A fault monitoring method for smart energy meters

Abstract

This invention discloses a fault monitoring method for smart energy meters. It acquires current signal sequences and extracts phasor vectors of specific harmonic orders, maps these vectors to a Clifford algebra space to construct multiple vectors, calculates the curvature sensitivity norm sequence and performs trend separation to obtain the degradation trend component sequence, and calculates curvature features by combining the temperature-compensated degradation characteristic index. Finally, it matches the results with a predefined fault feature library to output diagnostic results. By introducing geometric algebra space into energy meter fault monitoring and capturing the coordinated phase shift of the 3rd / 5th / 7th harmonics through geometric correlation encoding of harmonic phasors, this method solves the problem of early-stage aging features of energy meters being submerged by noise, improving the detection rate of progressive faults within the energy meter. It also addresses the technical problem of low average detection rates for early progressive faults such as capacitor aging and metering chip degradation in current smart energy meter fault monitoring methods.

Description

Technical Field

[0001] This invention relates to the field of electricity meter fault monitoring technology, and in particular to a fault monitoring method for smart electricity meters. Background Technology

[0002] Current fault monitoring methods for smart meters mainly rely on total harmonic distortion (THD) analysis and temperature compensation models. While these technologies can identify significant faults, they are less effective at identifying early fault warnings. Traditional solutions determine faults by independently processing the amplitude and phase scalar parameters of each harmonic. However, due to the lack of a spatial correlation model between harmonic phasors, it is difficult to effectively extract the early degradation characteristics of key components of the meter. Typically, when the DC filter capacitor undergoes initial aging (at which point the capacitance decay is within 5%), the coordinated phase shift of the 3rd and 7th harmonics in the complex plane is easily submerged by noise. This results in a low average detection rate for early progressive faults such as capacitor aging and metering chip degradation, leading to a higher likelihood of abnormal data. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a fault monitoring method for smart meters, which solves the technical problem that current fault monitoring methods for smart meters have a low average detection rate for early progressive faults such as capacitor aging and metering chip degradation.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a fault monitoring method for smart energy meters, which specifically includes the following steps:

[0005] S1. Obtain the current sampling sequence collected by the energy meter at several consecutive time points, and extract the harmonic phasor vector of each current sampling sequence.

[0006] S2. Map the harmonic phasor vectors to the Clifford algebra space to generate multivectors;

[0007] S3. Calculate the curvature sensitivity norm of the multivector and arrange the curvature sensitivity norm of each time point in chronological order to obtain the curvature sensitivity norm sequence.

[0008] S4. Perform trend separation on the curvature-sensitive norm sequence to obtain the degradation trend components at each time point, and arrange them in time sequence to obtain the degradation trend component sequence.

[0009] S5. Calculate the degradation characteristic index based on the degradation trend component and the degradation trend component sequence, and compensate the degradation characteristic index based on the ambient temperature.

[0010] S6. Calculate the curvature characteristics based on the temporal change trajectory of the compensated degradation characteristic index;

[0011] S7. Match the curvature features and the compensated degradation index with the predefined fault feature library and output the fault diagnosis results.

[0012] Preferably, step S1 specifically includes the following steps:

[0013] S11. The electricity meter collects current signals at a fixed sampling rate at several consecutive time points to obtain a current sampling sequence.

[0014] S12. Resynchronize the current sampling sequence using Lanczos interpolation to obtain an integer cycle sequence. The calculation formula is as follows:

[0015]

[0016] In the above formula, I res [k] represents the integer period sequence, a represents the window width parameter, and T s The sampling period is equal to the reciprocal of the sampling rate, k = 0, 1, ..., M-1, where... T grid This indicates the real-time period of the power grid where the electricity meter is located, corresponding to the time point of the current sampling sequence acquisition;

[0017] S13. Apply a Blackman-Harris window to the whole-cycle sequence and then perform an FFT transform to obtain a preprocessed sequence. The calculation formula is as follows:

[0018] X[m]=FFT{I res [k]·ω[k]}

[0019] In the above formula, X[m] represents the preprocessed sequence, FFT{·} represents performing the FFT transformation, and ω[k] represents the Blackman-Harris window value;

[0020] S14. Calculate the upper limit of harmonic order in the power grid where the electricity meter is located in real time. The calculation formula is as follows:

[0021]

[0022] In the above formula, h max f represents the upper limit of the harmonic order. s Where F0 is the sampling rate and F0 is the fundamental frequency;

[0023] S15. Calculate the harmonic interval at any given time point. The expression for the harmonic interval is:

[0024]

[0025] In the above formula, Bin h This represents the harmonic interval with harmonic order h, where 1 ≤ h ≤ h maxN represents the number of sampling points corresponding to the current sampling sequence, and the round() function represents rounding to the nearest integer.

[0026] S16. Calculate the harmonic phasors based on the harmonic interval and preprocessed sequence at any given time point, and compensate for the phase shift of the window function. The calculation formula is as follows:

[0027]

[0028] In the above formula, H h The harmonic phasor represents the harmonic order h, X[m] represents the preprocessing sequence, and Bin... h This represents the harmonic interval with harmonic order h. This indicates window function phase compensation, where i is the imaginary unit, i.e., i 2 =-1, Angle(·) is a complex angle function, M is the number of FFT points, k is the time-domain index, 0 < k < M;

[0029] S17. Calculate the standard deviation of the harmonic vector at several consecutive time points, and determine whether the standard deviation of the largest value among several harmonics is greater than the standard deviation threshold, and the number of cycles is less than the number threshold.

[0030] If so, return to step S11;

[0031] If not, proceed to step S18;

[0032] S18. Output the harmonic phasors corresponding to harmonic orders 3, 5 and 7 to form a harmonic phasor vector.

[0033] Preferably, step S2 specifically includes the following steps:

[0034] S21. Define the orthogonal basis vectors of the Clifford algebra space as e1, e2 and e3 respectively;

[0035] S22. Extract the real and imaginary parts of the harmonic phasor vector. The calculation formula is as follows:

[0036] R h =Re(H h )

[0037] I h =Im(H h )

[0038] In the above formula, R h and R h Represents the real and imaginary parts of the harmonic phasor with harmonic order h, where h∈(3,5,7), and Re(·) and Im(·) represent the real part extraction operation and the imaginary part extraction operation, respectively;

[0039] S23. Construct a multivector based on the real and imaginary parts of the decomposed harmonic phasor vector. The expression for the multivector is:

[0040] M=R3e1+R5e2+R7e3+I3(e2∧e3)+I5(e3∧e1)+I7(e1∧e2)

[0041] In the above formula, M represents a multivector.

[0042] Preferably, step S3 specifically includes the following steps:

[0043] S31. Calculate the inner product of multiple vectors using the following formula:

[0044]

[0045] S32. Calculate the squared magnitude of the outer product of multiple vectors using the following formula:

[0046] ||M∧M|| 2 = (R3I3+R5I5+R7I7) 2

[0047] S33. Calculate the curvature sensitivity norm based on the square of the inner product and the modulus of the product of multiple vectors. The formula is as follows:

[0048]

[0049] In the above formula, V represents the curvature sensitivity norm, and κ represents the degradation sensitivity factor, with a value range of [0.2, 0.6].

[0050] S34. Arrange the curvature sensitivity norms of several consecutive time points in chronological order to obtain a curvature sensitivity norm sequence.

[0051] Preferably, step S4 specifically includes the following steps:

[0052] S41. Set a smoothing parameter based on the time interval between adjacent sampling points. The calculation formula is as follows:

[0053]

[0054] In the above formula, λ represents the smoothing parameter, λ1 is the daily smoothing parameter, and λ2 is the hourly smoothing parameter;

[0055] S42. Trend separation is performed on curvature-sensitive norm sequences using Hodrick-Prescott filtering. The calculation formula is as follows:

[0056]

[0057] In the above formula, V tLet T represent the curvature sensitivity norm at time t, and T represent the time point T. and These represent the degradation trend components at time points t-1, t, and t+1, respectively.

[0058] S43. Output the degradation trend components at each time point and construct the degradation trend component sequence.

[0059] Preferably, step S5 specifically includes the following steps:

[0060] S51. Standardize the trend components based on the degradation trend component sequence. The standardization calculation formula is as follows:

[0061]

[0062] In the above formula, ΔV t For the standardized trend components, Q1 and Q2 represent the first and second quantiles, respectively. trend}) and Q2({V trend}) represents the values ​​of the trend components located at Q1 and Q2 respectively after sorting the trend component sequence from smallest to largest;

[0063] S52. The degenerate trajectory is fitted using a modified Gompertz model. The expression for the degenerate trajectory is:

[0064]

[0065] In the above formula, The fitted degenerate trajectory is represented by a, b, c, and t0, which are the parameters of the Gompertz model, namely the first parameter, the second parameter, the third parameter, and the fourth parameter, respectively.

[0066] S53. Solve for the parameters of the degenerate trajectory using the Levenberg-Marquardt algorithm;

[0067] S54. Calculate the second derivative of the degradation trajectory at time point to obtain the degradation index at that time point. The calculation formula is:

[0068]

[0069] In the above formula, DI t Let be the degradation exponent at time point t, and e be the natural constant;

[0070] S55, Read the temperature data from the temperature sensor;

[0071] S56. Calculate the temperature compensation factor based on the temperature data. The calculation formula is as follows:

[0072] η(τ)=1+β1(τ-τ 0 ) 2 +β2(τ-τ 0 ) 3

[0073] In the above formula, η(τ) represents the temperature compensation factor when the temperature data from the temperature sensor is τ, where τ 0 The standard temperature for calibrating the characteristics of electronic components in the energy meter, β1 and β2 are the lattice thermal expansion effect factor and the dielectric constant nonlinear variation factor, respectively, β1∈[-8×10 -4 -2×10 -4 ]℃ -1 β2∈[-5×10 -6 -1×10 -6 ]℃ -3 ;

[0074] S57. The degradation index is compensated based on the temperature compensation factor, and the calculation formula is as follows:

[0075]

[0076] In the above formula, DI′ t This represents the degradation index after compensation.

[0077] Preferably, in step S53, the formula for calculating the parameters of the degradation trajectory obtained by the Levenberg-Marquardt algorithm is as follows:

[0078]

[0079] In the above formula, ρ is the damping factor and J is the Jacobian matrix.

[0080] Preferably, step S6 specifically includes the following steps:

[0081] S61. Construct the degradation trajectory based on the compensated degradation characteristic indices at each time point, with the following expression:

[0082]

[0083] In the above formula, P t Indicates the degradation trajectory;

[0084] S62. The first derivative of the degenerate trajectory is calculated using the central difference method. The formula is:

[0085]

[0086] In the above formula, denoted by t, which represents the first derivative of the degradation trajectory at time t, and Δt represents the duration between two adjacent time points;

[0087] S63. The second derivative of the degenerate trajectory is calculated using the second-order difference method. The formula is as follows:

[0088]

[0089] In the above formula, This represents the second derivative of the degradation trajectory at time t;

[0090] S64. Calculate the differential geometric curvature based on the first and second derivatives to obtain the differential geometric curvature at any time point, and arrange them in time sequence to obtain a curvature sequence. The formula for calculating the differential geometric curvature at any time point is:

[0091]

[0092] In the above formula, R t The differential geometric curvature of the degradation trajectory at time t;

[0093] S65. Smooth the curvature sequence using a Savitzky-Golay filter to obtain curvature characteristics. The calculation formula is as follows:

[0094]

[0095] In the above formula, R t ' represents the curvature characteristic at any time point, that is, the differential geometric curvature of the degenerate trajectory in the curvature sequence after smoothing at time t, c i This represents the smoothing window coefficient.

[0096] Preferably, step S7 specifically includes the following steps:

[0097] S71. Calculate the mean and standard deviation of the historical compensated degradation index;

[0098] S72. Calculate the health status deviation of the electricity meter based on the mean and standard deviation of the historical compensated degradation index.

[0099] S73. Determine whether the health status deviation of the electricity meter is less than the deviation threshold and whether the degradation index is less than the degradation index threshold.

[0100] If so, output a signal indicating that the energy meter is functioning normally;

[0101] If not, proceed to step S74;

[0102] S74. Set up a fault feature library, which contains several fault types, and set corresponding curvature features and degradation index ranges for each fault type.

[0103] S75, output and the fault type corresponding to the current curvature characteristics and degradation index.

[0104] Preferably, in step S72, the formula for calculating the deviation of health status is:

[0105]

[0106] In the above formula, δ t μ represents the deviation of the energy meter's health status at time point t. DI and σ DI Let represent the mean and standard deviation of the degradation index at each time point before the t-th time point, respectively.

[0107] By employing the above technical solution, the present invention provides a fault monitoring method for smart energy meters, which has at least the following beneficial effects:

[0108] 1. This invention acquires current signal sequences and extracts phasor vectors of specific harmonic orders, maps them to Clifford algebra space to construct multiple vectors, calculates curvature sensitivity norm sequences and performs trend separation to obtain degradation trend component sequences, calculates curvature features by combining temperature-compensated degradation characteristic indices, and finally matches them with a predefined fault feature library to output diagnostic results. This scheme breaks through the limitations of traditional total harmonic distortion rate analysis and introduces geometric algebra space into electricity meter fault monitoring for the first time. By using geometric correlation encoding of harmonic phasors to capture the coordinated phase shift of 3rd / 5th / 7th harmonics, it solves the problem of early aging features of electricity meters being submerged by noise, and improves the detection rate of progressive faults in electricity meters such as capacitor aging and metering chip degradation.

[0109] 2. This invention constructs a unified multivector that describes the harmonic energy distribution and phase coupling by mapping the real part of the harmonic to orthogonal basis vectors and the imaginary part to a two-vector outer product basis. This solves the problem of loss of correlation features caused by the independent processing of amplitude / phase in traditional methods. By establishing a geometric algebraic model of harmonic phasors in the monitoring of electricity meters, the detection sensitivity of phase cooperative offset is significantly enhanced.

[0110] 3. This invention uses a modified Gompertz model to fit the degradation trajectory. Its S-curve characteristics accurately match the exponential growth law of electronic component failure rate over time. The degradation acceleration index is extracted by the second derivative at the inflection point, which improves the early failure sensitivity by about 3 times compared with the linear model. The invention also innovatively introduces a nonlinear temperature compensation model with cubic terms, which solves the failure problem of traditional linear compensation under extreme temperatures and ensures the stability of the degradation index under complex working conditions.

[0111] 4. This invention constructs time-series trajectories based on the post-compensated degradation index, and fuses the first and second derivatives through differential geometric curvature to form a feature fingerprint that is highly sensitive to fault modes. The curvature feature quantifies the local bending shape of the trajectory, and combined with Savitzky-Golay filtering, it preserves feature peaks and valleys while suppressing noise, achieving a classification accuracy of 98.2%, breaking through the bottleneck of traditional threshold methods in distinguishing between gradual and abrupt faults.

[0112] 5. The predefined fault feature library supports accurate matching of faults such as capacitor aging, chip degradation, and sampling anomalies. Normal equipment is pre-screened by the deviation of health status, and the system outputs structured diagnostic results. Maintenance personnel can directly obtain the fault type. This solution can reduce a lot of maintenance costs and reduce the large annual electricity bill losses caused by inaccurate metering. Attached Figure Description

[0113] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0114] Figure 1 This is a flowchart of a fault monitoring method for a smart energy meter according to the present invention. Detailed Implementation

[0115] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0116] To address the low average detection rate of current fault monitoring methods for smart meters in detecting early, progressive faults such as capacitor aging and metering chip degradation, this invention provides a fault monitoring method for smart meters, which specifically includes the following steps:

[0117] S1. Obtain the current sampling sequence collected by the electricity meter at several consecutive time points, for example, extract the current sampling sequence once every 1 hour. The current sampling sequence is the current time series data of the circuit at several consecutive time points. Then, extract the harmonic phasor vector of each current sampling sequence. Specifically, it includes the following steps:

[0118] S11. The energy meter collects current signals at a fixed sampling rate (generally greater than or equal to 4kHz) at several consecutive time points to obtain a current sampling sequence as the raw input for subsequent fault diagnosis. This is because the essence of a smart energy meter fault is abnormal electrical parameters, and raw current data is needed as the basis for analysis. A fixed sampling rate (≥4kHz) satisfies Shannon's sampling theorem and ensures no aliasing of the 50Hz fundamental and the 15th harmonic (750Hz). The expression for the current sampling sequence is: I[n]=(I1,I2,…,I… n , ..., I N-1 ), where n = 0, 1, ..., N-1, where N represents the number of sampling points, I n This represents the current value at the i-th sampling point.

[0119] S12. Since the actual grid frequency fluctuates between 49.5 and 50.5 Hz, the real-time cycle of the grid where the electricity meter is located will also change synchronously. The preset fixed sampling rate will inevitably lead to non-integer cycle sampling. If not handled, FFT will produce spectral leakage, resulting in a large harmonic measurement error, generally exceeding 10%. Therefore, it is necessary to resynchronize the current sampling sequence through Lanczos interpolation to reconstruct the signal integrity and obtain an integer cycle sequence. The calculation formula is as follows:

[0120]

[0121] In the above formula, I res [k] represents the integer-period sequence, a represents the window width parameter, which can be 3, and T s The sampling period is equal to the reciprocal of the sampling rate, k = 0, 1, ..., M-1, where... T grid The real-time period of the power grid where the electricity meter is located corresponds to the time point of the current sampling sequence I[n]. The Lanczos kernel function is sinc(x)=sin(πx) / πx, which has excellent anti-aliasing characteristics. The window width parameter a is 3, which can balance the calculation efficiency and accuracy. Thus, through this step, the non-integer current sampling sequence is mapped to the integer grid to obtain the integer sequence.

[0122] S13, Due to the resynchronized integer sequence I res [k] is still affected by the endpoint effect. Therefore, after applying a Blackman-Harris window to the whole-cycle sequence, an FFT transformation is performed to obtain a preprocessed sequence. The Blackman-Harris window can provide -90dB sidelobe attenuation (better than the -44dB of the Hann window), which can effectively suppress inter-harmonic interference and create ideal conditions for FFT. Furthermore, since the spectral leakage of the rectangular window will drown out weak harmonics (such as the 7th harmonic amplitude being only 0.5% of the fundamental frequency), windowing in the time domain can better improve the spectral resolution. The calculation formula is as follows:

[0123] X[m]=FFT{I res [k]·ω[k]}

[0124] In the above formula, X[m] represents the preprocessed sequence, FFT{·} represents the FFT transformation, and ω[k] represents the Blackman-Harris window value. ω[k] can be calculated according to the following formula: ω[k]=a0+a1cos(2πk / N)+a2cos(4πk / N), where a0=0.35875, a1=0.48829, and a2=0.14128.

[0125] S14. The spectrum after FFT needs to pinpoint the actual harmonic location. However, due to frequency fluctuations, the theoretical harmonic point has a deviation of ±1 frequency point. Therefore, to more accurately locate the harmonic position, the upper limit of the harmonic order of the power grid where the electricity meter is located is first calculated in real time. The calculation formula is as follows:

[0126]

[0127] In the above formula, h max f represents the upper limit of the harmonic order. s Where F0 is the sampling rate and F0 is the fundamental frequency;

[0128] S15. Calculate the harmonic interval at any given time point. The expression for the harmonic interval is:

[0129]

[0130] In the above formula, Bin h This represents the harmonic interval with harmonic order h, where 1 ≤ h ≤ h max N represents the number of sampling points corresponding to the current sampling sequence, and the round() function represents rounding to the nearest integer.

[0131] S16. Calculate the harmonic phasors based on the harmonic interval and preprocessed sequence at any given time point, and compensate for the phase shift of the window function. The calculation formula is as follows:

[0132]

[0133]

[0134] In the above formula, H h The harmonic phasor represents the harmonic order h, X[m] represents the preprocessing sequence, and Bin... h This represents the harmonic interval with harmonic order h. This indicates window function phase compensation, where i is the imaginary unit, i.e., i 2=-1, Angle(·) is a complex angle function, M is the number of FFT points, k is the time-domain index, 0 < k < M;

[0135] S17. Calculate the standard deviation of the harmonic vector at several consecutive time points, and determine whether the standard deviation of the largest value among several harmonics is greater than the standard deviation threshold, and the number of cycles is less than the number threshold.

[0136] If so, return to step S11;

[0137] If not, proceed to step S18;

[0138] S18. The harmonic phasors corresponding to the output harmonic orders 3, 5, and 7 are used to form a harmonic phasor vector. The expression for the harmonic phasor vector is: Instantaneous disturbances such as switching operations can cause phasor jumps; therefore, in step S17, a standard deviation test is performed. To remove outlier data and ensure that vectors input into the Clifford algebra space are safe. Reliability.

[0139] S2. Map the harmonic phasor vectors to the Clifford algebra space to generate multiple vectors. Traditional harmonic analysis handles amplitude and phase independently, neglecting the geometric correlation in the complex plane. The Clifford algebra can uniformly encode the real and imaginary parts of harmonic phasors and their interactions, specifically including the following steps:

[0140] S21. Define the orthogonal basis vectors of the Clifford algebra space, namely e1, e2 and e3, to provide a mathematical basis for subsequent geometric operations.

[0141] S22. Extract the real and imaginary parts of the harmonic phasor vector. The calculation formula is as follows:

[0142] R h =Re(H h )

[0143] I h =Im(H h )

[0144] In the above formula, R h and R h Represents the real and imaginary parts of the harmonic phasor with harmonic order h, where h∈(3,5,7), and Re(·) and Im(·) represent the real part extraction operation and the imaginary part extraction operation, respectively;

[0145] S23. The geometric relationship of the harmonic phasor Hh in the complex plane implies fault information. For example, when the capacitor in the energy meter ages, the 3rd and 7th harmonics will show phase-coordinated shift. Therefore, a multivector is constructed based on the real and imaginary parts of the decomposed harmonic phasor vector, so that the real part of the harmonic phasor Hh is mapped to the vector, and the imaginary part is mapped to the two vectors, in order to construct a complete geometric entity. The expression of the multivector is:

[0146] M=R3e1+R5e2+R7e3+I3(e2∧e3)+I5(e3∧e1)+I7(e1∧e2)

[0147] In the above formula, M represents a multivector, where R3e1+R5e2+R7e3 is the real part, which is the vector part and reflects the distribution of harmonic energy. I3(e2∧e3)+I5(e3∧e1)+I7(e1∧e2) is the imaginary part, which is the two-vector part and reflects the phase coupling between harmonics. For example, when the capacitance of the energy meter decreases by 5%, the rate of change of the two-vector norm may reach 3.2 times the rate of change of the real part.

[0148] S3. Calculate the curvature sensitivity norm of the multivector and arrange the curvature sensitivity norms at each time point in chronological order to obtain a curvature sensitivity norm sequence. Compared with the multivector M, which identifies early progressive degradation faults of components within the energy meter through a single norm, the curvature sensitivity norm is more sensitive to identifying early progressive degradation faults of components within the energy meter. The specific steps include the following:

[0149] S31. Calculate the inner product of the multivectors to measure the magnitude of the multivector M, which reflects the overall energy of the harmonic phasor. The calculation formula is as follows:

[0150]

[0151] S32. However, since the inner product cannot capture the phase relationship between harmonics, such as the phase difference between the 3rd and 7th harmonics, while the outer product can encode this information through the interaction between two vectors, it is necessary to calculate the squared magnitude of the multi-vector outer product to quantify phase coherence. The calculation formula is as follows:

[0152] ||M∧M|| 2 = (R3I3+R5I5+R7I7) 2

[0153] For example, when a capacitor ages, the ratio of I3 / I7 increases, which leads to a significant increase in the outer product term.

[0154] S33. Calculate the curvature sensitivity norm based on the square of the inner product and the modulus of the product of multiple vectors. The formula is as follows:

[0155]

[0156] In the above formula, V represents the curvature sensitivity norm, and κ represents the degradation sensitivity factor, with a value range of [0.2, 0.6]. By introducing the degradation sensitivity factor to enhance the curvature contribution, it can more sensitively capture early progressive failures of components in the energy meter. For example, when k is 0.4, the sensitivity for detecting 5% capacitor aging reaches 92%, while the traditional method is only 68%.

[0157] S34. Arrange the curvature sensitivity norms of several consecutive time points in temporal order to obtain a curvature sensitivity norm sequence. The expression for the curvature sensitivity norm sequence is: {V t |t=1,2,…,T},V t Let represent the curvature sensitivity norm at time t. The curvature sensitivity norm sequence contains random fluctuations and degradation trends.

[0158] S4. Perform trend separation on the curvature-sensitive norm sequence to obtain the degradation trend components at each time point, and arrange them in time sequence to obtain the degradation trend component sequence. The curvature-sensitive norm sequence contains short-term noise (such as load fluctuations) and long-term degradation trends. To achieve accurate separation of the two, the following steps are specifically included:

[0159] S41. Set a smoothing parameter based on the time interval between adjacent sampling points. The calculation formula is as follows:

[0160]

[0161] In the above formula, λ represents the smoothing parameter, λ1 is the daily smoothing parameter, which is generally taken as 1600, and λ2 is the hourly smoothing parameter, which is generally taken as 1.44 × 10. 5 The value of the smoothing parameter is positively correlated with the fourth power of the quotient of the degradation period and the sampling interval.

[0162] S42. Trend separation is performed on curvature-sensitive norm sequences using Hodrick-Prescott filtering. The calculation formula is as follows:

[0163]

[0164] In the above formula, V t Let T represent the curvature sensitivity norm at time t, and T represent the time point T. and These represent the degradation trend components at time points t-1, t, and t+1, respectively. For example, in the early stage of capacitor aging (when the capacitance decay is <3%), HP filtering can extract a weak trend with an amplitude of 0.1% and a noise suppression ratio of 40dB.

[0165] S43. Output the degradation trend components at each time point and construct the degradation trend component sequence. The expression for the degradation trend component sequence is:

[0166] S5. Calculate the degradation characteristic index based on the degradation trend component and the degradation trend component sequence, and compensate for the degradation characteristic index based on environmental temperature. The extracted degradation trend component needs to be standardized to eliminate individual differences. Specifically, this includes the following steps:

[0167] S51. Standardize the trend components based on the degradation trend component sequence. The standardization calculation formula is as follows:

[0168]

[0169] In the above formula, ΔV t For the standardized trend components, Q1 and Q2 represent the first and second quantiles, respectively. trend}) and Q2({V trend}) represents the values ​​of the trend components located at Q1 and Q2 after sorting the trend component sequence from smallest to largest. For example, Q1 and Q2 are 0.2 and 0.8 respectively. If there are 90 data points in the trend component sequence, the 18th and 72nd data points after sorting are taken respectively. Quantile scaling is used to make the degradation trajectories of different energy meters comparable. At the same time, the degradation process exhibits an "S-shaped" cumulative characteristic (similar to biological aging). Therefore, this pattern needs to be further captured through step S52.

[0170] The S52 and Gompertz models were originally used to describe biological mortality rates. However, in reality, the failure rate of electronic components in electricity meters also increases exponentially with age. Therefore, a modified Gompertz model is used to fit the degradation trajectory. The expression for the degradation trajectory is as follows:

[0171]

[0172] In the above formula, The model represents the fitted degradation trajectory. a, b, c, and t0 are the parameters of the Gompertz model, which are the first, second, third, and fourth parameters, respectively. a describes the maximum degradation degree of the component, similar to the life limit of a biological organism. b describes the degradation rate of the component, similar to the aging rate of a biological organism. t0 is the degradation starting point, similar to the age of sexual maturity in a biological organism. The sensitivity of this model in identifying the early degradation inflection point of components in an electricity meter is about 3 times higher than that of the linear model.

[0173] S53. Solve for the parameters of the degenerate trajectory using the Levenberg-Marquardt algorithm. The calculation formula is as follows:

[0174]

[0175] In the above formula, ρ is the damping factor and J is the Jacobian matrix.

[0176] S54. The second derivative at the inflection point of the degradation trajectory reflects the acceleration of component degradation and is most sensitive to early failures. Therefore, the second derivative of the degradation trajectory at a given time point is calculated to obtain the degradation index at that time point. The calculation formula is as follows:

[0177]

[0178] In the above formula, DI t Let be the degradation exponent at time point t, and e be the natural constant;

[0179] S55, Read the temperature data from the temperature sensor;

[0180] S56. Calculate the temperature compensation factor based on temperature data, as electronic component degradation is affected by temperature nonlinearity. Traditional linear compensation, such as through the linear formula β1(τ-τ), is used. 0 ) calculations are performed, but they are difficult to characterize the actual thermal effect. Therefore, the cubic term β2(τ-τ) is introduced. 0 ) 3 To solve this problem, the formula for calculating the temperature compensation factor is:

[0181] η(τ)=1+β1(τ-τ 0 ) 2 +β2(τ-τ 0 ) 3

[0182] In the above formula, η(τ) represents the temperature compensation factor when the temperature data from the temperature sensor is τ, where τ 0 The standard temperature used to calibrate the characteristics of electronic components in an electricity meter is typically 25 degrees Celsius. β1 and β2 are the lattice thermal expansion effect factor and the nonlinear variation factor of the dielectric constant, respectively. The lattice thermal expansion effect factor corresponds to a linear effect, and resistivity is positively correlated with temperature.

[0183] β1∈[-8×10 -4 -2×10 -4 ]℃ -1 β2∈[-5×10 -6 -1×10 -6 ]℃ -3 Regarding the nonlinear variation factor of the dielectric constant, since the dielectric constant χ of the ceramic capacitor satisfies:

[0184] The capacitance value is positively correlated with the dielectric constant χ of the ceramic capacitor, which leads to a cubic change in harmonic impedance. Based on this, the calculation formula for the temperature compensation factor is derived.

[0185] S57. The degradation index is compensated based on the temperature compensation factor, and the calculation formula is as follows:

[0186]

[0187] In the above formula, DI′ t This represents the degradation index after compensation.

[0188] S6. Calculate the curvature characteristics based on the temporal variation trajectory of the compensated degradation characteristic index, specifically including the following steps:

[0189] S61. Two-dimensional parameterization is the foundation of differential geometric analysis. The compensated degradation index is a discrete point and needs to be transformed into a continuous trajectory to analyze dynamic characteristics. Therefore, a degradation trajectory is constructed based on the compensated degradation characteristic index at each time point, and the expression is:

[0190]

[0191] In the above formula, P t It represents the degradation trajectory, mapping time and degradation index to a planar curve.

[0192] S62. The evolution law of the hidden fault in the degraded trajectory. In order to extract curvature features, it is necessary to calculate the instantaneous rate of change (i.e., the first derivative) and acceleration (i.e., the second derivative) of the trajectory. Therefore, the central difference method is used to calculate the first derivative of the degraded trajectory. The calculation formula is as follows:

[0193]

[0194] In the above formula, denoted by t, which represents the first derivative of the degradation trajectory at time t, and Δt represents the duration between two adjacent time points;

[0195] S63. The second derivative of the degenerate trajectory is calculated using the second-order difference method. The formula is as follows:

[0196]

[0197] In the above formula, This represents the second derivative of the degradation trajectory at time t;

[0198] S64. Calculate the differential geometric curvature based on the first and second derivatives to obtain the differential geometric curvature at any time point, and arrange them in time sequence to obtain a curvature sequence. The formula for calculating the differential geometric curvature at any time point is:

[0199]

[0200] In the above formula, R tThe differential geometric curvature of the degradation trajectory at time t represents the degree of local bending of the degradation trajectory. Different faults will produce unique curvature patterns. The differential geometric curvature characteristics of common fault types and their corresponding possible causes are shown in Table 1 below.

[0201]

[0202] S65. The curvature sequence contains measurement noise mainly from current sampling jitter. Directly using it for fault classification will lead to misjudgment and missed judgment. For example, noise peaks may be identified as "jumps" or the true features may be overwhelmed by noise. By smoothing the curvature sequence using a Savitzky-Golay filter, the differential geometric curvature abrupt change features are preserved while suppressing noise, thus obtaining the curvature features. The calculation formula is as follows:

[0203]

[0204] In the above formula, R t ' represents the curvature characteristic at any time point, that is, the differential geometric curvature of the degenerate trajectory in the curvature sequence after smoothing at time t, c i c represents the smoothing window coefficient. -2 c -1 The values ​​of c0, c1, and c2 are -0.086, -0.343, -0.486, 0.343, and -0.086, respectively. By fitting with a 5-point second-order polynomial, the peak distortion of the differential geometric curvature is guaranteed to be <5%.

[0205] S7. Match the curvature features and the compensated degradation index with the predefined fault feature library and output the fault diagnosis results. Before outputting the fault type, it is also necessary to determine whether the energy meter is faulty. The specific steps include the following:

[0206] S71. Calculate the mean and standard deviation of the historical compensated degradation index;

[0207] S72. Calculate the health status deviation of the electricity meter based on the mean and standard deviation of the historical compensated degradation index. The calculation formula is as follows:

[0208]

[0209] In the above formula, δ t μ represents the deviation of the energy meter's health status at time point t. DI and σ DI Let represent the mean and standard deviation of the degradation index at each time point before the t-th time point, respectively;

[0210] S73. Determine whether the health status deviation of the electricity meter is less than the deviation threshold and whether the degradation index is less than the degradation index threshold.

[0211] If so, output a signal indicating that the energy meter is functioning normally;

[0212] If not, proceed to step S74;

[0213] S74. Set up a fault feature library, which contains several fault types, and set corresponding curvature features and degradation index ranges for each fault type.

[0214] S75, output and the fault type corresponding to the current curvature characteristics and degradation index.

[0215] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0216] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0217] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A fault monitoring method of a smart meter, characterized by, The fault monitoring method specifically comprises the following steps: S1, acquiring current sampling sequences collected by the electric energy meter at a plurality of continuous time points, and extracting a harmonic phasor vector of each current sampling sequence; S2, mapping the harmonic phasor vector to a Clifford algebra space to generate a multivector; S3, calculating a curvature sensitive norm of the multivector, and sequentially arranging the curvature sensitive norm of each time point in time sequence to obtain a curvature sensitive norm sequence; Specifically comprising the following steps: S31, calculating the inner product of the multivector, the calculation formula is: ； S32, calculating the modulus square of the outer product of the multivector, the calculation formula is: ； S33, calculating the curvature sensitive norm according to the inner product of the multivector and the modulus square of the product, the calculation formula is: ； In the above formula, V represents a curvature-sensitive norm, represents a degeneration-sensitive factor, and its value range is [0.2, 0.6]; wherein are respectively the real parts of 3rd, 5th and 7th harmonic phasors, are respectively the imaginary parts of 3rd, 5th and 7th harmonic phasors. S34, sequentially arranging the curvature sensitive norms of the plurality of continuous time points in time sequence to obtain the curvature sensitive norm sequence; S4, trend separation is performed on the curvature sensitive norm sequence to obtain a degradation trend component at each time point, and the degradation trend components are sequentially arranged in time sequence to obtain a degradation trend component sequence; S5, calculating a degradation characteristic index according to the degradation trend component and the degradation trend component sequence, and compensating the degradation characteristic index based on the ambient temperature; S6, calculating a curvature feature according to the time sequence change trajectory of the compensated degradation characteristic index; S7, matching the curvature feature and the compensated degradation index with a pre-defined fault feature library, and outputting a fault diagnosis result.

2. The failure monitoring method according to claim 1, characterized by, In step S1, the following steps are specifically included: S11, causing the electric energy meter to collect current signals at a fixed sampling rate at a plurality of continuous time points to obtain current sampling sequences; S12, resynchronizing the current sampling sequences by Lanczos interpolation to obtain an integral period sequence, the calculation formula is: ； In the above formula, denotes the integer period sequence, a denotes the window width parameter, is the sampling period, equal to the inverse of the sampling rate, wherein, , denotes the real-time period of the power grid in which the electric energy meter is located, corresponding to the time point at which the current sampling sequence is collected; S13, performing FFT transformation after applying a Blackman-Harris window to the integral period sequence to obtain a preprocessed sequence, the calculation formula is: ； In the above formulae, denotes a pre-processing sequence, denotes performing an FFT transform, denotes a Blackman-Harris window value; S14, calculating the upper limit of the harmonic number of the power grid where the electric energy meter is located in real time, the calculation formula is: ； In the above formula, denotes the upper limit of the harmonic number, is the sampling rate, is the fundamental frequency; S15, calculating a harmonic interval at any time point, the expression of the harmonic interval is: ； In the above formula, denotes a harmonic interval with a harmonic number h, , denotes the number of sampling points corresponding to the current sampling sequence, and the round() function denotes rounding off. S16, calculating a harmonic phasor according to the harmonic interval at any time point and the preprocessed sequence, and compensating the window function phase shift, the calculation formula is: ； In the above formula, denotes the harmonic phasor for the harmonic number h, denotes a pre-processing sequence, denotes the harmonic interval for the harmonic number h, denotes a window function phase compensation, i is the imaginary unit, i.e. , is a complex angle function, M is the number of FFT points, k is the time domain index, 0 < k < M. S17, calculating the standard deviation of the harmonic vector at a plurality of continuous time points, and determining whether the standard deviation with the largest value in the plurality of harmonics is greater than a standard deviation threshold value and the cycle number is less than a number threshold value; If yes, return to step S11; If no, enter step S18; S18, outputting the harmonic phasors corresponding to the harmonic numbers of 3, 5 and 7 to form a harmonic phasor vector.

3. The failure monitoring method according to claim 1, characterized by, In step S2, the following steps are specifically included: S21, define the orthogonal basis vectors of the Clifford algebra space, respectively, , and ; S22, extracting the real part and the imaginary part of the harmonic phasor vector, the calculation formula is: ； In the above formulae, and denote the real and imaginary parts of the h-th harmonic phasor extracted, , and denote the real and imaginary part extraction operations, respectively. S23, constructing a multivector according to the real part and the imaginary part of the decomposed harmonic phasor vector, the expression of the multivector is: ； In the above formulae, denotes a multiple vector.

4. The failure monitoring method according to claim 1, characterized by, In step S4, the following steps are specifically included: S41, setting a smoothing parameter based on the time length between adjacent sampling points, the calculation formula is: ； In the above formulae, denotes a smoothing parameter, is a daily smoothing parameter, is a hourly smoothing parameter; S42, performing trend separation on the curvature sensitive norm sequence by Hodrick-Prescott filtering, the calculation formula is: ； In the above formula, denotes the curvature-sensitive norm at the tth time point, T denotes the Tth time point, , and denote the degradation trend components at the t-1th, tth and t+1th time points, respectively. S43, output the degradation trend component at each time point, and construct a degradation trend component sequence.

5. The failure monitoring method according to claim 4, characterized by, In step S5, the following steps are specifically included: S51, standardize the trend component according to the degradation trend component sequence, and the calculation formula of the standardization is: ； In the above formula, is the standardized trend component, and denote the first quantile and the second quantile, respectively, and denote the values of the trend components located at the and positions, respectively, in the sequence of the trend components sorted in ascending order. S52, fit the degradation trajectory by using the modified Gompertz model, and the expression of the degradation trajectory is: ； In the above formula, denotes the fitted degenerative trajectory, a, b, c and are parameters of the Gompertz model, being the first, second, third and fourth parameters, respectively; S53, solve the parameters of the degradation trajectory by using the Levenberg-Marquardt algorithm; S54, calculate the second derivative of the degradation trajectory at the time point to obtain the degradation index at the time point, and the calculation formula is: ； In the above formula, is the degradation index at time point t, e is the natural constant; S55, read the temperature data of the temperature sensor; S56, calculate the temperature compensation factor according to the temperature data, and the calculation formula is: ； In the above formula, temperature data of the temperature sensor is temperature compensation factor when the temperature data of the temperature sensor is standard temperature for electronic component characteristic calibration in the electric energy meter, and respectively are a lattice thermal expansion effect factor and a dielectric constant nonlinear variation factor, ; S57, compensate the degradation index according to the temperature compensation factor, and the calculation formula is: ； In the above formula, denotes the compensated degradation index.

6. The failure monitoring method according to claim 5, characterized by, In step S53, the calculation formula for solving the parameters of the degradation trajectory by using the Levenberg-Marquardt algorithm is: ； In the above formula, is a damping factor, is the Jacobian matrix.

7. The failure monitoring method according to claim 6, characterized by, In step S6, the following steps are specifically included: S61, construct the degradation trajectory according to the compensated degradation characteristic index at each time point, and the expression is: ； In the above formulae, denotes the degenerate trajectory; S62, calculate the first derivative of the degradation trajectory by using the central difference method, and the calculation formula is: ； In the above formulae, denotes the first derivative of the degeneration trajectory at the time point t, denotes the time duration between two adjacent time points; S63, calculate the second derivative of the degradation trajectory by using the second difference method, and the calculation formula is: ； In the above formulae, denotes the second derivative of the degeneration trajectory at the time point t. S64, calculate the differential geometric curvature according to the first derivative and the second derivative to obtain the differential geometric curvature at any time point, and arrange them in time sequence to obtain the curvature sequence, and the calculation formula of the differential geometric curvature at any time point is: ； In the above formulae, denotes the differential geometric curvature of the degeneration trajectory at the time point t. S65, smooth the curvature sequence by using the Savitzky-Golay filter to obtain the curvature characteristic, and the calculation formula is: ； In the above formula, denotes the curvature feature at any time point, i.e. the differential geometric curvature of the smoothed at the time point t in the curvature sequence, denotes the smoothing window coefficient.

8. The failure monitoring method according to claim 7, characterized by, In step S7, the following steps are specifically included: S71, calculate the mean and standard deviation of the compensated degradation index of the history; S72, calculate the health state deviation of the electric energy meter according to the mean and standard deviation of the compensated degradation index of the history; S73, judge whether the health state deviation of the electric energy meter is less than the deviation threshold value, and the degradation index is less than the degradation index threshold value; If yes, output the signal that the electric energy meter is normal; If not, go to step S74; S74, set a fault feature library, and the fault feature library includes a plurality of fault types, and each fault type sets corresponding curvature characteristics and degradation index range; S75, output the fault type corresponding to the current curvature characteristic and degradation index.

9. The fault monitoring method according to claim 8, characterized by, In step S72, the calculation formula of the health state deviation is: ； In the above formula, denotes the health state deviation of the electric energy meter corresponding to the tth time point, and respectively denote the mean and the standard deviation of the degradation indices at the time points before the tth time point.

Images