Whole-set ring main unit electrical variable abnormal energy spectrum measurement method

Abstract

The invention relates to the technical field of electrical variable monitoring, and further relates to a complete ring main unit electrical variable abnormal energy spectrum measurement method. The method comprises the following steps: step 1, constructing a substrate energy spectrum in a normal operation state; step 2, performing forward mode residual encoding based on the base energy spectrum to obtain a forward mode residual so as to highlight sudden increase abnormity of high-order harmonic energy; carrying out reverse mode residual encoding based on the base energy spectrum to form a reverse mode residual so as to capture slow change or drift abnormity of low-order harmonic energy; 3, performing bidirectional residual energy spectrum fusion to obtain an abnormal gain spectrum; step 4, constructing a self-adaptive judgment threshold value; and outputting the abnormal energy spectrum comprehensive index of the electrical variable of the complete ring main unit for alarm triggering. According to the invention, the sensitivity and anti-interference performance of anomaly detection are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electrical variable monitoring, and particularly relates to a method for measuring the energy spectrum of abnormal electrical variables of a complete set of ring network boxes. Background Technique

[0002] In the distribution network and industrial power consumption systems, as an important node device for branch power distribution, the ring network box undertakes multiple tasks such as bus sectioning, load transfer, and insulation protection. Due to its complex internal structure and variable operating environment, the electrical operating state of the ring network box is often affected by multiple factors, making it difficult to identify electrical abnormalities in a timely manner, and even potentially causing large-scale power outages in a short period. Therefore, the status monitoring of ring network boxes has always been an important research direction in the field of power operation and maintenance. Existing technologies usually rely on the monitoring of the fundamental effective value of voltage and current waveforms, harmonic analysis, or partial discharge pulse counting methods, hoping to reflect the health level of equipment operation through these indicators. However, traditional fundamental effective value monitoring can only cover the steady-state load characteristics of the equipment and cannot detect weak abnormalities such as the release of high-order harmonic energy caused by local contact looseness, insulation aging, or cable shielding defects. On the other hand, most conventional harmonic analyses are limited to the direct calculation based on Fourier transform or harmonic amplitude spectrum. Although it can reveal the increase in certain frequency components, it cannot further distinguish the transient impacts, long-term offsets, or coupling relationships caused by physical channel transmission behind these frequency components, making it difficult to effectively analyze the coupling abnormalities between the spike mutations in high-order harmonics and the chronic drifts in low-order harmonics. Although the partial discharge pulse counting technology can identify partial discharge events, it usually requires special high-frequency sensors to be installed inside or near the electrical equipment shell, which is significantly affected by the electromagnetic interference in the field environment and can only reflect high-frequency local abnormalities, and still cannot provide a complete monitoring perspective for the energy imbalance in the entire frequency band.

[0003] Existing methods based on energy spectrum analysis mostly adopt the method of superimposing the amplitudes of each harmonic component, which cannot incorporate the physical transfer relationship of the progressive harmonic energy along the order into the analysis system, and also ignore the significant impact of the relative intensity and distribution between harmonic orders on anomaly detection. This leads to problems such as low signal-to-noise ratio and difficulty in adapting the alarm threshold in the monitoring system when encountering complex situations where anomalies occur simultaneously at multiple orders. Some existing improvement methods attempt to introduce the idea of modal residuals, that is, by comparing the short-term increase and decrease trends or low-frequency drift trends of the energy spectra of each order of harmonics, to judge the abnormal mutation of harmonic energy at a certain order. However, in the specific implementation, relevant technologies often simply take the difference of harmonic amplitudes or perform simple moving averaging, ignoring the law of natural attenuation or amplification of the harmonic energy coupling intensity inside metal-enclosed equipment such as ring main units with frequency and order. This physical natural law is reflected in the frequency response curves of inductance, capacitance, and parasitic magnetic conductance. Conventional methods fail to convert this physical basic information into amplitude normalization or dynamic gain adjustment in residual measurement. Therefore, the calculation results of residuals often depend on artificially selected coefficients or empirical weighting values, lacking clear physical explanations and repeatability. Especially in complex working conditions where the voltage and current amplitudes are similar and partial discharge and environmental magnetic field noise are superimposed, such methods are more likely to have drawbacks such as fuzzy judgment, false alarm, or frequent misreporting. Summary of the Invention

[0004] The main object of the present invention is to provide a method for measuring the abnormal energy spectrum of electrical variables of a complete set of ring main units, which fully combines four types of electrical variable signals: the effective value of phase voltage, the effective value of phase current, the single-pulse energy of partial discharge, and the environmental magnetic flux density. By using base energy spectrum extraction, forward residual prediction, and backward residual drift monitoring, and with the help of complementary exponential amplification bounded by resonance orders, multi-scale coupling fusion of high-order harmonic spike anomalies and low-order harmonic slow-varying anomalies is achieved, thereby adaptively identifying energy imbalance problems within the entire frequency band and improving the sensitivity and anti-interference ability of anomaly detection.

[0005] To solve the above problems, the technical solution of the present invention is realized as follows: A method for measuring the abnormal energy spectrum of electrical variables of a complete set of ring main units, the method comprising: Step 1: Within a preset analysis window, perform unit conversion on the electrical variables of each harmonic order of the ring main unit, and calculate the average of the converted energies at all sampling points according to the same harmonic order to construct a base energy spectrum under normal operating conditions, which is used to characterize the reference energy distribution of each harmonic order; Step 2: Based on the base energy spectrum, perform forward modal residual coding to obtain a forward modal residual to highlight the sudden increase anomaly of high-order harmonic energy; based on the base energy spectrum, perform backward modal residual coding to form a backward modal residual to capture the slow change or drift anomaly of low-order harmonic energy; Step 3: According to the natural resonance order of the ring network cabinet bus, perform two-way residual energy spectrum fusion to achieve unified enhancement of multi-scale abnormal features and obtain an abnormal gain spectrum. Step 4: Construct an adaptive decision threshold; when the abnormal gain spectrum of a certain harmonic order exceeds the adaptive decision threshold, it is recorded as an effective abnormality; accumulate the abnormal gain spectra of all harmonic orders exponentially according to the proportion of the harmonic order numbers and take the average to output the abnormal energy spectrum comprehensive index of the complete set of ring network cabinet electrical variables for alarm triggering.

[0006] Furthermore, the electrical variables include: the effective value of phase voltage, the effective value of phase current, the single pulse energy of partial discharge, and the environmental magnetic flux density.

[0007] Furthermore, in Step 2, the process of performing forward mode residual coding specifically includes: for any target harmonic order, select a limited number of base energy spectra of low harmonic orders as the recursive set; distribute weights according to the relative magnitudes of the active powers corresponding to each harmonic order in the recursive set to form a theoretical prediction energy spectrum; subtract the theoretical prediction energy spectrum from the actual base energy spectrum of the target harmonic order to obtain a forward mode residual, which is used to highlight the sudden increase abnormality of high-order harmonic energy.

[0008] Furthermore, the low harmonic order is less than the target harmonic order; the value of the limited number is: , where is the natural resonance order of the ring network cabinet bus; if , then the selected low harmonic orders are only from 1 to ; otherwise, select the nearest harmonic orders as the low harmonic orders; is the target harmonic order.

[0009] Furthermore, in Step 2, the process of performing reverse mode residual coding includes: first, subtract the base energy spectra of the adjacent and higher-order harmonic order from the current harmonic order to obtain an immediate differential deviation; then, set a harmonic order window with a width equal to the inherent harmonic span of a single ring network cabinet in the low-order direction from the current harmonic order, and average all the base energy spectra within the harmonic order window; combine the immediate differential deviation and the window average deviation to form a reverse mode residual, which is used to sensitively capture the slow change or drift abnormality of low-order harmonic energy.

[0010] Furthermore, the process of combining the immediate differential deviation and the window average deviation to form a reverse mode residual includes: divide the immediate differential deviation and the window average deviation by their respective means of the target harmonic order under historical normal conditions. Determine whether the signs of the instantaneous differential deviation and the window average deviation are the same, and perform co-directional verification, including: if the signs are the same, retain their respective absolute values and add them to enhance the detection sensitivity of co-directional anomalies; if the signs are opposite, take the difference between the absolute values of the two and retain the actual sign of the result to avoid positive and negative cancellation masking the trend anomaly; define the result after co-directional verification as the reverse mode residual of the target harmonic order.

[0011] Further, step 3 includes: respectively performing complementary exponential amplification on the forward mode residual and the reverse mode residual according to the natural resonance order of the ring main unit bus, so that the high-order residuals are amplified incrementally according to the harmonic order ratio and the low-order residuals are amplified decrementally; then, taking the base energy spectrum of the harmonic order as the normalization reference, multiplying the forward residual and the reverse residual after performing complementary exponential amplification at the corresponding harmonic order to obtain the abnormal gain spectrum after bidirectional residual fusion, thereby realizing the unified enhancement of multi-scale abnormal features.

[0012] Further, the process of respectively performing complementary exponential amplification on the forward mode residual and the reverse mode residual includes: traversing the base energy spectrum to obtain the global maximum point of the base energy spectrum; confirming the harmonic order corresponding to this maximum point as the natural resonance order of the ring main unit bus; taking the natural resonance order as the boundary, dividing all harmonic orders into a high-order interval and a low-order interval; the high-order interval refers to all harmonic orders greater than the natural resonance order, and the low-order interval refers to all harmonic orders less than or equal to the natural resonance order; for each harmonic order in the high-order interval, construct a monotonically increasing exponential weight according to the ratio of its harmonic order to the natural resonance order to amplify the amplitude of the forward mode residual; for each harmonic order in the low-order interval, construct a monotonically decreasing exponential weight according to the ratio of the natural resonance order to its harmonic order to amplify the amplitude of the reverse mode residual; after taking the absolute values of the forward mode residual and the reverse mode residual respectively, introduce the corresponding exponential weights; taking the exponential weights as the exponents, perform power operations on the absolute values of the amplified forward mode residual and reverse mode residual, and at the same time retain the original positive and negative signs to obtain the forward residual and the reverse residual after complementary exponential amplification.

[0013] Further, the construction process of the adaptive decision threshold includes: for each harmonic order, call the mean and standard deviation of the abnormal gain spectrum under historical normal conditions to construct an adaptive decision threshold of three times the standard deviation.

[0014] A complete set of ring main unit electrical variable abnormal energy spectrum measurement method of the present invention has the following beneficial effects: It can achieve high-precision detection and quantification of abnormal electrical variables in a ring main unit under the background of multi-physical field signals. By introducing four multi-modal electrical variables, namely voltage, current, single-pulse energy of partial discharge, and ambient magnetic flux density, and combining bidirectional residual fusion based on natural resonance order division, the invention overcomes the limitation of traditional methods relying on single-band and single-modal signals, and realizes the simultaneous detection of high-order harmonic spike anomalies and low-order slow-varying drift anomalies.

[0015] In addition, the complementary exponential amplification strategy directly converts the physical attenuation characteristics in the frequency domain into dynamic gain adjustment, enabling the abnormal information in the high-order interval and the low-order interval to be amplified orderly according to their physical importance, and avoiding the interference introduced by artificial weighting or empirical coefficients. This dynamic normalization and abnormal amplification mechanism based on electromagnetic coupling and energy spectrum transfer enables the system to adaptively adjust the alarm sensitivity under various complex operating environments, maximally avoid false alarms or missed alarms, provide reliable, continuous, and interpretable health monitoring and early warning capabilities for the operation of the distribution network, and at the same time provide solid data support for operation and maintenance decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow chart of a method for measuring the energy spectrum of abnormal electrical variables in a complete set of ring main units provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the fusion principle of an abnormal gain spectrum integrating bidirectional modal residuals provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the principle of bidirectional modal residual coding integrating bidirectional modal residuals provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Refer to Figure 1 : A method for measuring the energy spectrum of abnormal electrical variables in a complete set of ring main units, the method comprising: Step 1: Within a preset analysis window, perform unit conversion on the electrical variables of each harmonic order of the ring main unit, and calculate the average according to the converted energy at all sampling points of the same harmonic order to construct a base energy spectrum under normal operating conditions, which is used to characterize the reference energy distribution of each harmonic order; Step 2: Based on the base energy spectrum, perform forward modal residual encoding to obtain the forward modal residual, so as to highlight the sudden increase anomaly of the high-order harmonic energy; based on the base energy spectrum, perform reverse modal residual encoding to form the reverse modal residual, so as to capture the slow change or drift anomaly of the low-order harmonic energy; Step 3: According to the natural resonance order of the ring main unit bus, perform two-way residual energy spectrum fusion to achieve unified enhancement of multi-scale anomaly features and obtain the anomaly gain spectrum; Step 4: Construct an adaptive decision threshold; when the anomaly gain spectrum of a certain harmonic order exceeds this adaptive decision threshold, it is recorded as a valid anomaly; exponentially accumulate the anomaly gain spectra of all harmonic orders according to the proportion of the harmonic order numbers, and take the average to output the anomaly energy spectrum comprehensive index of the complete set of ring main unit electrical variables for alarm triggering.

[0019] This embodiment provides a method for measuring the anomaly energy spectrum of ring main unit electrical variables by fusing two-way modal residual encoding. In the method for measuring the anomaly energy spectrum of ring main unit electrical variables in this embodiment, the overall principle can be regarded as a time-frequency mapping and residual enhancement process of the multi-field coupling energy evolution law inside the ring main unit. As a metal-enclosed electrical equipment at the end of the distribution network, the ring main unit has many branches, complex busbar routes, and intertwined magnetic and electric circuits, forming a series of energy residence channels with discrete harmonic order characteristics under power frequency excitation. When there is poor contact, insulation aging, load distortion or partial discharge in any branch, these energy channels will show observable power release or magnetic flux release at specific harmonic orders. To continuously monitor such weak and scattered anomaly signals, this method first constructs a preset analysis window, and uses synchronous sampling to collect multi-modal electrical variables such as voltage, current, partial discharge pulses and ambient electromagnetic noise in the same physical time period, and then adopts energy conservation conversion and dimensionless processing of physical fields to uniformly map the instantaneous energy of each harmonic order to the same scale; after eliminating the transient fluctuations through sample averaging, the base energy spectrum is formed. The base energy spectrum reflects the steady-state energy proportion of each harmonic order during normal operation, and it is physically equivalent to the static energy distribution baseline of the multi-field coupling inside the ring main unit. Subsequently, the method introduces forward modal residual encoding, regards the current high-order harmonic order as the end point, and regards the base energy spectra of several low harmonic orders as the energy transfer sources; according to the resonance characteristics of the bottom-up energy transfer in the electrical system and the electromagnetic coupling between harmonics, uses the relative weight of the low-order base energy to predict the steady value of the high-order energy under ideal coupling conditions, and then compares it with the actual observed value to generate the forward modal residual.

[0020] This residual essentially measures whether the high-order energy exceeds the reasonable upper limit that can be supplied by the low-order coupling. Therefore, it is extremely sensitive to harmonic spikes or high-frequency surges. Complementary to this, the reverse-mode residual encoding targets the change in energy stability of the low-order harmonics. By comparing the energy difference between the target order and the adjacent higher order, and superimposing the average offset within a complete harmonic span window in the lower-order direction, it comprehensively evaluates the gradual drift of energy in the low-frequency band. Since the magnetic conductive metal parts, insulating media, and line distribution parameters inside the ring network cabinet change slowly with temperature, load, and structural stress, the low-order energy usually reflects such slow-changing operating conditions. Therefore, the reverse-mode residual can capture the spectral shift caused by long-term degradation or chronic mismatch. The forward-mode residual and the reverse-mode residual correspond to high-speed anomalies and low-speed anomalies respectively, but their magnitudes and sign values are different. If directly superimposed, the high-frequency anomalies will be offset or masked by the low-frequency drift. Therefore, this method sets up a two-way residual energy spectrum fusion mechanism: First, determine the natural resonance order in the base energy spectrum. The resonance order corresponds to the dominant resonance frequency point jointly affected by the ring network cabinet structure parameters and external load. The frequency band higher than the resonance order is called the high-order interval, and the frequency band lower than or equal to the resonance order is called the low-order interval. Then, apply an exponentially increasing amplification to the forward-mode residual with respect to the order ratio, so that the amplification factor increases exponentially for each increase in order; apply an exponentially decreasing amplification to the reverse-mode residual with respect to the order ratio, so that the lower the order, the greater the amplification factor, but not exceeding the reference amplitude at the resonance order.

[0021] Before amplification, take the absolute value of the residual to unify the dimension, and then restore it according to the original sign to retain the direction information of energy weakening or energy strengthening. Through this complementary exponential amplification, the spikes of the forward residual are more likely to rise in the high-order interval, while the subtle changes of the reverse residual are highlighted in the low-order interval. The two are multiplied point by point according to the order in the energy spectrum space to obtain the abnormal gain spectrum. The abnormal gain spectrum converts the high-order spikes and low-order drifts into multiplicative gains in the same frequency domain, thus highlighting multi-scale anomalies synchronously. Since a single frequency point may still be affected by random noise, transient shocks, or measurement errors, this method continues to perform statistical filtering on the abnormal gain spectrum, accumulates the mean and standard deviation of the gains at each order under long-term normal operating conditions, and constructs an adaptive decision threshold using the three-standard-deviation principle. The selection of three standard deviations stems from the 99.7 percentile theory in the empirical distribution, which is sufficient to balance the false alarm rate and the miss rate in continuous observation samples. When the abnormal gain spectrum of any harmonic order crosses the threshold, it is recognized as a valid anomaly. This recognition not only indicates that the energy at this frequency point breaks through the statistical stability band but also means that the ring network cabinet has an over-limit coupling in the corresponding frequency domain channel. To measure the overall health level, the method further exponentiates each order's abnormal gain spectrum according to the proportion of the harmonic order number. The exponential weight makes the high-order surges and low-order drifts maintain an importance matching their respective electromagnetic hazard risks in the overall evaluation. Subsequently, the weighted gain spectrum is averaged to obtain the comprehensive index of the abnormal energy spectrum.

[0022] The comprehensive index of abnormal energy spectrum has the same isomorphic scale as the base energy spectrum and can be directly mapped to the percentage of energy balance deviation of the ring network cabinet. The larger this index is, the more serious the deviation of the full-spectrum energy is, the more abnormal the coupling relationship of each harmonic channel is, and the lower the overall safety margin of the power distribution equipment is. By calculating the comprehensive index of abnormal energy spectrum in real time and comparing it with the set threshold, the alarm device can be driven to prompt the maintenance personnel in the first time. At the same time, the index time series can be input into the intelligent operation and maintenance system to analyze the deterioration rate, predict the fault lead time and plan the maintenance cycle. To sum up, this method establishes a steady-state energy baseline through the base energy spectrum, uses forward modal residual coding and backward modal residual coding to correspond to high-frequency sudden increase and low-frequency drift anomalies, implements complementary index amplification and fusion in the high and low intervals divided by the resonance order, maps the abnormal information of different time scales and different frequency spectrum scales to the abnormal gain spectrum, then screens out effective anomalies through the adaptive decision threshold, and finally quantifies the overall health state of the equipment with the comprehensive index of abnormal energy spectrum, which can significantly reduce false alarms while ensuring sensitivity and meet the comprehensive requirements of real-time, credibility and interpretability for on-line monitoring of ring network cabinets.

[0023] Furthermore, the electrical variables include: the effective value of phase voltage, the effective value of phase current, the single-pulse energy of partial discharge, and the ambient magnetic flux density.

[0024] The effective value of phase voltage and the effective value of phase current respectively represent the instantaneous energy supply and load distribution of the circuit transmission channel. The single-pulse energy of partial discharge directly corresponds to the transient migration of charges in the insulating medium, and the ambient magnetic flux density reflects the leakage distribution characteristics of the electromagnetic field during the spatial conduction process. The base energy spectrum is: ; where is the harmonic order; is the time sampling serial number; is the effective value of the th-order phase voltage; is the rated line voltage effective value of the ring network cabinet; is the effective value of the th-order phase current; is the rated phase current effective value of the ring network cabinet; is the th-order RLC equivalent impedance; is the reference impedance; is the th-order single-pulse energy of partial discharge; is the upper limit of the allowable partial discharge energy specified by the manufacturer; is the th-order ambient magnetic flux density; is the reference magnetic flux density under no-load condition; is the number of sampling points within the analysis window; is the maximum harmonic order; is the order base energy spectrum.

[0025] Further, in step 2, the process of performing forward modal residual encoding specifically includes: for any target harmonic order, select a limited number of base energy spectra of low harmonic orders as the recurrence set; distribute weights according to the relative magnitudes of the active powers corresponding to each harmonic order in the recurrence set to form a theoretical predicted energy spectrum; subtract the theoretical predicted energy spectrum from the actual base energy spectrum of the target harmonic order to obtain the forward modal residual, which is used to highlight the sudden increase anomaly of high-order harmonic energy.

[0026] When the ring network cabinet is in steady-state operation, the energy in the base energy spectrum shows a decreasing distribution from low to high among different harmonic orders. This decrease not only reflects the natural suppression of high-frequency components by conductor resistance, inductance, and distributed capacitance, but also reflects the step-by-step absorption of active power by the load end. From the perspective of electromagnetic fields, low harmonic orders have stronger resident magnetic energy and electrical energy. They excite high-order components upward through the coupling path formed by copper bars, switch contacts, and cable shielding layers, and the high-order components dissipate energy near the core edge, partial discharge voids, and frame lap joints, and finally dissipate in the form of heat, sound, or corona. Therefore, once there is a significant gap between the high-order energy inferred from the input power observable at low harmonic orders and the actually measured high-order energy, it means that there are additional injection sources or additional blocking sources in the energy transfer chain, and the corresponding physical anomalies often manifest as rapid evolution problems such as loose current-carrying connections, carbonization of insulation coatings, enhanced eddy currents in metal parts, or discontinuous grounding.

[0027] To quantify this gap, forward modal residual encoding first needs to find a set of base energy spectra of low harmonic orders for each target harmonic order, and this set of base energy spectra is called the recurrence set. The order depth of the recurrence set is determined by half of the natural resonance order of the ring network cabinet because the frequency spectrum section below the resonance order carries most of the magnetic potential and electric potential that can be transferred upward, and further expansion downward will introduce far-low-frequency components with a transfer coefficient approaching zero to the target harmonic order, interfering with the prediction. After selecting the recurrence set, the system needs to know the current actual contribution ratio of energy of these low harmonic orders. For this purpose, active power is introduced as the distribution weight. Active power involves the effective value of phase voltage, the effective value of phase current, and the phase angle at the same time. The instantaneous power integration completed within the power frequency cycle can eliminate the interference of reactive power and harmonic distortion on energy evaluation, so it becomes a reliable indicator for measuring the load of the real energy channel. Normalizing the active powers of each harmonic order in the recurrence set can obtain a cluster of weight coefficients with a sum of one. After multiplying them by the amplitudes of their respective base energy spectra and summing them up, the theoretical predicted energy spectrum is generated.

[0028] The theoretically predicted energy spectrum physically represents the energy upper limit of the target harmonic order under the current load and coupling state, and can also be regarded as the expected value of the ideal coupling scenario. Then, the theoretically predicted energy spectrum is directly subtracted from the actual base energy spectrum of the target harmonic order to obtain the forward modal residual. If the forward modal residual is positive and significantly greater than the historical variance, it indicates that there is an abnormal energy gain in the target harmonic order. This gain cannot be derived from the normal coupling of low-order energy, and often indicates that the occurrence of electric arcs or local discharges on the metal surface suddenly increases. When the single-pulse energy of local discharges significantly accumulates in the high-frequency channel, it will form a sharp harmonic peak. If the forward modal residual is negative and the amplitude is too large, it indicates that the normal transport of energy to the target harmonic order is inhibited. This may be due to the increase in resistance caused by the temperature rise of the branch conductor, the appearance of an oxide film at the contact point, or the current shunting caused by the damage of the shielding layer, thus weakening the excitation amplitude of the high-order harmonics. To ensure the dynamic adaptability of the forward modal residual encoding, the method updates the data window of the recursive set in each sampling period, so that the base energy spectrum of the low harmonic order and the corresponding active power are refreshed in real time with the change of the load. At the same time, a sliding average is used to slightly smooth the theoretically predicted energy spectrum to suppress the error amplification caused by short-term fluctuations.

[0029] Since the forward modal residual is essentially a measure of the difference between theory and reality, as long as there is an unbalanced injection or blockage in the electromagnetic coupling chain of the ring network cabinet, the residual will immediately appear and show an exponential growth trend as the fault expands. In the subsequent two-way residual energy spectrum fusion, the absolute value of the forward modal residual will be exponentially amplified according to the ratio of the harmonic order to the resonance order, further enhancing the significance of the spike anomaly in the high-order interval, so as to ensure that the overall measurement method can lock and alarm the sudden increase of high-order harmonic energy with a time resolution of milliseconds, and realize the early diagnosis and trend assessment of the fast failure mode of the ring network cabinet.

[0030] The forward modal residual is: ; where is the forward modal residual of the th order; is the recursion depth, defined as ; is the active power of the th order; is the voltage-current phase angle of the th order at the th sampling point; and are both integer subscript indexes.

[0031] Furthermore, the low harmonic order is less than the target harmonic order; the value of the limited quantity is: where is the natural resonance order of the ring network cabinet bus; if , the selected low harmonic orders are only from 1 to ; otherwise, select the nearest harmonic orders as the low harmonic orders; is the target harmonic order.

[0032] In order to ensure that the forward modal residual coding can not only make full use of the energy transfer information of the low harmonic orders, but also not introduce far-low frequency noise with a coupling coefficient approaching zero with the target harmonic order due to too deep recurrence order, a clear order depth limit rule is set for the recurrence set. The core idea of this rule is to use the natural resonance order of the ring network cabinet bus as the physical upper limit of the energy coupling channel, and then reduce this upper limit to a safe range according to the binary principle.

[0033] The specific method is as follows: During operation, first find the global peak position of the energy amplitude by using the base energy spectrum scanning, and identify the harmonic order where the peak is located as the natural resonance order; then divide the natural resonance order by two and take the integer part to obtain an order depth upper limit, which is called the maximum recurrence depth. The maximum recurrence depth can be regarded as the farthest distance at which high-order energy can still maintain a significant coupling contribution when tracing back to the low-order channel. Physically, it corresponds to the position where the combined inductance of the busbar and cable and the distributed capacitance attenuate the high-frequency components to the critical threshold. When the system starts to perform forward recurrence prediction on a certain target harmonic order, it will first judge the relative relationship between the target order and the maximum recurrence depth. If the value after subtracting one from the target order is less than the maximum recurrence depth, it means that the target order itself is in the low-frequency region below half of the resonance order. In this case, it is not necessary to trace back too many orders to cover the main energy supply source, so the recurrence set directly takes all the low harmonic orders from the fundamental order to the order before the target order; conversely, if the target order is high enough so that the value after subtracting one from it is not less than the maximum recurrence depth, the system only selects several low harmonic orders closest to the target order, and the number is equal to the maximum recurrence depth. In this way, the recurrence set can form a dense energy transfer gradient near the target order, and will not include far-low-frequency components that lack direct coupling with the target order into the weight calculation due to an overly wide window, thus effectively suppressing noise amplification and error propagation. Next, the system reads the active power values of each low harmonic order in the recurrence set within the current analysis window, normalizes these active powers to form a set of relative weight coefficients. The weight coefficients physically reflect the contribution ratio of different low harmonic orders to the excitation of high-order energy, and also reflect the actual energy supply capacity of each channel under the current load state. The system then multiplies these weight coefficients by the base energy spectrum amplitudes of the corresponding low harmonic orders one by one and sums them up to obtain the theoretical predicted energy spectrum. This predicted value is equivalent to the steady-state amplitude that the low-order energy can maintain at the target order after cascading upward along the electromagnetic coupling path under ideal conditions. Essentially, it is a dynamic upper limit based on instantaneous load, structural coupling, and energy conservation constraints. Subsequently, the system compares the theoretical predicted energy spectrum with the base energy spectrum actually measured at the target harmonic order, and the difference between the two is the forward mode residual.

[0034] Since the order depth of the recursive set follows the principle of the half value of the natural resonance order for adaptive adjustment within each sampling period, the forward modal residual can always evaluate the sudden increase in energy on the most reasonable physical scale. When immediate faults such as mass partial discharges, arc bridging, or loose current-carrying connections occur inside the ring network cabinet, the high-order energy will break through the upper limit predictable by low-order coupling within an extremely short time, and the forward modal residual will rise significantly and trigger subsequent exponential amplification and fusion. In the case of normal load swings or harmless transients, the real-time supply changes of low-order energy will also be synchronously reflected in the weight coefficients, thus keeping the forward modal residual stable and avoiding false alarms. Through this self-adaptive and physically well-constrained low harmonic order selection mechanism, while ensuring the high-sensitivity capture ability of high-order anomalies, the method effectively reduces the interference caused by random noise, load fluctuations, and environmental magnetic field disturbances, ensuring that the encoded output of the forward modal residual is both real-time and interpretable, providing solid theoretical support and engineering feasibility for the full-spectrum anomaly detection of electrical variables in the ring network cabinet.

[0035] Further, in step 2, the process of performing reverse modal residual encoding includes: first, taking the difference between the base energy spectra of adjacent and one higher harmonic order and the current harmonic order to obtain the immediate differential deviation; then, setting a harmonic order window with a width equal to the inherent harmonic span of a single ring network cabinet in the low-order direction from the current harmonic order, and averaging all the base energy spectra within the harmonic order window; combining the immediate differential deviation with the window average deviation to form the reverse modal residual for sensitively capturing the slow change or drift anomaly of low-order harmonic energy.

[0036] The reverse modal residual coding undertakes the task of monitoring the slow change or abnormal drift of low-order harmonic energy. Its implementation process fully unfolds around the inherent characteristic that the energy in the ring main unit flows back step by step from high to low under electromagnetic steady state. First, after the system completes a new round of base energy spectrum refresh in each sampling period, it will calculate the instantaneous differential deviation for all harmonic orders in sequence. The specific method is to directly subtract the base energy spectrum value of the current harmonic order from the adjacent and higher-order base energy spectrum value, and the obtained difference is the instantaneous differential deviation. This difference represents the instantaneous change direction and amplitude of the energy distribution slope between two adjacent frequency channels. When the ring main unit is in a healthy state, due to continuous structural impedance and uniform load distribution, the energy decay rate between the low-order channel and the immediately adjacent higher-order channel will remain stable; once factors such as load reconstruction, contact aging, or distribution capacitance change cause the magnetic field reflux to be blocked, the absolute value of the instantaneous differential deviation will produce a slow and continuous drift, becoming the first front-line indicator for detecting low-frequency anomalies. To eliminate single-point random fluctuations and capture the true slow change trend, the system will then expand a harmonic order window with a width equal to the inherent harmonic span of a single ring main unit in the low-order direction of the current harmonic order. The inherent harmonic span is a structural characteristic quantity obtained through resonance scanning during the factory commissioning stage, which usually reflects the harmonic interval determined by the comprehensive parameters of the busbar section length, bushing path, and shielding circuit. It is the smallest complete unit of electromagnetic coupling inside the ring main unit. Taking this span as the window width not only conforms to the physical coupling granularity but also avoids the risk of masking local information with an overly wide window.

[0037] Arithmetic averaging is performed on all base energy spectrum values within this harmonic order window to obtain the window average deviation. This average value depicts the overall drift direction of the energy baseline within a certain number of lower orders from the current order. If the long-term aging of the ring main unit causes the overall energy of the low-order channel to gradually accumulate or decay, the window average deviation will show a stable positive or negative offset, corroborating the trend of the instantaneous differential deviation. Next, the system needs to combine the instantaneous differential deviation and the window average deviation to form the reverse modal residual. The combination strategy follows the principle of enhancing in the same direction and suppressing in the opposite direction: when their signs are the same, it indicates that the local slope change and the overall baseline drift diverge or converge together, meaning that the low-order harmonic energy shows a unidirectional abnormal trend. At this time, the system adds the absolute values of the two and assigns the original sign to amplify the abnormal signal in the same direction; when their signs are opposite, it shows that the local mutation may be just a transient caused by random perturbation or load swing, and the overall drift does not respond synchronously. At this time, the system takes the difference between the absolute values of the two and retains the sign of the difference to avoid positive and negative cancellation masking the true slow change information. The combined result is the reverse modal residual, which not only retains the cumulative effect of low-order energy slip but also introduces the real-time feedback of the slopes of adjacent orders, achieving sensitive capture of low-frequency anomalies.

[0038] The reverse modal residuals are then fed into the bidirectional residual energy spectrum fusion step, where they are amplified exponentially in the low-order interval bounded by the resonant order and complement the forward modal residuals in the high-order interval, ultimately jointly constituting the abnormal gain spectrum. Since the reverse modal residuals mainly respond to slow processes such as insulation moisture diffusion, current-carrying conductor temperature rise, and gradual change of distributed capacitance, their change rate is much smaller than that of high-order spikes, but they have a more direct indicative significance for equipment life and thermal stability. Through the combined processing of the instantaneous differential deviation and the window average deviation, the method introduces a dual filtering mechanism in the low-order harmonic channel that can not only filter out random noise but also accumulate trend information, enabling the abnormal energy spectrum comprehensive index to not only quickly alarm for fast discharge events but also give quantitative hints in advance for chronic faults such as insulation deterioration and grounding imbalance, thereby ensuring that the energy balance deviation of the ring main unit is monitored in real time throughout its entire life cycle and providing continuous and reliable data support for operation and maintenance decision-making.

[0039] The reverse modal residuals are: ; where is the reverse modal residual of the th order; is the mean value of the base energy spectrum of the th order under long-term normal operating conditions; is the width of the backward-looking window, and its value is equal to the harmonic order span corresponding to a single ring main unit; is the base energy spectrum of the th order; is the harmonic order variable within the backward-looking window.

[0040] Furthermore, the process of forming the reverse modal residuals by combining the instantaneous differential deviation and the window average deviation includes: dividing the instantaneous differential deviation and the window average deviation by the mean values of their corresponding target harmonic orders under historical normal operating conditions respectively; judging whether the signs of the instantaneous differential deviation and the window average deviation are the same for co-directionality verification, including: if the signs are the same, retain their absolute values and add them to enhance the detection sensitivity of co-directional anomalies, if the signs are opposite, take the difference between their absolute values and retain the actual sign of the result to avoid positive and negative cancellation masking trend anomalies; define the result after co-directionality verification as the reverse modal residual of the target harmonic order.

[0041] The instantaneous differential deviation essentially characterizes the instantaneous change of the energy density gradient between adjacent harmonic channels. If this gradient remains stable for a long time, it means that the low-order harmonic energy continuously flows back to higher orders according to the normal coupling relationship. However, once the gradient suddenly increases or decreases, it often indicates that there is an abnormality in the local reactance or capacitance channel, causing the energy transmission to be blocked or extra energy to be injected. The window-averaged deviation, on the other hand, is responsible for measuring the overall drift of the energy baseline within a certain order bandwidth. It calculates the average deviation by extending the current harmonic order towards the lower order by an interval equal to the inherent harmonic span of the ring network box, thereby filtering out the random fluctuations of a single order and making the observation results more reflect the common trend of multi-order coupling. To enable the comparison of the magnitudes of the two deviations within the same statistical scale and eliminate the dimension difference, the method first adopts the normalization step of dividing each by the mean value of the target harmonic order under historical normal operating conditions. This operation is equivalent to mapping the energy spectrum amplitude to the unitized relative deviation space, so that subsequent comparisons only focus on the "deviation percentage" rather than the "absolute amplitude", thus avoiding the energy baseline difference between high and low load levels from masking the abnormality.

[0042] After normalization, the system performs a co-directionality check on the two deviations. The underlying physical logic is the law of energy balance and the principle of random vector superposition: if the signs of the instantaneous differential deviation and the window-averaged deviation are the same, it indicates that the local slope change and the overall energy baseline drift point in the same direction. Whether it is energy accumulation or energy decay, both are jointly driving the low-order harmonic energy to move in the same direction, which means that there are continuous and consistent structural changes in the current path, insulation medium, or magnetic flux leakage, showing obvious trend characteristics. To quickly amplify this trend in the subsequent abnormal gain spectrum, the system chooses to retain the absolute values of both and add them together, which is equivalent to superimposing the contributions of co-directional energy imbalance and introducing non-linear gain to increase the detection sensitivity.

[0043] Conversely, if the signs of the two are opposite, it indicates that there is a cancellation effect between the local slope change and the overall drift. It may be a short-term anomaly triggered by transient load fluctuations, measurement noise, or minor partial discharges, or it may be a latent fault that has just occurred at certain locations and has not caused a baseline shift over a larger range. To avoid misjudging such small-scale disturbances as trend anomalies, the system adopts a strategy of taking the absolute value difference between the two and retaining the actual sign: when the local mutation amplitude is less than the baseline drift, the difference will be significantly weakened and ultimately further suppressed in the anomaly gain spectrum; when the local mutation amplitude is greater than the baseline drift, the difference retains the remaining energy direction, indicating that the small-scale rapid deterioration has broken through the background noise and is worthy of further tracking. Through such an adaptive co-directionality check, the method not only ensures that large co-directional drifts are quickly amplified but also quantitatively describes the degree of cancellation when the signs are opposite, thus avoiding false alarms while maintaining high sensitivity. Finally, the value after the co-directionality check is defined as the reverse mode residual of the target harmonic order and enters the two-way residual energy spectrum fusion process. Since the reverse mode residual is already a normalized and sign-analyzed net drift index, it can provide a clear low-frequency anomaly weight when multiplied by the forward mode residual, enabling the anomaly gain spectrum to take into account both high-order spikes and low-order slow-varying fault characterization mechanisms. The final output anomaly energy spectrum comprehensive index can then respond hierarchically to both rapid discharges and chronic aging. This combination process essentially integrates the three ideas of energy gradient conservation, statistical normalization, and sign logic operation, making the health criterion of the ring main unit in the low-order harmonic domain retain physical interpretability while possessing the robustness and real-time performance required for engineering applications.

[0044] Further, step 3 includes: according to the natural resonance order of the ring main unit bus, respectively performing complementary exponential amplification on the forward mode residual and the reverse mode residual, so that the high-order residuals are amplified incrementally according to the harmonic order ratio while the low-order residuals are amplified decrementally; then, taking the base energy spectrum of the harmonic order as the normalization reference, multiplying the forward residual and the reverse residual after performing complementary exponential amplification at the corresponding harmonic orders to obtain the anomaly gain spectrum after two-way residual fusion, thereby realizing the unified enhancement of multi-scale anomaly features.

[0045] Two-way residual energy spectrum fusion utilizes the physical interface provided by the natural resonance order of the ring main unit bus to respectively assign complementary exponential weights to the forward mode residual and the reverse mode residual, so that the high-order residuals obtain an amplification strength that monotonically increases with the order ratio, while the low-order residuals obtain an amplification strength that monotonically decreases with the order ratio, thereby ensuring that the spike anomalies that break out rapidly in the high-frequency range and the drift anomalies that accumulate slowly in the low-frequency range can be simultaneously enhanced and expressed in the energy spectrum dimension.

[0046] Specifically, the system first retrieves the global peak of the energy amplitude in the base energy spectrum and defines the harmonic order where the peak is located as the natural resonance order. This order corresponds to the dominant resonance point generated by the coupling of the busbar, switch contact, and cable distribution parameters, and is the critical watershed for the energy transfer along the frequency axis within the ring network cabinet. For the high-order interval above the resonance order, the method applies an exponential weight with the ratio of the order to the resonance order as the base to the forward modal residual at each harmonic order. This means that the higher the order, the faster the exponential amplification factor grows, and the amplitude of the high-order spike anomaly can be multiplied in the energy spectrum. For the low-order interval below or equal to the resonance order, the method applies an exponential weight with the ratio of the resonance order to the order as the base to the reverse modal residual at each harmonic order, so that the lower the order, the greater the amplification factor, but it is still strictly constrained by the weight at the resonance order to prevent the low-frequency noise from being infinitely amplified. After completing the complementary exponential amplification, the system uses the base energy spectrum of each harmonic order as the normalization reference, aiming to eliminate the amplitude drift that may be introduced during the amplification process and ensure that the amplified forward residual and reverse residual are still on a physically comparable energy scale.

[0047] After normalization, the two types of residuals are multiplied point by point at the corresponding harmonic orders to obtain the abnormal gain spectrum after bidirectional residual fusion. During the multiplicative fusion process, the forward residual in the high-order interval, due to being exponentially enhanced, forces the energy contribution of the high-frequency spike anomaly into the product result, while the reverse residual in the low-order interval, due to the superposition of its decreasing weight and the long-term accumulation effect, continuously provides the weight of the low-frequency drift anomaly in the product. The product operation has a gating effect: only when both the high-frequency spike and low-frequency drift anomalies at the same harmonic order coexist, the abnormal gain spectrum will show a significant amplification; if one type of anomaly is insufficient, the product is suppressed by the smaller value of the other type of anomaly, thereby reducing the false alarm rate. This design is equivalent to implementing a multi-scale collaborative detector in the frequency domain in terms of signal processing, which not only maintains the millisecond-level response speed to high-frequency surges but also provides the long-term integration ability for low-frequency slow changes. The finally generated abnormal gain spectrum is an energy anomaly map spanning all harmonic orders, and its amplitude directly reflects the imbalance degree of the internal electromagnetic coupling chain of the ring network cabinet at different scales.

[0048] Further, the process of performing complementary exponential amplification on the forward modal residual and the reverse modal residual respectively includes: traversing the base energy spectrum to obtain the global maximum point of the base energy spectrum; confirming the harmonic order corresponding to this maximum point as the natural resonance order of the ring network box bus; dividing all harmonic orders into a high-order interval and a low-order interval with the natural resonance order as the boundary; the high-order interval refers to all harmonic orders greater than the natural resonance order, and the low-order interval refers to all harmonic orders less than or equal to the natural resonance order; for each harmonic order in the high-order interval, construct a monotonically increasing exponential weight according to the ratio of its harmonic order to the natural resonance order for amplifying the amplitude of the forward modal residual; for each harmonic order in the low-order interval, construct a monotonically decreasing exponential weight according to the ratio of the natural resonance order to its harmonic order for amplifying the amplitude of the reverse modal residual; after taking the absolute values of the forward modal residual and the reverse modal residual respectively, introduce the corresponding exponential weights; take the power operation on the absolute values of the amplified forward modal residual and reverse modal residual with the exponential weights as the power, while retaining the original positive and negative signs, to obtain the forward residual and the reverse residual after complementary exponential amplification.

[0049] The bus, switch contacts and cable shielding layer of the ring network box form a high-order transmission chain that contains both skin resistance and distributed inductance and capacitance. Any local impedance mismatch on the link will leave an uneven imprint in the energy spectrum with frequency as the coordinate. When the harmonic frequency increases, the conductor reactance dominates the energy climb, and the magnetic potential is more easily driven towards gaps and sharp geometric edges; when the harmonic frequency decreases, the return current caused by the distributed capacitance and leakage conductance path becomes significant, and the polarization relaxation and thermal breakdown accumulation in the insulating medium change the energy baseline. The natural resonance order is exactly the intersection point of these two trends: the structural wave resistance is the lowest, the local standing wave is the strongest, and the energy may either be drawn away along the high-frequency channel or deposited along the low-frequency channel. The global maximum point of the base energy spectrum can stably lock this intersection point in a statistical sense because the power density near the resonance point is most likely to be jointly enhanced by the load and the environment during normal operation.

[0050] Taking this resonance order as the frequency axis segmentation reference, the rapidly rising anomalies and the slowly subsiding anomalies can be projected onto the high-order interval and the low-order interval respectively. If a linear gain is still used to superimpose the two types of residuals at this time, the "short board effect" will surely occur: the high-order spikes are much larger than the low-order drifts in instantaneous amplitude, while the low-order drifts far exceed the spike duration in time integration, and the two cannot be simultaneously visualized under the same scale. The proposed complementary exponential amplification strategy is precisely to solve this problem of essential scale mismatch. Its core is to make the amplification weight of the high-order interval increase exponentially with the order ratio, so that the positive modal residuals obtain a non-linear jump when the frequency increases; at the same time, make the amplification weight of the low-order interval decrease exponentially with the ratio, so that the negative modal residuals remain amplified when the frequency decreases, avoiding being suppressed by the high-order large-amplitude spikes. The choice of the exponential function rather than the non-linear slope is not a mathematical stunt, but is homologous to the skin depth attenuation of the transmission line resonance, ferromagnetic eddy current loss, and the dielectric capacitance energy storage curve: these natural processes are approximately linear in the logarithmic coordinate system, and their increase and decrease rates change exponentially on the ordinary linear scale. Fixing this natural increase and decrease rate as the weight base, the amplitude of the amplified residuals will be isomorphic to the physical dissipation law, and the energy conservation relationship will not be destroyed due to artificial coefficients. As for the process of taking the absolute value of the residuals, then performing the power operation, and finally restoring the sign, although it seems to be a pure mathematical step, it is essentially constructing a mirror-symmetric non-linear mapping: the absolute value operation projects both energy gain and energy loss onto the positive number axis, and the power operation completes the amplitude expansion on the positive number axis; restoring the sign rewrites the energy direction information back to the spectrum, so that the subsequent product still retains the power flow direction. In this way, if the high-order spikes correspond to an abrupt increase in energy, the exponential amplification will push them to an extremely high amplitude; if the low-order drifts correspond to energy attenuation, the sign of the exponentially amplified value is negative, and the same-order points of the high-frequency positive residuals and the low-frequency negative residuals balance each other during the product operation. Only when there is an asymmetric energy injection and a dissipation imbalance simultaneously in the entire channel, will the abnormal gain spectrum be significantly lifted due to the coupling of the same signs.

[0051] In other words, complementary exponential amplification fuses alarm signals with two different time constants and different energy directions through non-linear mapping in the same frequency domain plane, constructing an effect similar to a coherent detector: it is highly sensitive to synchronization imbalance and automatically suppresses isolated spikes or isolated drifts. Looking deeper, the exponential base is determined by the ratio of harmonic orders, and the harmonic order structure is determined by electrical geometry, ultimately making the amplitude field of the gain spectrum correspond one-to-one with the actual three-dimensional physical geometry field. In engineering, when a local discharge cluster appears inside the ring main unit cabinet, the positive-mode residuals in the high-order interval will be pulled above the threshold by the exponential weight; if at the same time insulation dampness causes the reverse-mode residuals in the low-order interval to slowly shift, the energy peak synthesized by multiplying the two types of anomalies will be much higher than any single anomaly, triggering maintenance in advance. If there is only environmental noise jittering at the high-frequency end, since the low-order residuals do not shift significantly, the product cannot be synchronously amplified, and the false alarm probability drops significantly. In summary, the core principle of complementary exponential amplification is not simple amplification but energy scale matching based on resonant partitioning, non-linear isomorphic mapping, and sign-coupled gating. It places the high-order instantaneous spikes responsible for by the positive-mode residuals and the low-order slow drifts responsible for by the reverse-mode residuals in a unified exponential weight system, using the natural resonant order as the physical anchor point, making the abnormal gain spectrum a comprehensive health metric with both time and frequency dual resolutions, providing a highly self-consistent and interpretable theoretical framework for early fault detection and trend assessment of the ring main unit cabinet in a multi-field coupling environment.

[0052] Furthermore, the construction process of the adaptive decision threshold includes: for each harmonic order, call the mean and standard deviation of the abnormal gain spectrum under historical normal operating conditions to construct an adaptive decision threshold of three times the standard deviation.

[0053] Comprehensive index of abnormal energy spectrum is ; wherein, is the mean of the abnormal gain spectrum of the th order under historical normal operating conditions; is the standard deviation of the abnormal gain spectrum of the th order under historical normal operating conditions; is a conditional indicator function, taking a value of 1 if the expression inside the condition holds, otherwise taking 0; is the abnormal gain spectrum after fusing the two-way residuals of the th order.

[0054] The following is an embodiment of the present invention in practice. The object is a ring main unit cabinet with a rated voltage of , a rated current of , a bus inductance of , a bus capacitance of , and a copper bar AC resistance of . Industrial frequency 。The analysis window length is synchronous sampling points, and the maximum harmonic order is taken as 。The partial discharge reference energy ,and the no-load reference magnetic flux density 。

[0055] Based on the formula where , 。

[0056] Typical sampling values (units normalized according to the formula): : , ; : , ; : , ; : , ; : , ; After substitution, we can get ,  ,  ,  ,  。Excluding the fundamental wave, the maximum amplitude appears at ,so the natural resonance order 。

[0057] The maximum recursion depth 。The recursion set is 。The active power of each order , , 。The normalized weight 。The theoretically predicted energy spectrum 。The forward mode residual 。

[0058] The immediate difference deviation 。The back view window width ;The window average deviation 。If the signs are the same, they are all positive, and the cumulative sum in the same direction gives 。

[0059] The base of the exponential weight value in the high-order interval ,and the base in the low-order interval 。Positive residual after magnification ,Negative residual 。The sign remains positive.

[0060] Abnormal gain 。Set the historical mean ,Standard deviation ,The three - times standard - deviation threshold is 。 Is marked as a valid anomaly. After averaging the full spectrum, a comprehensive index is obtained, which is higher than the engineering warning line ,and the system prompts "The energy at the resonance order exceeds the limit, and the insulation and grounding integrity need to be checked."

[0061] Reference Figure 2 ,such as Figure 2 shown by the central vertical dotted line. By traversing the base energy spectrum, the global maximum point of the base energy spectrum is obtained, and the harmonic order corresponding to this maximum point is confirmed as the natural resonance order of the ring - network box bus. This natural resonance order is used as a dividing line to divide all harmonic orders into a high - order interval and a low - order interval. As Figure 2 shown in the following annotation, the high - order interval refers to all harmonic orders greater than the natural resonance order, which is located in the right - hand region of the figure; the low - order interval refers to all harmonic orders less than or equal to the natural resonance order, which is located in the left - hand region of the figure. The two intervals are respectively identified by rectangular boxes, ensuring the clarity and accuracy of the interval division. For each harmonic order in the high - order interval, a monotonically increasing exponential weight is constructed according to the ratio of its harmonic order to the natural resonance order, which is used to amplify the amplitude of the positive - mode residual. As Figure 2 shown by the right - hand arrow and the "increasing amplification" annotation, the amplification factor of the high - order harmonics increases with the increase of the harmonic order. For each harmonic order in the low - order interval, a monotonically decreasing exponential weight is constructed according to the ratio of the natural resonance order to its harmonic order, which is used to amplify the amplitude of the negative - mode residual. As Figure 2 shown by the left - hand arrow and the "decreasing amplification" annotation, the amplification factor of the low - order harmonics increases with the decrease of the harmonic order. As Figure 2 shown by the solid curve in the figure, taking the base energy spectrum of the harmonic order as the normalization reference, the positive residual and the negative residual after complementary - index amplification are multiplied at the corresponding harmonic order to obtain the abnormal - gain spectrum after bi - directional residual fusion. The dotted line in the figure represents the original residual, and the solid line represents the abnormal - gain spectrum after complementary - index amplification and fusion. The comparison between the two clearly shows the enhancement effect of the fusion process.

[0062] Reference Figure 3 ,such as Figure 3As shown in the upper part, the specific implementation process of the forward modal residual encoding includes: for any target harmonic order, first select a limited number of base energy spectra of low harmonic orders as the recurrence set, where the low harmonic orders are less than the target harmonic order. Subsequently, weights are assigned according to the relative magnitudes of the active powers corresponding to each harmonic order in the recurrence set to form a theoretical predicted energy spectrum. The difference between the theoretical predicted energy spectrum and the actual base energy spectrum of the target harmonic order is obtained to get the forward modal residual. Figure 3 The change trend of the forward modal residual is shown through the coordinate system in, and this residual is used to highlight the sudden increase anomaly of the high-order harmonic energy. As Figure 3 As shown in the lower part, the implementation process of the reverse modal residual encoding includes: first, the difference between the base energy spectra of the adjacent and higher-order harmonic order and the current harmonic order is obtained to get the immediate differential deviation; then, a harmonic order window with a width equal to the inherent harmonic span of a single loop network box is set in the direction from the current harmonic order to the lower order, and the average of all the base energy spectra within the harmonic order window is obtained to get the window average deviation.

[0063] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring the abnormal energy spectrum of electrical variables of a complete set of ring network boxes, characterized in that, The method includes: Step 1: Within a preset analysis window, normalize the electrical variables of each harmonic order in the ring main unit, calculate the average of the converted energies at all sampling points according to the same harmonic order, and construct the base energy spectrum under normal operating conditions to represent the reference energy distribution of each harmonic order; Step 2: Based on the base energy spectrum, perform forward modal residual coding to obtain the forward modal residual to highlight the sudden increase anomaly of high-order harmonic energy; based on the base energy spectrum, perform reverse modal residual coding to form the reverse modal residual to capture the slow change or drift anomaly of low-order harmonic energy; Step 3: According to the natural resonance order of the ring main unit bus, perform two-way residual energy spectrum fusion to achieve unified enhancement of multi-scale anomaly features and obtain the anomaly gain spectrum; Step 4: Construct an adaptive decision threshold; when the anomaly gain spectrum of a certain harmonic order exceeds the adaptive decision threshold, it is recorded as an effective anomaly; exponentially accumulate and average the anomaly gain spectra of all harmonic orders according to the proportion of the harmonic order numbers, and output the comprehensive anomaly energy spectrum index of the electrical variables of the complete set of ring main units for alarm triggering.

2. The method for measuring the abnormal energy spectrum of electric variables of a complete set of ring network boxes according to claim 1, wherein, The electrical variables include: the effective value of phase voltage, the effective value of phase current, the single pulse energy of partial discharge, and the environmental magnetic flux density.

3. A complete set of ring network box electrical variable abnormal energy spectrum measurement method according to claim 2, characterized in that, In Step 2, the process of performing forward modal residual coding specifically includes: for any target harmonic order, select a limited number of base energy spectra of low harmonic orders as the recurrence set; distribute weights according to the relative magnitudes of the active powers corresponding to each harmonic order in the recurrence set to form the theoretical predicted energy spectrum; subtract the theoretical predicted energy spectrum from the actual base energy spectrum of the target harmonic order to obtain the forward modal residual, which is used to highlight the sudden increase anomaly of high-order harmonic energy.

4. The method for measuring the energy spectrum of abnormal electrical variables of a complete set of ring network boxes according to claim 3, characterized in that, The order of the low harmonics is less than the order of the target harmonics; the defined quantity takes values as follows: , where is the natural resonance order of the ring network cabinet bus; if , then the selected order of the low harmonics is only from 1 to ; otherwise, select the nearest harmonic orders as the order of the low harmonics; is the order of the target harmonics.

5. A complete set of ring network box electrical variable abnormal energy spectrum measurement method according to claim 4, characterized in that, In Step 2, the process of performing reverse modal residual coding includes: first subtract the base energy spectra of the adjacent and higher-order harmonic order from the current harmonic order to obtain the instantaneous differential deviation; then set a harmonic order window with a width equal to the inherent harmonic span of a single ring main unit in the low-order direction from the current harmonic order, and calculate the average of all base energy spectra within the harmonic order window; combine the instantaneous differential deviation and the window average deviation to form the reverse modal residual, which is used to sensitively capture the slow change or drift anomaly of low-order harmonic energy.

6. The method for measuring the abnormal energy spectrum of electrical variables of a complete set of ring network boxes according to claim 5, characterized in that, The process of combining the instantaneous differential deviation and the window average deviation to form the reverse modal residual includes: divide the instantaneous differential deviation and the window average deviation by their respective means of the target harmonic order under historical normal operating conditions; Judge whether the signs of the instantaneous differential deviation and the window average deviation are the same for co-directionality verification, including: if the signs are the same, retain their absolute values and add them to enhance the detection sensitivity of co-directional anomalies; if the signs are opposite, take the difference between their absolute values and retain the actual sign of the result to avoid masking the trend anomaly due to positive and negative cancellation; define the result after co-directionality verification as the reverse modal residual of the target harmonic order.

7. The method for measuring the abnormal energy spectrum of electric variables of a complete set of ring network boxes according to claim 6, characterized in that, Step 3 includes: according to the natural resonance order of the ring network cabinet bus, respectively performing complementary exponential amplification on the forward modal residual and the reverse modal residual, so that the high-order residuals are amplified incrementally according to the harmonic order ratio and the low-order residuals are amplified decrementally; subsequently, taking the base energy spectrum of the harmonic order as the normalization reference, multiplying the forward residual and the reverse residual after performing complementary exponential amplification at the corresponding harmonic order to obtain the abnormal gain spectrum after the fusion of the bidirectional residuals, thereby realizing the unified enhancement of multi-scale abnormal features.

8. A complete set of ring network box electrical variable abnormal energy spectrum measurement method according to claim 7, characterized in that, The process of respectively performing complementary exponential amplification on the forward modal residual and the reverse modal residual includes: traversing the base energy spectrum to obtain the global maximum point of the base energy spectrum; confirming the harmonic order corresponding to this maximum point as the natural resonance order of the ring network cabinet bus; taking the natural resonance order as the boundary, dividing all harmonic orders into a high-order interval and a low-order interval; the high-order interval refers to all harmonic orders greater than the natural resonance order, and the low-order interval refers to all harmonic orders less than or equal to the natural resonance order; for each harmonic order in the high-order interval, constructing a monotonically increasing exponential weight according to the ratio of its harmonic order to the natural resonance order for amplifying the amplitude of the forward modal residual; for each harmonic order in the low-order interval, constructing a monotonically decreasing exponential weight according to the ratio of the natural resonance order to its harmonic order for amplifying the amplitude of the reverse modal residual; after taking the absolute values of the forward modal residual and the reverse modal residual respectively, introducing the corresponding exponential weights; taking the exponential weights as the exponents, performing a power operation on the absolute values of the amplified forward modal residual and reverse modal residual while retaining the original positive and negative signs to obtain the forward residual and the reverse residual after complementary exponential amplification.

9. A method for measuring the abnormal energy spectrum of electrical variables of a complete set of ring network cabinets according to claim 8, characterized in that, The process of constructing the adaptive decision threshold includes: for each harmonic order, calling the mean and standard deviation of the abnormal gain spectrum under the historical normal working conditions to construct an adaptive decision threshold of three times the standard deviation.

Images