Intelligent monitoring method based on track-type electric energy meter

Abstract

The application discloses a kind of intelligent monitoring methods based on guide rail type electric energy meter, it is related to electric energy meter monitoring technical field, including the following steps: real-time acquisition current waveform data in load loop by high-precision current sensor built-in guide rail type electric energy meter;Determine the position of zero-crossing point based on zero-crossing detection logic, record the time stamp corresponding to each detected zero-crossing point, around each detected zero-crossing point, extract a fixed length small window waveform data section, form a zero-crossing neighborhood waveform subset.The application intelligently predicts current waveform distortion by machine learning model, and adaptively adjusts the determination sensitivity of zero-crossing point according to the distortion degree, reduces the misjudgment caused by current waveform distortion, avoids the problems such as relay misoperation and load switching error caused by polarity misjudgment, effectively improves the intelligent level and reliability of guide rail type electric energy meter, ensures the efficient and stable operation of equipment in complex power grid environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric energy meter monitoring, and particularly relates to an intelligent monitoring method based on a guide rail type electric energy meter. BACKGROUND

[0002] The intelligent monitoring of the guide rail type electric energy meter refers to realizing real-time monitoring and data analysis of the electric energy meter and the monitored circuit by integrating intelligent sensors, data acquisition modules and communication modules in the electric energy meter. Its functions not only include traditional electric energy measurement, but also can automatically collect various power parameters such as current, voltage and power factor, and detect abnormal conditions in the power grid in real time, such as overload, short circuit and power fluctuation, and transmit data to the cloud or a local management platform through wireless communication technology. The intelligent monitoring system can remotely monitor the running state of the power equipment, diagnose and warn faults through big data analysis and artificial intelligence algorithms, identify potential problems in advance, optimize power resource scheduling, and support remote control and automatic management, thereby improving the efficiency, safety and reliability of the power system.

[0003] The current zero-crossing distortion of the guide rail type electric energy meter refers to the phenomenon of abnormal deviation, delay or waveform distortion when the current waveform normally passes through the zero point (i.e. from positive to negative or from negative to positive). In theory, the alternating current waveform should pass through the zero point smoothly and symmetrically, but when there are nonlinear loads (such as frequency converters, switching power supplies, large motor starts) or circuit faults, the current zero-crossing point may be distorted, jump or deviate. For the guide rail type electric energy meter, monitoring the current zero-crossing distortion is important, on the one hand, it can improve the ability to identify abnormal load working conditions and timely discover potential equipment aging, poor grounding or power quality problems; on the other hand, it can also ensure the accuracy of electric energy metering, because if the zero-crossing distortion is not detected and compensated in time, it will cause errors in the calculation of active and reactive power, especially affecting the accurate measurement of dynamic load power and the reliability of subsequent intelligent analysis. Therefore, zero-crossing distortion monitoring is an important part of the intelligent function of the guide rail type electric energy meter to improve system stability, measurement accuracy and early fault warning capability.

[0004] The prior art has the following disadvantages:

[0005] In the prior art, when potential distortion occurs in the current waveform during passing through the zero-crossing point, the traditional track-type electric energy meter usually has difficulty in timely and accurately distinguishing the change of current polarity, thereby causing misjudgment of the internal control module when performing load management operations (such as load switching, fault isolation, etc.). Due to the misjudgment of polarity, the built-in load relay of the track-type electric energy meter may have abnormal switching, which is specifically manifested as power-off without cause in the case of normal load operation, or refusal to act when it needs to be disconnected. The above abnormal switching phenomenon not only interrupts the continuous power supply of normal load, causing abnormal operation or failure of terminal equipment, but also may cause large-scale power interruption, failure of protection mechanism and other serious safety hazards in extreme cases, thereby seriously affecting the operation stability of the track-type electric energy meter and the reliability of system application.

[0006] The above information disclosed in the background section is only intended to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute the prior art known to those of ordinary skill in the art. SUMMARY

[0007] The purpose of the present application is to provide an intelligent monitoring method based on a track-type electric energy meter, which ensures high-quality input of the current waveform through a high-precision current sensor and high-sampling-rate data acquisition. The zero-crossing point is accurately located using zero-crossing detection logic, and the waveform subset is extracted around each zero-crossing point, ensuring accurate capture of the current waveform in the zero-crossing neighborhood. Further feature extraction and deep analysis effectively quantify the distortion degree of the current waveform. The machine learning model intelligently predicts based on these features, accurately identifies the distortion of the current waveform, and adaptively adjusts the judgment sensitivity of the zero-crossing point according to the distortion degree, reducing misjudgment caused by distortion and avoiding problems such as relay misoperation and load switching errors caused by polarity misjudgment, effectively improving the intelligent level and reliability of the track-type electric energy meter, and ensuring efficient and stable operation of the equipment in complex power grid environments, to solve the problems in the above background.

[0008] In order to achieve the above purpose, the present application provides the following technical solutions: an intelligent monitoring method based on a track-type electric energy meter, comprising the following steps:

[0009] The current waveform data in the load loop is collected in real time by a high-precision current sensor (such as a shunt, a Hall element, etc.) built in the track-type electric energy meter at a high sampling rate (such as 20 kHz or higher);

[0010] The position of the zero-crossing point is determined based on zero-crossing detection logic (for example, finding a pair of sampling points where the current value changes from positive to negative or from negative to positive), and the timestamp corresponding to each detected zero-crossing point is recorded;

[0011] A small window waveform data segment (such as the sampling data within 1 ms before and after) of a fixed length is extracted around each detected zero-crossing point to form a zero-crossing neighborhood waveform subset;

[0012] For the current waveform data in each zero-crossing neighborhood waveform subset, potential current waveform distortion features are extracted therefrom, and after deep analysis of the extracted features through feature engineering techniques, the current waveform distortion of the zero-crossing point is quantified based on the analyzed features;

[0013] The processed features are inputted into a machine learning model (such as a lightweight neural network, SVM or decision tree model) that has been trained in advance, and the machine learning model is used to intelligently predict the current waveform distortion in the zero-crossing neighborhood waveform subset;

[0014] When the zero-crossing neighborhood waveform subset is identified to have current waveform distortion, the sensitivity of the zero-crossing point is automatically relaxed according to the degree of current waveform distortion, specifically: the actual zero-crossing determination threshold is adaptively shifted according to the degree of current waveform distortion to reduce the misjudgment caused by distortion.

[0015] Preferably, the position of the zero-crossing point is determined based on zero-crossing detection logic, and the specific steps are as follows:

[0016] In the collected current waveform data, polarity change detection is first performed, i.e., the polarity of two consecutive sampling points is compared, and when the current sampling point has a positive current value and the next sampling point has a negative current value, or the current sampling point has a negative current value and the next sampling point has a positive current value, it is preliminarily determined that there is a zero-crossing phenomenon;

[0017] Next, the crossover section is determined, i.e., the section between the two consecutive sampling points where the polarity change is detected is regarded as the zero-crossing transition zone, and it is determined that the actual zero-crossing point falls within this section;

[0018] Finally, the position of the zero-crossing point is calculated, and the linear interpolation or quadratic interpolation method is used to calculate the exact position of the current signal crossing the zero axis to improve the accuracy of the zero-crossing point determination and reduce the interference of distortion or noise.

[0019] Preferably, the length of the small window waveform data segment is flexibly adjusted according to the response requirements of the intelligent monitoring system and the sampling rate, and the length is adjusted to cover the complete transition waveform before and after the zero-crossing point.

[0020] Preferably, for each current waveform data in the zero-crossing neighborhood waveform subset, potential current waveform distortion features are extracted therefrom, wherein the extracted features include the tail length of the current signal near the zero point, and a zero-crossing tail reference value is generated after deep analysis of the extracted features through feature engineering techniques, and the distortion of the current waveform is quantified by the zero-crossing tail reference value.

[0021] Preferably, the specific steps of generating the zero-crossing tail reference value by deep analysis of the tail length of the current signal near the zero point through feature engineering techniques are as follows:

[0022] In each zero-crossing neighborhood waveform subset, first define the tail section near the zero point, specifically a sequence of sampling points that continuously satisfy the condition that the absolute value of the current is lower than a set small threshold (for example, 1% of the rated maximum current);

[0023] After the tail section is extracted, an energy decay curve is constructed based on the current amplitude variation, and the nonlinear cumulative amount of the current amplitude in the tail section is accumulated by integration, and the cumulative expression is:

[0024]

[0025] Ψ x is the total nonlinear energy of the tail section current amplitude; I i represents the current amplitude of the i-th sampling point in the tail section; Δ|I i represents the current amplitude increment of adjacent sampling points, Δ|I i | = |I i |I i-1 |; λ is the nonlinear amplitude index of energy accumulation, usually taking 1.5-2.5, used to emphasize the small residual energy, γ is the mutation index of current amplitude increment, which can be taken as 1.2-2.0, used to capture the mutation trend, and ξ is the amplitude variation weight factor, used to regulate the influence proportion of mutation change, which can be set according to environmental experience, such as in the range of 0.3-0.5;

[0026] After obtaining the total nonlinear energy of the tail section, it is further nonlinearly normalized with the maximum current amplitude in the overall zero-crossing neighborhood and the tail section length to generate the final zero-crossing tail reference value, and the generated expression is:

[0027]

[0028] C tt is the zero-crossing tail reference value, α, β, θ are exponential adjustment factors (usually set between 1.0-2.0), used to strengthen the relative weight relationship of the nonlinear energy total amount of the current amplitude, the maximum current amplitude and the time length; I maxwherein, I max is the maximum current amplitude in the zero-crossing neighborhood waveform subset (for normalizing the waveform energy scale), L is the number of effective sampling points in the tail section (for reflecting the tail duration length).

[0029] Preferably, the zero-crossing tail reference value is input into a machine learning model trained in advance, and the zero-crossing tail reference value and the current waveform distortion risk are analyzed by the machine learning model to output a distortion risk coefficient, and the current waveform distortion in the zero-crossing neighborhood waveform subset is intelligently predicted based on the distortion risk coefficient.

[0030] Preferably, the distortion risk coefficient generated when the current waveform distortion at the zero-crossing point is intelligently predicted by the machine learning model is compared and analyzed with a pre-set distortion risk reference threshold value to identify the current waveform distortion in the zero-crossing neighborhood waveform subset, and the specific identification steps are as follows:

[0031] If the distortion risk coefficient is greater than the distortion risk reference threshold value, the zero-crossing neighborhood waveform subset is identified as a current waveform distortion subset;

[0032] If the distortion risk coefficient is less than or equal to the distortion risk reference threshold value, the zero-crossing neighborhood waveform subset is identified as a current waveform normal subset.

[0033] Preferably, when the current waveform distortion in the zero-crossing neighborhood waveform subset is identified, the actual zero-crossing determination threshold is adaptively shifted according to the distortion degree of the current waveform, and the specific steps are as follows:

[0034] After detecting the current waveform distortion in the zero-crossing neighborhood waveform subset, first, the difference between the distortion risk coefficient and the distortion risk reference threshold value is quantified, and a sensitivity adjustment factor is calculated, and the expression is:

[0035]

[0036] wherein, Δμ is the sensitivity adjustment factor, used to reflect the dynamic influence of the actual distortion degree on the zero-crossing determination threshold, η is a sensitivity adjustment reference coefficient, which is a positive real number, usually between 0.1 and 0.3, used to control the adjustment amplitude, C tt is the distortion risk coefficient generated when the machine learning model predicts the current waveform distortion at the zero-crossing point, κ0 is the distortion risk reference threshold value, φ is a nonlinear amplification index, φ > 1, usually between 1.5 and 2.0, used to enhance the adjustment response when the distortion degree is high, and if C tt ≤ κ0, then Δμ = 0, i.e. the sensitivity is not adjusted when the distortion does not exceed the reference threshold value;

[0037] After obtaining the sensitivity adjustment factor Δμ, further adaptive offset is carried out based on the standard zero-crossing judgment threshold to generate a new actual zero-crossing judgment threshold, and the generated expression is:

[0038] ∈' = ∈0·(1 + Δμ)

[0039] Where ∈0 is the initial set standard zero-crossing judgment threshold (for example, ±5 mA), and ∈' is the actual zero-crossing judgment range dynamically adjusted according to the current distortion risk. Through this adaptive offset mechanism, when the current waveform is severely distorted, the actual zero-crossing judgment threshold is appropriately relaxed, and the current amplitude range allowed for zero-crossing judgment is expanded, thereby effectively reducing the zero-crossing misjudgment caused by waveform tailing and oscillation; when the distortion risk is small, the judgment sensitivity is maintained, and the high-precision identification performance of the system is ensured.

[0040] In the above technical solution, the technical effects and advantages provided by the present application are:

[0041] The present application ensures high-quality input of the current waveform through high-precision current sensors and high-sampling-rate data acquisition. The zero-crossing points are accurately located using zero-crossing detection logic, and waveform subsets are extracted around each zero-crossing point to ensure accurate capture of the current waveform in the zero-crossing neighborhood. Further feature extraction and deep analysis effectively quantify the distortion degree of the current waveform. The machine learning model intelligently predicts based on these features, accurately identifies the distortion of the current waveform, and adaptively adjusts the judgment sensitivity of the zero-crossing point according to the distortion degree, reduces the misjudgment caused by distortion, avoids relay misoperation and load switching errors caused by polarity misjudgment, and effectively improves the intelligent level and reliability of the guide rail type electric energy meter, ensuring efficient and stable operation of the equipment in complex power grid environments. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments or prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0043] Figure 1 The present application is based on the intelligent monitoring method of the guide rail type electric energy meter. DETAILED DESCRIPTION

[0044] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive concept to those skilled in the art.

[0045] The present application provides a smart monitoring method based on a track-type electric energy meter as shown in the specification Figure 1 The smart monitoring method based on a track-type electric energy meter comprises the following steps:

[0046] The high-precision current sensor (such as a shunt, a Hall element, etc.) built in the track-type electric energy meter is used to collect current waveform data in the load loop in real time at a high sampling rate (such as 20 kHz or above);

[0047] The collected data should cover a complete AC cycle to ensure the detailed capture of slight fluctuations near the zero-crossing point. Real-time, continuous and accurate acquisition of current waveform data is the basis for subsequent zero-crossing detection and waveform feature analysis. Ensuring that the sampling precision and sampling rate are high enough can maximize the restoration of the real state of the current waveform near the zero-crossing point and provide reliable data support for distortion identification.

[0048] The position of the zero-crossing point is determined based on zero-crossing detection logic (for example, finding a pair of sampling points at which the current value changes from positive to negative or from negative to positive), and the timestamp corresponding to each detected zero-crossing point is recorded;

[0049] The position of the zero-crossing point is determined based on zero-crossing detection logic, and the specific steps are as follows:

[0050] In the collected current waveform data, first, polarity change detection is performed, that is, the polarity of two consecutive sampling points is compared. When it is detected that the current value of the current sampling point is positive and the current value of the next sampling point is negative, or the current sampling point is negative and the next sampling point is positive, it is preliminarily determined that there is a zero-crossing phenomenon. Then, the crossover section is determined, that is, the section between the two consecutive sampling points where the polarity change is detected is regarded as the zero-crossing transition zone, and it is determined that the actual zero-crossing point falls within this section. Finally, the crossover point position is calculated. Through linear interpolation or quadratic interpolation of the current values of the two sampling points in the section, the accurate position where the current signal truly crosses the zero axis is calculated to improve the accuracy of the zero-crossing point determination and reduce the interference of distortion or noise.

[0051] When recording the timestamp corresponding to each detected zero-crossing point, the time positioning of the zero-crossing point can be further optimized by linear interpolation and other methods to improve the accuracy of zero-crossing identification. Accurate identification of the zero-crossing time of the current waveform is the key to distinguishing the polarity change of the current and synchronously controlling the action. Accurate zero-crossing point identification can lay a time reference for subsequent extraction of zero-crossing neighborhood waveform features and dynamic monitoring, and directly affects the accuracy and timeliness of the entire intelligent monitoring system.

[0052] Around each detected zero-crossing point, a small window waveform data segment (such as sampling data within 1 ms before and after) of a fixed length is extracted to form a zero-crossing neighborhood waveform subset;

[0053] The length of the small window waveform data segment can be flexibly adjusted according to the system response requirement and the sampling rate to cover the complete transition waveform before and after the zero-crossing point. By locally amplifying the data near the zero-crossing point, the potential distortion characteristics of the waveform can be captured. Establishing a zero-crossing neighborhood waveform subset can effectively focus on the analysis of the local characteristics related to the zero-crossing, improve the pertinence and sensitivity of the distortion detection, and reduce the computational redundancy brought by the full waveform analysis.

[0054] For the current waveform data in each zero-crossing neighborhood waveform subset, potential current waveform distortion features are extracted therefrom. After in-depth analysis of the extracted features through feature engineering techniques, the current waveform distortion of the zero-crossing point is quantified based on the analyzed features.

[0055] For the current waveform data in each zero-crossing neighborhood waveform subset, potential current waveform distortion features are extracted therefrom. The extracted features include the tail length of the current signal near the zero point. After in-depth analysis of the extracted features through feature engineering techniques, a zero-crossing tail reference value is generated, and the distortion of the current waveform is quantified through the zero-crossing tail reference value.

[0056] In an ideal case, the current waveform should cross the zero point in a continuous, smooth, and rapid manner, that is, the current value near the zero-crossing point should quickly transition from positive to negative or from negative to positive. However, when the current signal exhibits a tail phenomenon near the zero point, that is, the current value remains near zero for a long time without quickly completing the polarity reversal, this usually reflects that the waveform has a potential distortion. The reason is that the tail phenomenon is often caused by the superposition of high-frequency noise components, harmonic distortion, or transient discontinuity due to the nonlinear characteristics of the load, resulting in uneven energy release during the zero-crossing process, thereby disrupting the clear switching of normal positive and negative polarity. Therefore, a longer tail length near the zero-crossing point can be used as an important feature indicator to identify potential distortion in the current waveform, indicating abnormal power quality or abnormal system load state.

[0057] The specific steps for generating the zero-crossing tail reference value through feature engineering techniques to analyze the tail length of the current signal near the zero point are as follows:

[0058] In each zero-crossing neighborhood waveform subset, first define the tail section near the zero point, specifically a sequence of sampling points that continuously satisfy the condition that the absolute value of the current is below a set small threshold (e.g., 1% of the rated maximum current);

[0059] After the tail section is extracted, an energy decay curve is constructed based on the change in current amplitude. The nonlinear accumulation of the current amplitude in this tail section is accumulated by integration, and the accumulated expression is:

[0060]

[0061] Ψ = ∑ i = 1 n | I i | λ exp ( - γ | I i - I i - 1 | ) ξ x is the nonlinear energy sum of the tail section current amplitude; I i represents the current amplitude of the i-th sampling point of the tail section; Δ|I i | represents the current amplitude increment of adjacent sampling points, Δ|I i | = | I i - I i - 1 | i | = | I i - I i - 1 | i-1 |; λ is the nonlinear amplitude index of energy accumulation, usually taking 1.5-2.5, for emphasizing the small residual energy, γ is the mutation index of current amplitude increment, which can take 1.2-2.0, for capturing the mutation trend, and ξ is the amplitude change weight factor, for regulating the influence proportion of mutation change, which can be set according to environmental experience, such as 0.3-0.5;

[0062] By constructing the energy decay curve, the residual energy density and fluctuation intensity inside the tail section are directly quantified, laying a foundation for subsequent distortion severity assessment.

[0063] After obtaining the nonlinear energy sum of the tail section, it is further nonlinearly normalized with the maximum current amplitude in the overall zero-crossing neighborhood and the tail section length to generate the final crossing tail reference value, and the generated expression is:

[0064]

[0065] , wherein: C tt is the crossing tail reference value, α, β, θ are exponential adjustment factors (usually set between 1.0-2.0), for strengthening the relative weight relationship of the nonlinear energy sum of the current amplitude, the maximum current amplitude and the time length; I max is the maximum current amplitude in the waveform subset of the zero-crossing neighborhood (for normalizing the waveform energy scale), and L is the effective sampling point number in the tail section (for reflecting the tail duration length);

[0066] Through the tensor normalization of energy-amplitude-time, the transient energy release of the current waveform, the tail existence time and the scale difference are comprehensively considered, the highly sensitive crossing tail reference value is generated, which can accurately quantify the severity of potential distortion of the current waveform at the zero-crossing point, and has strong distinguishing ability and adaptability.

[0067] From the cross-tail reference value, the greater the performance value of the cross-tail reference value generated after deep analysis of the tail length of the current signal near zero point by feature engineering technology, the greater the risk of potential current waveform distortion in the zero-crossing neighborhood waveform subset, and vice versa. The reason is that the cross-tail reference value is a quantitative index generated after feature extraction, non-linear energy accumulation and normalization analysis of the tail section in the zero-crossing neighborhood of the current signal. The greater the performance value, the higher the residual current energy in the tail section, the more obvious the oscillation or discontinuity, and the longer the tail duration, which comprehensively reflects the greater the distortion degree of the waveform near the zero-crossing point. Because in the ideal current zero-crossing process, the current should pass through the zero point smoothly and quickly, the tail length is short and the energy decays quickly. Once there is high-frequency interference, harmonic distortion or load nonlinearity effect, the current will produce abnormal stagnation, oscillation or energy accumulation near the zero point, resulting in a significant increase in the cross-tail reference value. Therefore, the greater the performance value of the cross-tail reference value, the higher the risk of potential current waveform distortion at the zero-crossing point; otherwise, the smaller the cross-tail reference value, the closer the waveform to the ideal state, and the lower the distortion risk.

[0068] The processed features are input into a machine learning model (such as a lightweight neural network, SVM or decision tree model) trained in advance, and the current waveform distortion in the zero-crossing neighborhood waveform subset is intelligently predicted by the machine learning model.

[0069] The cross-tail reference value is input into a machine learning model trained in advance, and the cross-tail reference value and current waveform distortion risk are analyzed by the machine learning model, and the distortion risk coefficient is output. Based on the distortion risk coefficient, the current waveform distortion in the zero-crossing neighborhood waveform subset is intelligently predicted.

[0070] The machine learning model trained in advance refers to an intelligent prediction model generated by pre-training a large amount of historical waveform data and fixing the training parameters for potential distortion phenomena in the zero-crossing neighborhood current waveform. Specifically, in the model training stage, the system first extracts key characteristic features from each zero-crossing neighborhood waveform subset based on large-scale current waveform data collected under various operating conditions, especially the crossing tail reference value as the core feature input. At the same time, through artificial labeling or auxiliary diagnosis, each group of data is associated with a specific distortion risk label (for example: normal, slight distortion, moderate distortion, severe distortion). During the training process, the machine learning model learns that there is a positive correlation between the crossing tail reference value and the current waveform distortion risk, that is, the larger the crossing tail reference value, the more serious the waveform tail phenomenon near the zero-crossing point, the more energy remains, and the greater the oscillation amplitude, so the current waveform distortion risk is also higher. In order to better fit this positive relationship, supervised learning algorithms such as support vector machine (SVM), gradient boosting tree (GBDT), and lightweight neural network (LightweightNN) are used in the training stage to iteratively optimize the model's internal weight parameters to sensitively capture feature trends, and maintain high generalization ability under different distortion levels. After the model is trained, it is tested on a strict validation set (evaluation accuracy, recall rate, area under the curve AUC, etc.) to ensure its consistency and interpretability in judging distortion risk under different tail features. Finally, the weights and decision rules are fixed to form a reasoning model that can be directly applied to real-world scenarios. Because the positive mapping relationship between the crossing tail reference value and the distortion risk level is explicitly established during the training stage, the machine learning model trained in advance can achieve high-precision, positive-consistent intelligent distortion prediction based on input features in actual use.

[0071] In the actual application stage, when the guide rail type electric energy meter detects a zero-crossing neighborhood waveform subset, the system extracts the crossing tail reference value from the subset and inputs it into the machine learning model trained in advance. Based on the feature-risk positive relationship learned during training, the model intelligently analyzes the input crossing tail reference value and positively infers a continuous numerical form of distortion risk coefficient. The distortion risk coefficient is usually defined in the interval of 0 to 1, where the value closer to 1 indicates a higher potential distortion risk of the current waveform; the value closer to 0 indicates that the waveform is closer to the ideal state and the distortion risk is lower. Through the positive analysis logic, the model can naturally reflect the direct pull effect of the crossing tail reference value change on the distortion risk, without complex reverse reasoning or special transformation, greatly improving the intuitiveness and reliability of the judgment.

[0072] The greater the performance value of the cross-tail reference value generated by the deep analysis of the tail length of the current signal near zero point through feature engineering technology, that is, the greater the performance value of the distortion risk coefficient generated by the intelligent prediction of the current waveform distortion of the zero-crossing point through the machine learning model, the greater the risk of potential current waveform distortion in the zero-crossing neighborhood waveform subset, and vice versa.

[0073] The distortion risk coefficient generated by the intelligent prediction of the current waveform distortion of the zero-crossing point through the machine learning model is compared and analyzed with the pre-set distortion risk reference threshold to identify the current waveform distortion in the zero-crossing neighborhood waveform subset. The specific identification steps are as follows:

[0074] If the distortion risk coefficient is greater than the distortion risk reference threshold, the zero-crossing neighborhood waveform subset is identified as a current waveform distortion subset.

[0075] If the distortion risk coefficient is less than or equal to the distortion risk reference threshold, the zero-crossing neighborhood waveform subset is identified as a current waveform normal subset.

[0076] When it is identified that there is current waveform distortion in the zero-crossing neighborhood waveform subset, the determination sensitivity of the zero-crossing point is automatically relaxed according to the distortion degree of the current waveform. Specifically, the actual zero-crossing determination threshold is adaptively shifted according to the distortion degree of the current waveform to reduce the false judgment caused by distortion.

[0077] When it is identified that there is current waveform distortion in the zero-crossing neighborhood waveform subset, the actual zero-crossing determination threshold is adaptively shifted according to the distortion degree of the current waveform. The specific steps are as follows:

[0078] After detecting that there is current waveform distortion in the zero-crossing neighborhood waveform subset, first, the difference between the distortion risk coefficient and the distortion risk reference threshold is quantified, and a sensitivity adjustment factor is calculated. The expression is:

[0079]

[0080] where Δμ is the sensitivity adjustment factor, which is used to reflect the dynamic influence of the actual distortion degree on the zero-crossing determination threshold, η is the sensitivity adjustment reference coefficient, which is a positive real number, usually between 0.1 and 0.3, and is used to control the adjustment amplitude, C tt is the distortion risk coefficient generated by the machine learning model for predicting the current waveform distortion of the zero-crossing point, κ0 is the distortion risk reference threshold, φ is a nonlinear amplification index, φ > 1, usually between 1.5 and 2.0, and is used to enhance the adjustment response when the distortion degree is high, and if C tt≤ κ0, then let Δμ = 0, that is, the sensitivity is not adjusted when the distortion does not exceed the reference threshold;

[0081] When the distortion risk coefficient is less than or equal to the distortion risk reference threshold, the sensitivity adjustment factor Δμ = 0 is set, which is mainly based on the dual consideration of system operation stability and detection accuracy. The distortion risk reference threshold is usually set according to a large amount of actual operation data, representing the maximum normal fluctuation range of the current waveform in the zero-crossing neighborhood that can be accepted. When the distortion risk coefficient does not exceed the distortion risk reference threshold, it means that the current current waveform may have weak disturbances, but the overall is still in the normal or tolerable change level, which is not enough to cause substantial impact on the accuracy of zero-crossing point determination. Therefore, in order to avoid unnecessarily relaxing the sensitivity and causing the determination accuracy to decrease, the system keeps the original standard determination sensitivity unchanged in this case, that is, Δμ = 0, to ensure that the electric energy meter still has high precision and high reliability in detecting normal waveforms, thereby achieving the best balance between stability and adaptability.

[0082] Through this formula, a quantitative sensitivity adjustment factor is dynamically generated according to the positive deviation between the distortion risk coefficient and the distortion risk reference threshold, so that the higher the distortion risk, the greater the relaxation of the zero-crossing determination sensitivity, thereby better adapting to different degrees of waveform distortion.

[0083] After obtaining the sensitivity adjustment factor Δμ, further adaptive offset is made based on the standard zero-crossing determination threshold to generate a new actual zero-crossing determination threshold, and the generated expression is:

[0084] ∈' = ∈0·(1+Δμ)

[0085] where ∈0 is the initially set standard zero-crossing determination threshold (e.g. ±5mA), and ∈' is the actual zero-crossing determination range dynamically adjusted according to the current distortion risk. Through this adaptive offset mechanism, when the current waveform is severely distorted, the actual zero-crossing determination threshold is relaxed, expanding the current amplitude range allowed for zero-crossing determination, thereby effectively reducing the zero-crossing misjudgment caused by waveform tailing and oscillation; and when the distortion risk is small, the determination sensitivity is kept strict to ensure the high precision identification performance of the system;

[0086] By the action of the sensitivity adjustment factor on the standard zero-crossing determination threshold, the zero-crossing determination strategy is dynamically adjusted according to the distortion risk, which can improve the robustness of the system under high risk conditions, and maintain high measurement accuracy under low risk conditions, ensuring the intelligent adaptability and reliability of the guide rail type electric energy meter in complex power quality environment.

[0087] Through the above intelligent monitoring method, the measurement accuracy and stability of the guide rail type electric energy meter in a complex current waveform environment can be significantly improved. First, through the high-precision current sensor and high-sampling-rate real-time data acquisition, the high-quality input of the current waveform is ensured, providing a reliable basis for subsequent analysis. Then, the zero-crossing point position is accurately determined through the zero-crossing detection logic, and the waveform subset is extracted around each zero-crossing point, ensuring the detailed capture of the current waveform in the zero-crossing neighborhood. Further feature extraction and deep analysis effectively quantify the distortion of the current waveform, and the machine learning model can accurately identify the distortion of the current waveform through intelligent prediction of these features, and automatically adjust the zero-crossing point determination sensitivity according to the distortion degree, reduce the misjudgment caused by distortion, thereby avoiding the problems of relay misoperation, load switching error and other problems caused by polarity misjudgment, improving the intelligent level and reliability of the guide rail type electric energy meter, and finally ensuring the efficient and stable operation of the equipment in a complex power grid environment.

[0088] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the most recent real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0089] The above only describes certain exemplary embodiments of the present application in a descriptive manner, and it is self-evident that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above figures and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present application.

[0090] It should be understood that the size of the sequence number of each process in various embodiments of the present application does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0091] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0092] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the above-described system, device and unit can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0093] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0094] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0095] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any skilled person in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0096] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above figures and description are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

Claims

1. A method for intelligent monitoring of a track-based electric energy meter, characterized in that, The method comprises the following steps: Real-time acquisition of current waveform data in the load loop through a high-precision current sensor built in the track-type electric energy meter; Determination of the position of the zero-crossing point based on zero-crossing detection logic, recording of the time stamp corresponding to each detected zero-crossing point, extraction of a small window waveform data segment of a fixed length around each detected zero-crossing point, and formation of a zero-crossing neighborhood waveform subset; For the current waveform data in each zero-crossing neighborhood waveform subset, potential current waveform distortion features are extracted therefrom, the extracted features are analyzed in depth through feature engineering technology, the current waveform distortion of the zero-crossing point is quantified based on the analyzed features, the processed features are input into a machine learning model trained in advance, and the machine learning model is used to intelligently predict the current waveform distortion in the zero-crossing neighborhood waveform subset; When recognizing that there is current waveform distortion in the zero-crossing neighborhood waveform subset, the zero-crossing point determination sensitivity is automatically relaxed according to the degree of current waveform distortion, specifically: the actual zero-crossing determination threshold is adaptively shifted according to the degree of current waveform distortion, and the misjudgment caused by distortion is reduced. For the current waveform data in each zero-crossing neighborhood waveform subset, potential current waveform distortion features are extracted therefrom, wherein the extracted features include the tail length of the current signal near the zero point, a crossing tail reference value is generated after the extracted features are analyzed in depth through feature engineering technology, and the crossing tail reference value is used to quantify the distortion of the current waveform. The specific steps of generating the crossing tail reference value after in-depth analysis of the tail length of the current signal near the zero point through feature engineering technology are as follows: In each zero-crossing neighborhood waveform subset, first define the tail section near the zero point, specifically a sequence of sampling points that continuously satisfy the condition that the absolute value of the current is below a set small threshold; After the trailing section is extracted, an energy decay curve is constructed based on the change in current amplitude. The nonlinear cumulative amount of current amplitude within the trailing section is accumulated by integration. The expression for the accumulation is: ,in: The total nonlinear energy of the current amplitude in the tail section; Indicates the trailing section number i Current amplitude at each sampling point; This represents the increment of current amplitude between adjacent sampling points. ; The nonlinear amplitude exponent for energy accumulation. The abrupt change index is a measure of the current amplitude increment, used to capture abrupt change trends. This is a weighting factor for amplitude changes, used to regulate the proportion of influence from abrupt changes; After obtaining the nonlinear energy of the tail section, it is further normalized with the maximum current amplitude and the tail section length in the overall zero-crossing neighborhood to generate the final crossing tail reference value. The expression generated is: wherein: is the crossing tail reference value, , , is an exponential adjustment factor for strengthening the relative weight relationship of the nonlinear energy of the current amplitude, the maximum current amplitude, and the time length; is the maximum current amplitude in the waveform subset of the zero-crossing neighborhood, L is the number of effective sampling points in the tail section. The crossing tail reference value is input into a machine learning model trained in advance, the crossing tail reference value and the current waveform distortion risk are analyzed in a forward direction through the machine learning model, a distortion risk coefficient is output, and the current waveform distortion in the zero-crossing neighborhood waveform subset is intelligently predicted based on the distortion risk coefficient.

2. The track-based electric energy meter based intelligent monitoring method according to claim 1, characterized in that, The specific steps of determining the position of the zero-crossing point based on the zero-crossing detection logic are as follows: In the acquired current waveform data, first perform polarity change detection, i.e., compare the polarity of two consecutive sampling points, when the current value of the current sampling point is positive and the current value of the next sampling point is negative, or the current sampling point is negative and the next sampling point is positive, preliminarily determine that there is a zero-crossing phenomenon; Perform crossing section determination, i.e., regard the section between the two consecutive sampling points where the polarity change is detected as the zero-crossing transition zone, and determine that the actual zero-crossing point falls within this section; Perform crossing point position calculation, i.e., use linear interpolation on the current values of the two sampling points in the section to calculate the exact position where the current signal truly crosses the zero axis.

3. The track-based electric energy meter based intelligent monitoring method according to claim 1, characterized in that, The length of the small window waveform data segment is flexibly adjusted according to the response requirements of the intelligent monitoring system and the sampling rate, and the complete transition waveform before and after the zero-crossing point is covered as the standard.

4. The track-based electric energy meter based intelligent monitoring method according to claim 1, characterized in that, The distortion risk coefficient generated by intelligently predicting the current waveform distortion of the zero-crossing point through the machine learning model is compared and analyzed with the pre-set distortion risk reference threshold to identify the current waveform distortion in the zero-crossing neighborhood waveform subset, and the specific identification steps are as follows: If the distortion risk coefficient is greater than the distortion risk reference threshold, the zero-crossing neighborhood waveform subset is identified as a current waveform distortion subset; If the distortion risk coefficient is less than or equal to the distortion risk reference threshold, the zero-crossing neighborhood waveform subset is identified as a current waveform normal subset.

5. The track-based electric energy meter based intelligent monitoring method according to claim 4, characterized in that, When the current waveform distortion in the zero-crossing neighborhood waveform subset is identified, the actual zero-crossing judgment threshold is adaptively offset according to the distortion degree of the current waveform, and the specific steps are as follows: After detecting the current waveform distortion in the zero-crossing neighborhood waveform subset, the difference between the distortion risk coefficient and the distortion risk reference threshold is quantified, and the sensitivity adjustment factor is calculated, and the expression is as follows: wherein: is a sensitivity adjustment factor, used to reflect the dynamic influence of actual distortion degree on zero-crossing determination threshold, is a sensitivity adjustment reference coefficient, a positive real number, used to control the adjustment amplitude, is a distortion risk coefficient generated when the machine learning model predicts the current waveform distortion of the zero-crossing point, is a distortion risk reference threshold, is a non-linear amplification exponent, is used to enhance the adjustment response when the distortion degree is high, if then that is, the sensitivity is not adjusted when the distortion does not exceed the reference threshold; After obtaining the sensitivity adjustment factor Then, based on the standard zero-crossing decision threshold, an adaptive offset is performed to generate a new actual zero-crossing decision threshold, and the generated expression is: wherein: is an initial set standard zero-crossing decision threshold, is an actual zero-crossing decision range dynamically adjusted according to current distortion risk.

Images