to the queue priority, a graded early warning instruction for the terminal block is output from the instruction issuance queue.

10. A real-time monitoring and fault prediction system for an electricity metering box, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method according to any one of claims 1 to 9.

Images