Damped oscillation wave working condition identification method of intelligent electric energy meter

Abstract

The invention belongs to the technical field of electric power measurement, and discloses a damped oscillation wave working condition identification method of an intelligent electric energy meter, which comprises the following steps: acquiring sampling signals of voltage and current; preprocessing the sampling signals of the voltage and the current to obtain characteristic data; calculating a score corresponding to the voltage sampling signal by combining the dynamic threshold with the feature data; and judging whether the score is greater than a threshold value, if so, marking as an oscillation working condition, if not, judging whether the current sampling signal meets an amplitude change condition and whether the current and voltage sampling signals meet a phase mutation condition, and if so, marking as the oscillation working condition. According to the invention, the recognition accuracy can be improved, and misjudgment is reduced.

Description

Technical Field

[0001] This invention relates to the field of power metering technology, and in particular to a method for identifying damped oscillation wave operating conditions in smart energy meters. Background Technology

[0002] In modern power grids, the widespread use of power electronic equipment has led to the frequent occurrence of damped oscillations, typically with frequencies between 0.1 and 3 kHz and attenuation coefficients between 0.01 and 0.5. These oscillations can cause metering errors in electricity meters, ranging from 5% to 30%. Existing electricity meters mainly rely on voltage sag / boost protection mechanisms, but their ability to identify oscillating transient interferences is insufficient, often misinterpreting them as normal load fluctuations. Although the international standard IEC 61000-4-30 defines relevant test methods for oscillations, it does not provide a real-time identification scheme suitable for embedded terminals.

[0003] There are two main methods for identifying oscillation conditions: one is to use the threshold comparison method, but this method has a high false positive rate of over 40%; the other is to rely on hardware filters, but this method has a response delay of over 20ms and cannot quantify oscillation characteristics. Summary of the Invention

[0004] In order to achieve accurate identification of damped oscillation waves and improve the accuracy of electricity meter measurement, this invention proposes a method for identifying the operating conditions of damped oscillation waves in smart electricity meters.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for identifying the damped oscillation wave operating condition of a smart energy meter, comprising the following steps:

[0006] Step 1: Acquire the sampling signals of voltage and current;

[0007] Step 2: Preprocess the sampled voltage and current signals to obtain feature data;

[0008] Step 3: Calculate the score corresponding to the voltage sampling signal by combining dynamic threshold with feature data; the calculation formula is:

[0009] Score=Σ[w i *M i (X i )];

[0010] Where Score represents the rating, M i (X i X represents the i-th feature data of the voltage sampling signal. i Feature membership function, w i Let be the weight coefficient of the i-th feature;

[0011] Step 4: Determine if the score is greater than the threshold. If it is, mark it as an oscillation condition. If it is not, determine if the current sampling signal meets the amplitude change condition and if the current and voltage sampling signals meet the phase change condition. If both are met, mark it as an oscillation condition.

[0012] Preferably, in step 2, the obtained feature data includes oscillation frequency, attenuation coefficient, attenuation time, and amplitude change rate;

[0013] In step 3, the formula for calculating the score corresponding to the voltage sampling signal is as follows:

[0014] Score = w1F(f osc )+w2 G(α)+w3H(T duration )+w4I(r);

[0015] Where Score represents the rating, w1, w2, w3, and w4 represent the weighting coefficients corresponding to the oscillation frequency, decay coefficient, decay time, and amplitude change rate, respectively, and F(f osc ) represents the oscillation frequency f osc The Gaussian membership function, G(α), represents the Sigmoid function of the attenuation coefficient α, and H(T) represents the Gaussian membership function. duration ) represents the decay time T of the voltage signal. duration The saturated linear function, I(r), represents a piecewise function related to the rate of change of amplitude r; where:

[0016] I(r) = 0, (r < 0.05);

[0017] I(r)=(r-0.05) / 0.25, (0.05≤r≤0.3);

[0018] I(r) = 1 (r > 0.3).

[0019] Preferably, the formula for calculating the amplitude change rate r is: r = ΔU / U base Where ΔU represents the peak-to-peak value of the first oscillation period, U base This represents the fundamental voltage.

[0020] Preferably, step 3 further includes the following steps: determining the current environmental condition type and obtaining the corresponding weight coefficient, and calculating the score corresponding to the voltage sampling signal based on the obtained weight coefficient.

[0021] Preferably, in step 3, the environmental conditions include: residential areas in summer, industrial areas, commercial areas in winter, and new energy access areas.

[0022] Preferably, in step 3, each weight coefficient is obtained through optimization using an LSTM network, specifically by:

[0023] Step 0.1: Obtain voltage and current sampling signals under the same environmental condition type from the power grid historical database;

[0024] Step 0.2: Preprocess the sampled signals of each voltage and current to obtain feature data, and construct a dataset based on the feature data corresponding to each sampled signal of each voltage and current;

[0025] Step 0.3: Input the feature data of the dataset into the LSTM network for training to obtain the weight coefficients corresponding to each feature data.

[0026] Step 0.4: Obtain the voltage and current sampling signals under the next environmental condition from the power grid historical database. Repeat steps 0.2 to 0.3 until the weighting coefficients for all environmental condition types are obtained.

[0027] Preferably, in step 2, the specific steps for preprocessing the voltage and current sampling signals to obtain feature data include:

[0028] Step 2.1: Extract the oscillation frequency f osc ;

[0029] Step 2.2: Extract the envelope of the voltage signal using Hilbert transform, and calculate the attenuation coefficient α by fitting the exponential decay curve through nonlinear regression;

[0030] Step 2.3: Extract the decay time T of the voltage signal duration The decay time is the time it takes for the oscillation amplitude to decay to 10% of the fundamental amplitude;

[0031] Step 2.4: Calculate the rate of change of amplitude r. The formula is: r = ΔU / U base Where ΔU represents the peak-to-peak value of the first oscillation period, U base This represents the fundamental voltage.

[0032] Preferably, in step 2, the fitting formula for the attenuation coefficient is:

[0033] A(t) = A0e -αt cos(2πf osc t);

[0034] Where A(t) represents the envelope of the voltage sampling signal extracted by the Hilbert transform; A0 represents the initial amplitude; α represents the attenuation coefficient; and f osc t represents the oscillation frequency, and t represents time.

[0035] Preferably, in step 4, the amplitude change condition is:

[0036] |ΔI / I base|>δ;

[0037] The phase change condition is:

[0038] ω>φ;

[0039] Where ΔI represents the peak-to-peak value of the current sampling signal, I base φ represents the fundamental current, ω represents the phase difference between the current sampling signal and the voltage sampling signal, δ represents the amplitude change threshold, and φ represents the phase change threshold.

[0040] Preferably, δ = 5% and φ = 15.

[0041] Compared with the prior art, the present invention has the following advantages:

[0042] 1. This invention provides a method for identifying damped oscillation wave operating conditions in smart energy meters. It achieves multi-dimensional feature extraction through joint analysis in the time and frequency domains and fuses key features through dynamic weight decision-making, enabling millisecond-level accurate identification of damped oscillation waves and reducing the false judgment rate to below 1%, which is a significant improvement compared to the false judgment rate of existing technologies (over 30%).

[0043] 2. In this invention, the weight coefficients of feature fusion are adaptively optimized by training an LSTM model with historical data. Therefore, they can be automatically adjusted according to changes in season and load type, further improving the accuracy of identifying oscillating wave conditions.

[0044] 3. This invention requires no new equipment; it only needs to acquire voltage and current signals from the existing energy meter's AD module to achieve identification. It is fully compatible with existing smart meter architectures (such as the STM32F4 series), reducing deployment costs and difficulties. Attached Figure Description

[0045] Figure 1 A flowchart illustrating a method for identifying damped oscillation wave conditions in a smart energy meter, as proposed in Embodiment 1 of this invention.

[0046] Figure 2 This is a schematic diagram of a typical waveform of a damped oscillating wave.

[0047] Figure 3 A block diagram of the dynamic weight adjustment mechanism in this embodiment of the invention. Detailed Implementation

[0048] To make the technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments and accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments; based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] Example 1

[0050] like Figure 1 As shown, Embodiment 1 of the present invention provides a method for identifying the damped oscillation wave operating condition of a smart energy meter, comprising the following steps:

[0051] Step 1: Acquire the voltage and current sampling signals.

[0052] Specifically, data can be acquired from the ADC module of the electricity meter. During acquisition, the ADC module synchronously acquires voltage and current waveforms at a sampling rate of no less than 10kHz. For example... Figure 2 The figure shown is a typical waveform diagram of a damped oscillating wave.

[0053] Step 2: Preprocess the voltage and current sampling signals to obtain feature data.

[0054] In step 2, the specific steps for preprocessing the voltage and current sampling signals to obtain feature data include:

[0055] Step 2.1: Extract the oscillation frequency f osc Specifically, by analyzing the sampled signal using a sliding window FFT, the oscillation frequency f can be extracted. osc .

[0056] Step 2.2: Extract the envelope of the voltage signal using Hilbert transform, and calculate the attenuation coefficient α by fitting the exponential decay curve through nonlinear regression.

[0057] In step 2.2, the fitting formula for the attenuation coefficient is:

[0058] A(t) = A0e -αt cos(2πf osc t); (1)

[0059] Where A(t) represents the envelope of the voltage sampling signal extracted by the Hilbert transform; A0 represents the initial amplitude (amplitude at t=0); α represents the attenuation coefficient, which controls the rate of amplitude attenuation; the larger α is, the faster the amplitude attenuates. -αt f represents the exponential decay term, indicating that the amplitude decreases exponentially with time. osc t represents the oscillation frequency, and t represents time.

[0060] Step 2.3: Extract the decay time T of the voltage signal duration The decay time is the time it takes for the oscillation amplitude to decay to 10% of the fundamental amplitude;

[0061] Step 2.4: Calculate the rate of change of amplitude r. The calculation formula is:

[0062] r=ΔU / U base (2)

[0063] Where ΔU represents the peak-to-peak value of the first oscillation period, U base This represents the fundamental voltage, which is equal to the initial amplitude A0 before decay. Specifically, ΔU = A max -A min A max and A min These represent the maximum and minimum values ​​of the amplitude, respectively.

[0064] Furthermore, this embodiment also includes a step of bandpass filtering the voltage sampling signal, with a filtering frequency of 100Hz-5kHz.

[0065] Step 3: Calculate the score corresponding to the voltage sampling signal; the calculation formula is:

[0066] Score=Σ[w i *M i (X i (3)

[0067] Where Score represents the rating, M i (X i X represents the i-th feature data of the voltage sampling signal. i Feature membership function, w i Let be the weight coefficient of the i-th feature;

[0068] Furthermore, in this embodiment, the calculation formula for the score corresponding to the voltage sampling signal in step 3 is as follows:

[0069] Score = w1 F(f osc )+w2G(α)+w3 H(T duration )+w4 I(r); (4)

[0070] Where Score represents the rating, w1, w2, w3, and w4 represent the weighting coefficients corresponding to the oscillation frequency, decay coefficient, decay time, and amplitude change rate, respectively, and F(f osc G(α) represents the Gaussian membership function for the oscillation frequency, G(α) represents the Sigmoid function for the attenuation coefficient, and H(T) represents the Gaussian membership function for the oscillation frequency. duration ) represents the decay time T of the voltage signal.duration The saturated linear function, I(r), represents a piecewise function related to the rate of change of amplitude r; where:

[0071] I(r) = 0, (r < 0.05);

[0072] I(r)=(r-0.05) / 0.25, (0.05≤r≤0.3); (5)I(r)=1, (r>0.3).

[0073] Specifically, step 3 further includes the following steps: determining the current environmental condition type, determining the weighting coefficient based on the environmental condition type, and then calculating the score corresponding to the voltage sampling signal based on the weighting coefficient.

[0074] In step 3, the environmental conditions include: residential areas in summer, industrial areas, commercial areas in winter, new energy access areas, etc.

[0075] Furthermore, in step 3 of this embodiment, the weight coefficients are obtained through optimization using an LSTM network.

[0076] In this embodiment, the weight coefficients w1, w2, w3, and w4 are dynamically adapted weight coefficients, trained using historical data.

[0077] The LSTM model enables adaptive optimization; therefore, the score in this embodiment can be automatically adjusted according to changes in season and load type, thereby improving the accuracy of operating condition identification.

[0078] In step 3 of this embodiment, each weight coefficient is obtained through LSTM network optimization, referencing... Figure 3 The specific method is as follows:

[0079] Step 0.1: Obtain voltage and current sampling signals under the same environmental condition type from the power grid historical database;

[0080] Step 0.2: Preprocess the sampled signals of each voltage and current to obtain feature data, and construct a dataset based on the feature data corresponding to each sampled signal of each voltage and current;

[0081] Step 0.3: Input the feature data of the dataset into the LSTM network for training to obtain the weight coefficients corresponding to each feature data.

[0082] Step 0.4: Obtain the voltage and current sampling signals under the next environmental condition from the power grid historical database. Repeat steps 0.2 to 0.3 until the weighting coefficients for all environmental condition types are obtained.

[0083] In the historical database, each voltage and current sampling signal is labeled with an environmental tag, corresponding to different operating conditions. By using the environmental tags of the data, historical data under the same environmental conditions can be extracted, and then the corresponding weight coefficients can be obtained through deep learning network optimization.

[0084] Table 1 shows several sets of weight coefficients obtained by training an LSTM network after constructing datasets based on different environmental conditions.

[0085] Table 1 Typical weight allocation strategies

[0086]

[0087] Step 4: Determine if the score is greater than the threshold. If it is, mark it as an oscillation condition. If it is not, determine if the current sampling signal meets the amplitude change condition and if the phase of the current and voltage sampling signals meets the phase change condition. If both conditions are met, mark it as an oscillation condition.

[0088] Specifically, in this embodiment, the scoring threshold is set to 0.85.

[0089] Specifically, in step 4, the amplitude change condition is as follows:

[0090] |ΔI / I base |>δ; (6)

[0091] Specifically, the phase transition condition is as follows:

[0092] ω>φ; (7)

[0093] Where ΔI represents the peak-to-peak value of the current sampling signal, I base φ represents the fundamental current, ω represents the phase difference between the current sampling signal and the voltage sampling signal, δ represents the amplitude change threshold, and φ represents the phase change threshold.

[0094] Furthermore, in this embodiment, the amplitude change threshold is set to δ = 5%, and the phase change threshold is set to φ = 15.

[0095] In this embodiment, when the voltage sampling signal of the voltage channel fails to detect oscillation, the current sampling signal is also combined for judgment. By confirming through dual channels, false judgments can be avoided and the recognition rate can be improved.

[0096] Furthermore, this embodiment also includes the following steps:

[0097] Step 5: For oscillating conditions, freeze the accumulated electricity to improve the accuracy of electricity measurement. For non-oscillating conditions, perform normal measurement and upload the measurement data to the main station.

[0098] Table 2 shows the accuracy and response data of the identification method used in this embodiment. Table 2 illustrates that by integrating four key features and using dynamically adaptive weight coefficients to calculate scores, the present invention can effectively identify various oscillation conditions with an identification rate of over 95% and can reduce the false positive rate to 0.7%, which is a significant improvement compared to the false positive rate of the prior art (over 30%).

[0099] Table 2. Identification accuracy for different oscillation conditions

[0100] Operating conditions Recognition accuracy Response time Motor switching oscillation 98.2％ 8.3ms Capacitor Resonance 97.5％ 9.1ms Normal load fluctuations False positive rate: 0.7% -

[0101] In summary, this invention provides a method for identifying damped oscillation wave conditions in smart energy meters. Based on multi-dimensional feature extraction and dynamic weight decision-making, and through joint time-frequency domain analysis, it can achieve millisecond-level accurate identification of damped oscillation waves.

[0102] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for identifying the damped oscillation wave operating condition of a smart energy meter, characterized in that, Includes the following steps: Step 1: Acquire the sampling signals of voltage and current; Step 2: Preprocess the voltage and current sampling signals to obtain feature data; Step 3: Calculate the score corresponding to the voltage sampling signal by combining dynamic threshold with feature data; The calculation formula is: Score=Σ[w i *M i (X i )]; Where Score represents the rating, M i (X i X represents the i-th feature data of the voltage sampling signal. i Feature membership function, w i Let be the weight coefficient of the i-th feature; Step 4: Determine if the score is greater than the threshold. If it is, mark it as an oscillation condition. If it is not, determine if the current sampling signal meets the amplitude change condition and if the current and voltage sampling signals meet the phase change condition. If both are met, mark it as an oscillation condition.

2. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 1, characterized in that, In step 2, the obtained feature data includes oscillation frequency, attenuation coefficient, attenuation time, and amplitude change rate. In step 3, the formula for calculating the score corresponding to the voltage sampling signal is as follows: Score=w1 F(f osc )+w2 G(α)+w3H(T duration )+w4I(r); Where Score represents the rating, w1, w2, w3, and w4 represent the weighting coefficients corresponding to the oscillation frequency, decay coefficient, decay time, and amplitude change rate, respectively, and F(f osc ) represents the oscillation frequency f osc The Gaussian membership function, G(α), represents the Sigmoid function of the attenuation coefficient α, and H(T) represents the Gaussian membership function. duration ) represents the decay time T of the voltage signal. duration The saturated linear function, I(r), represents a piecewise function related to the rate of change of amplitude r; where: I(r)＝0，(r<0.05)； I(r)＝(r-0.05) / 0.25，(0.05≤r≤0.3)； I(r)＝1，(r>0.3)。 3. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 2, characterized in that, The formula for calculating the rate of change of amplitude r is: r = ΔU / U base Where ΔU represents the peak-to-peak value of the first oscillation period, U base This represents the fundamental voltage.

4. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 1, characterized in that, Step 3 further includes the following steps: determining the current environmental condition type and obtaining the corresponding weight coefficient, and calculating the score corresponding to the voltage sampling signal based on the obtained weight coefficient.

5. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 4, characterized in that, In step 3, the environmental conditions include: residential areas in summer, industrial areas, commercial areas in winter, and new energy access areas.

6. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 2, characterized in that, In step 3, each weight coefficient is obtained through optimization using an LSTM network. The specific method is as follows: Step 0.1: Obtain voltage and current sampling signals under the same environmental condition type from the power grid historical database; Step 0.2: Preprocess the sampled signals of each voltage and current to obtain feature data, and construct a dataset based on the feature data corresponding to each sampled signal of each voltage and current; Step 0.3: Input the feature data of the dataset into the LSTM network for training to obtain the weight coefficients corresponding to each feature data. Step 0.4: Obtain the voltage and current sampling signals under the next environmental condition from the power grid historical database. Repeat steps 0.2 to 0.3 until the weighting coefficients for all environmental condition types are obtained.

7. The method for identifying damped oscillation wave operating conditions of a smart energy meter according to claim 1, characterized in that, In step 2, the specific steps for preprocessing the voltage and current sampling signals to obtain feature data include: Step 2.1: Extract the oscillation frequency f osc ; Step 2.2: Extract the envelope of the voltage signal using Hilbert transform, and calculate the attenuation coefficient α by fitting the exponential decay curve through nonlinear regression; Step 2.3: Extract the decay time T of the voltage signal duration The decay time is the time it takes for the oscillation amplitude to decay to 10% of the fundamental amplitude; Step 2.4: Calculate the rate of change of amplitude r. The formula is: r = ΔU / U base Where ΔU represents the peak-to-peak value of the first oscillation period, U base This represents the fundamental voltage.

8. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 7, characterized in that, In step 2, the fitting formula for the attenuation coefficient is: A(t)=A0e -αt cos(2πf osc t); Where A(t) represents the envelope of the voltage sampling signal extracted by the Hilbert transform; A0 represents the initial amplitude; α represents the attenuation coefficient; and f osc t represents the oscillation frequency, and t represents time.

9. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 1, characterized in that, In step 4, the amplitude change condition is: ∣ΔI / I base ∣>d; The phase change condition is: ω>φ; Where ΔI represents the peak-to-peak value of the current sampling signal, I base φ represents the fundamental current, ω represents the phase difference between the current sampling signal and the voltage sampling signal, δ represents the amplitude change threshold, and φ represents the phase change threshold.

10. The method for identifying the damped oscillation wave operating condition of a smart energy meter according to claim 9, characterized in that, δ = 5%, φ = 15.

Images