Noise reduction method for metering signal of direct-current electric energy meter
By combining adaptive variational modal decomposition and wavelet threshold denoising technology, the filtering parameters of the DC energy meter meter meter signal are optimized, and the problem of poor filtering of the DC energy meter meter signal in the existing technology is solved, and more efficient noise is achieved Removal and metering accuracy.
Patent Information
- Application Number
- CN202410940676.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The parameter settings of existing DC energy meter metering signal filtering and noise reduction methods lack optimization, resulting in poor filtering effect and affecting the measurement accuracy.
Combining adaptive variational modal decomposition (AVMD) and wavelet threshold denoising (AWTD), by establishing a local signal envelope entropy optimization model and mutual information analysis, the effective component and noise component are distinguished, and the signal-to-noise ratio and root mean square Error, smoothness, and correlation coefficient are used as evaluation indicators to optimize the wavelet threshold denoising parameters to achieve optimal filtering and noise reduction of noise components.
It significantly improves the filtering and noise reduction effect of the DC energy meter metering signal, ensures the accuracy of the metering and signal-to-noise ratio, and reduces the root mean square error.
Smart Images

Figure CN118939937A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of signal noise reduction, and in particular to a DC electric energy meter measurement signal noise reduction method based on the combination of adaptive variational mode decomposition and wavelet threshold denoising. Background Art
[0002] At present, large-scale distributed new energy is connected to the distribution network, and the pace of terminal power replacement is accelerating. Among them, distributed photovoltaics and electric vehicles, as typical distributed new energy elements, have grown rapidly in recent years. Compared with AC distribution networks, DC distribution networks have the characteristics of high power supply capacity, low equipment investment loss, and flexible control methods. They are the distribution network form for the safe consumption of large-scale distributed new energy in the future. In this context, electric energy metering has changed from traditional AC form to DC form.
[0003] At present, there are three main DC energy metering algorithms, namely, the average value method, the effective value method, and the time domain integration method. However, most of the current distributed new energy equipment is connected to the grid through power electronic devices such as DC / DC converters, and has significant nonlinear characteristics, resulting in a large amount of ripple in the DC output voltage and current signals, that is, containing high-order AC harmonic components. In this case, the average value method and the effective value method will have measurement deviations, and the greater the ripple content, the greater the measurement error, resulting in inaccurate measurement, affecting the fairness and rationality of electricity billing. The time domain integration method can reduce the impact of ripple on measurement by increasing the sampling frequency. For example, a "DC electric energy meter and electric energy measurement method" disclosed in Chinese patent documents, with the announcement number CN117388570A, uses a DC isolation capacitor to isolate the first voltage signal and the first current signal of the DC, and the corresponding AC voltage signal and AC current signal can be obtained. In this way, the product of the first voltage signal and the first current signal is subtracted from the product of the first voltage signal and the AC current signal, and the product of the AC voltage signal and the first current signal, and then the product of the AC voltage signal and the AC current signal is added to obtain the instantaneous power signal of the detection object. Finally, the instantaneous power signal is integrated using the integration module to obtain the cumulative power consumption of the detection object. However, the computing performance of the metering chip limits the increase in the sampling frequency. Therefore, it is of great significance to carry out high-quality noise reduction research on DC electric energy metering data.
[0004] Filtering setting is an effective way to achieve noise reduction of DC power metering data. Generally, the ripple content of the output signal can be reduced from the hardware level by adding a filter circuit. However, the size of the inductor and capacitor in the filter circuit will affect the filtering effect. When the inductor and capacitor are small, the filtering effect is poor, the output signal ripple content is high, and the measurement accuracy is affected; when the inductor and capacitor are large, the filtering effect is good, but it will cause the output signal to be distorted, affecting the measurement accuracy. Therefore, digital filtering came into being. Classic digital filtering technology includes classic filtering methods such as median filtering, limiting filtering, and sliding average method, but it is not effective in dealing with environmental conditions such as rapid changes and multiple interferences. Compared with the classic filtering method, the overall performance of modern filtering methods has been significantly improved, including wavelet denoising, wavelet threshold denoising (WTD), and classical mode decomposition (EMD). Summary of the invention
[0005] The present invention mainly solves the problems that the parameter setting of the prior art method for filtering and denoising the voltage and current signals at the output end of a DC power load lacks optimization and the filtering and denoising effect needs to be improved; a DC electric energy meter measurement signal denoising method is provided, which combines adaptive variational mode decomposition with wavelet threshold denoising (AVMD-AWTD) to further improve the filtering and denoising effect of the DC electric energy meter measurement signal, thereby ensuring the accuracy of DC electric energy measurement.
[0006] The above technical problems of the present invention are mainly solved by the following technical solutions: A method for reducing noise of a DC electric energy meter measurement signal comprises the following steps: S1: Establish an adaptive variational mode decomposition parameter optimization model for the output DC voltage / current signal with the goal of minimizing the local signal envelope entropy; S2: Calculate and solve the model, and distinguish the effective component from the noise component according to the maximum mutual information coefficient between each modal component and the original sampling data sequence obtained by the solution; S3: For the noise component, an adaptive wavelet threshold denoising parameter optimization model is established with composite index evaluation to filter and reduce the noise component; S4: The improved harmony search algorithm is used to solve the adaptive variational mode decomposition parameter optimization model and the adaptive wavelet threshold denoising parameter optimization model, and the optimal parameter combination scheme is obtained by solving the model.
[0007] The AVMD parameter optimization model of the DC voltage / current signal at the output end is established with the goal of minimizing the local signal envelope entropy, and the model is solved to achieve the optimal decomposition of each IMF. On this basis, the effective component and the noise component are reliably distinguished through the maximum mutual information coefficient of each IMF and the original sampling data sequence. For the noise component, WTD is used for filtering and denoising. In order to ensure the WTD denoising effect, an AWTD parameter optimization model is established with the composite index evaluation of signal-to-noise ratio, root mean square error, smoothness, and correlation coefficient, considering the wavelet basis function, decomposition layer number, threshold function, threshold criterion, etc. The model is solved to obtain the optimal parameter combination scheme, thereby achieving the optimal filtering and denoising of the noise component.
[0008] Preferably, the modal components after modal decomposition of the original sampling data sequence of the DC voltage / current at the output end are subjected to Hilbert transformation to obtain a new modal component sequence; Dividing each new modal component sequence by the sum of all new modal component sequences obtains a proportional coefficient of each data point in the new modal component sequence after Hilbert transformation; The signal envelope entropy value of the modal component of the original sampling data sequence of the DC voltage / current at the output end after the variational modal decomposition is calculated according to the proportional coefficient of all data points.
[0009] Preferably, the constraints of the adaptive variational modal decomposition parameter optimization model are set, and the quadratic penalty factor and the number of modal decompositions in the model need to satisfy their respective upper and lower limit constraints.
[0010] The quadratic penalty factor is a positive real number greater than or equal to 0 and less than positive infinity.
[0011] Preferably, when the maximum mutual information coefficient is greater than or equal to a maximum mutual information coefficient threshold, the corresponding modal component is a valid component; when the maximum mutual information coefficient is less than the maximum mutual information coefficient threshold, the corresponding modal component is a noise component.
[0012] As a preference, the maximum mutual information coefficient γ MIC (X i ; O) The calculation expression is: Where N is the number of sample points of each decomposed modal sequence and the original signal; p(x i , o) is the joint probability density of the i-th decomposition mode sequence and the original signal; p(x i ) is the marginal probability density of the i-th decomposition modal sequence; p(o) is the marginal probability density of the i-th original signal; The grid method is used to calculate the marginal probability density and joint probability density of each decomposed modal sequence and the original signal; a and b are the number of rows and columns of the grid, respectively.
[0013] Preferably, the adaptive wavelet threshold denoising parameter optimization model uses wavelet basis function, decomposition layer number, threshold function and threshold criterion as decision variables; the composite index evaluation is obtained by weighted summation of signal-to-noise ratio, root mean square error, smoothness and correlation coefficient.
[0014] For the noise component, WTD is used for filtering and denoising. To ensure the WTD denoising effect, an AWTD parameter optimization model is established with the composite index of signal-to-noise ratio, root mean square error, smoothness, and correlation coefficient as the best evaluation, considering wavelet basis function, decomposition layer number, threshold function, threshold criterion, etc., and the model is solved to obtain the optimal parameter combination scheme, thereby achieving the optimal filtering and denoising of the noise component.
[0015] Preferably, the constraint conditions for setting the adaptive wavelet threshold denoising parameter optimization model include: The small basis wave function is in the set of small basis wave functions; The number of decomposition levels is between the upper and lower limits of the set decomposition levels; The threshold function is in the threshold function set; Threshold criteria are in the threshold criteria set.
[0016] Preferably, the improved harmony search algorithm comprises: Establishing an initial harmony memory library, wherein each harmony group in the initial harmony memory library includes an initial decomposition parameter group and its corresponding local signal envelope entropy value; According to a preset dynamic adjustment strategy, the harmony in the initial harmony memory library is updated; Iterative optimization is performed to determine whether a preset convergence condition group is satisfied, and when the preset convergence condition group is satisfied, the quadratic penalty factor and modal decomposition number corresponding to the optimal harmony are output, that is, the optimal decomposition parameter group is output.
[0017] Preferably, the dynamic adjustment strategy includes: Generate a random number. If the random number is greater than the probability considered by the harmony memory library, regenerate a decomposition parameter group that meets the preset conditions from the initial harmony memory library, obtain each decomposition mode through variational mode decomposition, calculate the local envelope entropy of the corresponding decomposition mode, and use it as the first harmony; otherwise, select a harmony from the initial harmony memory library as the first harmony; If the first harmony comes from the initial harmony memory library, a random number is regenerated. If the random number is less than the pitch adjustment probability, the secondary penalty factor and the modal decomposition number in the first harmony are adjusted according to the pitch adjustment bandwidth. Otherwise, no adjustment is made to form a second harmony. According to the quadratic penalty factor of the second harmony and the modal decomposition number, each decomposition mode is obtained through variational modal decomposition; The local envelope entropy of the corresponding decomposition mode is calculated to obtain the objective function value corresponding to the second harmony. If the corresponding objective function value is better than the worst harmony in the initial harmony memory bank, the second harmony replaces the worst harmony to update the initial harmony memory bank.
[0018] Preferably, the expression for adjusting the secondary penalty factor and the modal decomposition number in the first harmony according to the pitch adjustment bandwidth is: K2=K1+bw(g) α2=α1+bw(g) In the formula, PAR min and PAR max Dynamically adjust the lower and upper limits for the pitch adjustment probability, respectively; bw min and bw max Dynamically adjust lower and upper limits for the tone adjustment bandwidth, respectively; g is the current iteration number; g max is the maximum number of iterations; K1 and K2 are the modal decomposition numbers of the first and second harmony respectively; α1 and α2 are the quadratic penalty factors for the first and second harmony, respectively.
[0019] The beneficial effects of the present invention are: 1. Establish a parameter optimization model for the variational modal decomposition of the DC voltage / current signal at the output end, and combine it with mutual information analysis to achieve adaptive distinction between the effective modal components and noise modal components of the original signal.
[0020] 2. Establish a wavelet threshold denoising parameter optimization model with optimal composite evaluation indicators such as signal-to-noise ratio, root mean square error, smoothness, and correlation coefficient to achieve optimal filtering and denoising of noise modal components, thereby ensuring the accuracy of DC power metering. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 This is the original signal diagram of the DC voltage at the output end of the present invention.
[0022] Figure 2 This is the original signal diagram of the DC current at the output end of the present invention.
[0023] Figure 3 It is a sequence diagram of each mode after the DC voltage original signal AVMD of the present invention.
[0024] Figure 4It is a sequence diagram of each mode of the DC current original signal AVMD of the present invention.
[0025] Figure 5 It is a comparison chart of data before and after the DC voltage noise reduction of the present invention.
[0026] Figure 6 It is a comparison chart of data before and after the direct current noise reduction of the present invention. DETAILED DESCRIPTION
[0027] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.
[0028] Example: A DC electric energy meter measurement signal denoising method of the present embodiment is a DC electric energy meter measurement signal denoising method based on the combination of adaptive variational mode decomposition and wavelet threshold denoising (AVMD-AWTD). First, an adaptive variational mode decomposition (AVMD) parameter optimization model of the DC voltage and current signals at the output end with the goal of minimizing the local signal envelope entropy is established, and the model is solved to achieve the optimal decomposition of each intrinsic mode function (IMF). On this basis, the effective component and the noise component are reliably distinguished through the maximum mutual information coefficient between each IMF and the original sampling data sequence. For the noise component, the wavelet threshold denoising (WTD) method is used for filtering and denoising. In order to ensure the WTD denoising effect, an AWTD parameter optimization model is established with the composite index of signal-to-noise ratio, root mean square error, smoothness, and correlation coefficient as the best evaluation, considering the wavelet basis function, decomposition layer number, threshold function, threshold criterion, etc., and the model is solved to obtain the optimal parameter combination scheme, thereby achieving the optimal filtering and denoising of the noise component. Finally, the effectiveness of the proposed method is verified through example simulation analysis.
[0029] In this embodiment, before establishing the AVMD parameter optimization model, the relevant principles of DC power measurement are analyzed. In this embodiment, three mainstream methods of DC power measurement are used, namely, the average value method, the effective value method, and the time domain integration method. The three DC power measurement methods are as follows: 1) The expression for measuring DC electric energy using the average value method is: In the formula, m is the electric energy calculated using the average value method; is the average value of the voltage within the Δt time; is the average value of the current within the Δt time; P1 is the average value of the power within the Δt time.
[0030] 2) The expression for measuring DC electric energy using the effective value method is: W2=P2·Δt=UI·Δt Where, W2 is the electric energy value calculated using the effective value method; U is the effective value of the voltage within the time Δt; I is the effective value of the current within the Δt time; P2 is the effective value of the power within the Δt time.
[0031] 3) The time domain integration method measures DC power. The time domain integration method is also called the instantaneous power integration method. Its mathematical expression is: Where W3 is the integrated electric energy value of instantaneous voltage and current within the unit period T; u(t) is the instantaneous voltage; i(t) is the instantaneous current; P3 is the instantaneous power.
[0032] In the actual sampling process, instantaneous power is not sampled continuously. There is a sampling interval. The sampling interval is set to Δt. Sampling is performed within the unit sampling period T. The number of samplings is set to N. The instantaneous power calculated by the time domain integration method can be expressed as: The expression of electric energy in unit sampling period is: When the value of Δt is relatively small and the value of N is relatively large, the electric energy measured by the time domain integration method can be expressed by the above formula.
[0033] During the rectification and filtering process, the inductance L and capacitance C of the filter circuit will affect the filtering effect. At the same time, due to the impedance characteristics and high-power input and output during the charging process of the electric vehicle battery, ripples will appear in the DC output voltage signal.
[0034] The ripple content in the DC output voltage can be expressed by the ripple coefficient, and the ripple coefficient is divided into two definitions, one is the ripple effective value coefficient, and the other is the ripple peak coefficient. Both are the ratios of the effective value and peak value of the AC voltage to the average value of the DC voltage.
[0035] The expression of ripple effective value coefficient is: In the formula, K rms is the ripple effective value coefficient; U M is the effective value of AC voltage; U OUTis the average value of AC output.
[0036] The expression for the ripple peak factor is: Where, X PV is the ripple peak factor; U PV is the peak value of the AC voltage.
[0037] The energy metering error model analysis under the influence of ripple is established to analyze the error caused by ripple on the metering. The energy metering error model analysis under the influence of ripple includes the following steps: From the formulas of the three DC power measurement methods, namely the average value method, the effective value method and the time domain integration method, it can be seen that the DC power in an ideal situation can be obtained by multiplying the instantaneous values of the DC voltage and current. For DC power with ripple components, the concept of AC measurement power needs to be cited. In the actual charging process, the ripple is mainly composed of 6k±1 harmonics, of which the main components are 11th, 13th and 15th harmonics, which are determined by the characteristics of the three-phase full bridge during the rectification process. Assuming that the ripple signal does not contain other components, only the 11th, 13th and 15th harmonic components (the fundamental frequency is the industrial frequency ω), the DC charging voltage expression of the electric vehicle is: Among them, U0 is the ideal DC charging voltage; U 11 , U 13 and U 15 The voltages of the 11th, 13th and 15th harmonic components respectively; α 11 , α 13 and α 15 They are the initial phases of the 11th, 13th and 15th voltage harmonic components respectively.
[0038] The DC charging current expression of electric vehicles is: Where, I0 is the ideal DC charging current; I 11 ,I 13 and I 15 The currents of the 11th, 13th and 15th harmonic components respectively; β 11 , β 13 and β 15 They are the initial phases of the 11th, 13th and 15th current harmonic components respectively.
[0039] When the effective value method is used to calculate the apparent power of electric energy, the expression of the effective value of the output voltage is: The mathematical expression of the effective value of the output current is: Calculate the apparent power S of electric energy, which includes DC power and other harmonic distortion powers. The expression is: Where D I is the current distortion power, which is specifically expressed as: D U is the voltage distortion power, which is specifically expressed as: D H is the apparent power of the harmonic signal, and the specific expression is: S k is the DC component power.
[0040] From the expression of apparent power S, it can be seen that when measuring the output power of a DC charging pile, the power measured by the effective value method consists of DC component power, voltage and current distortion power, and harmonic signal power. The output power measured by the time domain integration method is: Where P0 is the DC active power; P 11 , P 13 and P 15 They are the 11th, 13th and 15th harmonic active powers respectively.
[0041] In order to facilitate the modeling and calculation of measurement errors, the specific impact of ripple on the error is intuitively displayed, the formula is simplified, and the order of harmonics is represented by h.
[0042] According to the concept of AC power, the active power P of the DC power used to charge electric vehicles under the influence of ripple is W for: P W =U0I0+U h I h cos(α h -β h ) Reactive power Q of DC power used to charge electric vehicles under ripple w for: Q w =U h I h sin(α h -β h ) Apparent power S of DC power used to charge electric vehicles under ripple W for: S w =U OM I OM Where, U0 is the ideal DC charging voltage component; I0 is the ideal DC charging current component; U OM is the effective value of voltage; I OM is the effective value of current; U h is the voltage effective value of the hth harmonic component; I h is the effective value of the current of the hth harmonic component; α h is the initial phase of the hth voltage harmonic component; β h is the initial phase of the hth harmonic component of current.
[0043] The calculation expressions for the effective values of the output voltage and current can be simplified to: The calculation expression of the effective value of the output voltage can be simplified as follows: The calculation expression of the effective value of the output current can be simplified as follows: The average power P1 can be expressed as: The effective power P2 can be expressed as: P2=S W =U OM I OM The instantaneous power P3 can be expressed as: After comparison, it can be known that P2 obtained by the effective value method can be used as the AC "apparent" power for calculation. Among them, P3 obtained by the time domain integration method can be used as the AC "active" power for calculation. In the calculation process of AC power, it is stipulated that the power solved by the time domain integration method can be used as the standard power.
[0044] Therefore, when using the above average value method, effective value method, and time domain integration method to analyze and solve the DC output power measurement error, P3 obtained by the time domain integration method can be set as the standard theoretical power, and the error sizes corresponding to P1, P2, and P3 can be solved respectively. The DC power measurement error solved by the voltage and current average value method can be expressed as ε 31 It means that it can be regarded as the error value of P1 relative to P3: The DC energy measurement error solved by the effective value method can be expressed as ε 31 It means that it can be regarded as the error value of power P2 relative to P3: Assume that in the DC output signal, the proportion of the ripple in the voltage and current signals relative to the DC effective value is ρ uh and ρ ih , the harmonic phase difference is
[0045] The ratio of ripple to DC effective value in voltage signal ρ uh The expression is: The ratio of ripple to the DC effective value in the current signal ρ ih The expression is: The harmonic phase difference is The expression is: The relative error of the two energy measurement algorithms can be expressed as: According to the inequality: It can be seen that only ρ uh =ρ ih When the DC power error ε is solved by the average value method 31 The DC energy error ε obtained by using the effective value method 32 Reached maximum value.
[0046] Specifically, a DC power meter measurement signal noise reduction method of this embodiment includes the following steps: S1: establishing an output DC voltage and current signal AVMD parameter optimization model with the goal of minimizing the local signal envelope entropy, and solving the model to achieve the optimal decomposition of each IMF.
[0047] Based on the three common mainstream methods and the error caused by Wenbo measurement, an AVMD parameter optimization model of the DC voltage and current signals at the output end is established with the goal of minimizing the local signal envelope entropy, and the model is solved to achieve the optimal decomposition of each IMF.
[0048] Establish the objective function of the AVMD parameter optimization model.
[0049] Define the minimum signal envelope entropy of all IMFs after VMD of the original DC voltage and current signals at the output as the local signal envelope entropy. The objective function of the AVMD parameter optimization model is to minimize the local signal envelope entropy of the original VMD signals of the DC voltage and current at the output, as shown in the following formula: Among them, E i is the signal envelope entropy value of the ith modal component after the variational modal decomposition of the original sampling data sequence of the DC voltage / current at the output end. The specific expression is: Among them, p i,t is the proportional coefficient of the tth data point in the new sequence after the i-th modal component is Hilbert transformed. The specific expression is: Among them, y i (t) is the new sequence of the i-th modal component after Hilbert transformation. The specific expression is: In the formula, x i (τ) is the i-th modal component sequence; T is the total number of points in the new sequence formed by each modal component after Hilbert transformation; K is the modal decomposition number.
[0050] Set the constraints for the AVMD parameter optimization model.
[0051] The quadratic penalty factor and modal decomposition number in the model are variables to be optimized and need to satisfy their respective upper and lower limit constraints: K min ≤K≤K max , K∈N + 0≤α<+∞,α∈R + Where: α is the quadratic penalty factor; K is the modal decomposition number; K min and K max are the lower and upper limits of the modal decomposition number respectively; N + is a set of positive integers; R + is the set of positive real numbers.
[0052] S2: Reliable distinction between effective components and noise components is achieved through the maximum mutual information coefficient between each IMF and the original sampling data sequence.
[0053] The solution of this embodiment determines and distinguishes the effective component from the noise component by calculating the maximum mutual information coefficient between each modal component and the original signal. The maximum mutual information coefficient calculation formula is as follows: Where N is the number of sample points of each decomposed modal sequence and the original signal; p(x i , o) is the joint probability density of the i-th decomposition mode sequence and the original signal; p(x i ) is the marginal probability density of the i-th decomposition modal sequence; p(o) is the marginal probability density of the i-th original signal.
[0054] This embodiment uses a grid method to calculate the marginal probability density and the joint probability density of each decomposed modal sequence and the original signal, where a and b are the number of rows and columns of the grid, respectively.
[0055] According to the preset maximum mutual information coefficient threshold, the components after adaptive variational mode decomposition of the original signal are divided into effective components and noise components.
[0056] When the maximum mutual information coefficient γ MIC (X i ; O) greater than or equal to the maximum mutual information coefficient threshold γ o When , the i-th modal component is the effective component; When the maximum mutual information coefficient γ MIC (X i ; O) is less than the maximum mutual information coefficient threshold γ o When , the i-th modal component is the noise component.
[0057] S3: For the noise component, WTD is used for filtering and noise reduction.
[0058] In order to ensure the WTD noise reduction effect, an AWTD parameter optimization model is established, which takes the composite index of signal-to-noise ratio, root mean square error, smoothness and correlation coefficient into consideration, and considers wavelet basis function, decomposition level, threshold function and threshold criterion.
[0059] Construct the objective function of the AWTD parameter optimization model.
[0060] The solution of this embodiment takes the optimal composite evaluation index of wavelet threshold denoising performance as the objective function, which is as follows: max f=maxη Among them, η is a composite evaluation index of wavelet threshold denoising that takes into account signal-to-noise ratio, root mean square error, smoothness, and correlation coefficient.
[0061] η=w′1η SNR +w′2ηRMSE +w′3η r +w′4η R Among them, w′1, w′2, w′3 and w′4 are weight coefficients of the signal-to-noise ratio, root mean square error, smoothness and correlation coefficient indicators respectively. The solution of this embodiment adopts the hierarchical analysis method for solution.
[0062] Specifically, the signal-to-noise ratio η SNR The calculation formula is as follows: Among them, O(i) is the original signal data; G(i) is the reconstructed signal after wavelet threshold denoising; N is the total number of points in the data sequence.
[0063] Specifically, the root mean square error η RMSE The calculation formula is as follows: Specifically, the smoothness η r The calculation formula is as follows: Specifically, the correlation coefficient η R The calculation formula is as follows: Set the constraints of the AWTD parameter optimization model. The decision variables of the AWTD parameter optimization model are wavelet basis function, decomposition level, threshold function and threshold criterion, and set their constraints accordingly.
[0064] The constraints for the wavelet basis function are: Ψ(t)∈T Among them, Ψ(t) is the wavelet basis function; Ψ is the set of wavelet basis functions.
[0065] The constraints on the number of decomposition levels are: d min ≤d≤d max Where, d is the number of decomposition levels; d min and d max are the lower and upper limits of the number of decomposition levels, respectively.
[0066] The constraints for the threshold function are: T hf ∈Ω Among them, T hf is the threshold function; Ω is a set of threshold functions.
[0067] The constraints for the threshold criterion are: H∈Γ Among them, H is the threshold criterion; Γ is the threshold criterion set. S4: An improved harmony search algorithm is proposed to solve the AVMD parameter optimization model and the AWTD parameter optimization model. The optimal parameter combination scheme is obtained by solving the model, thereby achieving optimal filtering and denoising of the noise component.
[0068] Get the preset algorithm parameter values, including the harmony memory sample capacity HM, the harmony memory consideration probability HMCR, and the pitch adjustment probability dynamic adjustment upper limit PAR max , Pitch adjustment probability dynamic adjustment lower limit PAR min , Tone adjustment bandwidth dynamic adjustment upper limit bw max And the tone adjustment bandwidth dynamically adjusts the lower limit bw min And other algorithm-related parameter values.
[0069] An initial harmony memory library with a sample capacity of HM is established, and each harmony group in the initial harmony memory library includes an initial decomposition parameter group and its corresponding local signal envelope entropy value.
[0070] The initial decomposition parameter group is randomly extracted and generated within the preset quadratic penalty factor and modal decomposition number variable value range, and each decomposition mode is obtained by variational modal decomposition, and the local envelope entropy is calculated according to the local envelope entropy calculation formula, wherein the preset quadratic penalty factor variable value range is [α min , α max ], the preset range of the modal decomposition number variable is [K min , K max ].
[0071] According to a preset dynamic adjustment strategy, the harmony in the initial harmony memory library is updated.
[0072] Specifically, a random number [0, 1] is generated. If the random number is greater than the probability HMCR considered by the harmonic memory bank, a quadratic penalty factor and a modal decomposition number that meet the preset quadratic penalty factor and modal decomposition number variable range requirements are regenerated from outside the initial harmonic memory bank to form a new decomposition parameter group, and each decomposition mode is obtained through variational modal decomposition.
[0073] The local envelope entropy is calculated according to the local envelope entropy calculation formula to form a new harmony as the first harmony, otherwise a harmony is selected from the initial harmony memory library as the first harmony.
[0074] If the first harmony comes from the initial harmony memory library, a random number of [0, 1] is regenerated. If the random number is less than the pitch adjustment probability PAR, the quadratic penalty factor and modal decomposition number in the first harmony are adjusted according to the pitch adjustment bandwidth bw. Otherwise, no adjustment is made to form the second harmony.
[0075] According to the quadratic penalty factor of the second harmony and the modal decomposition number, each decomposed mode is obtained through variational modal decomposition.
[0076] The local envelope entropy is calculated according to the local envelope entropy calculation formula to obtain the objective function value corresponding to the second harmony. If the corresponding objective function value is better than the worst harmony in the initial harmony memory library, the second harmony replaces the worst harmony to update the initial harmony memory library.
[0077] According to the preset dynamic adjustment strategy, the harmony in the initial harmony memory library is updated, and it is iteratively optimized to determine whether the preset convergence condition group is met. When the preset convergence condition group is met, the quadratic penalty factor and modal decomposition number corresponding to the optimal harmony are output, that is, the optimal decomposition parameter group is output.
[0078] Among them, the secondary penalty factor and the modal decomposition number in the first harmony are adjusted according to the pitch adjustment bandwidth bw, which is specifically performed through the following strategy formula: K2=K1+bw(g) α2=α1+bw(g) In the formula, g is the current iteration number; g max is the maximum number of iterations; K1 and K2 are the modal decomposition numbers of the first and second harmony respectively; α1 and α2 are the quadratic penalty factors for the first and second harmony, respectively.
[0079] According to the above steps, the optimal decomposition parameter group of the AVMD parameter optimization model can be solved. Similarly, the AWTD parameter optimization model can be solved by using the above steps to optimize the wavelet basis function, decomposition layer number, threshold function, threshold criterion and other parameters.
[0080] In order to enable those skilled in the art to better understand the solution of this embodiment, a specific calculation example and simulation result analysis are provided. Specifically, this embodiment uses the DC voltage and current simulation data at the output end of the DC charging pile to verify the method.
[0081] Usually, the DC charging pile adopts constant voltage or constant current charging mode. In addition to the DC fundamental component, the output signal has high harmonic content, including 11th harmonic, 13th harmonic and 15th harmonic, with harmonic content of 5%, 3% and 1% respectively. Therefore, according to the formula of this embodiment, the simulation data of the original signal of the DC voltage and DC current at the output end is calculated.
[0082] Wherein, U0 is the fundamental component of the DC voltage at the output end, and in this embodiment, the values are 500V; I0 is the fundamental component of the DC current at the output end, and in this embodiment, the value is 60A; r 11 、r 13 、r 15 The percentages of the 11th harmonic, 13th harmonic and 15th harmonic are 5%, 3% and 1% respectively in this embodiment; In this embodiment, the harmonic orders k1=11, k2=13, k3=15; the power frequency f=50 Hz; and the sampling frequency t=1 ms.
[0083] According to the above parameters, the simulation sequence of the original signal of DC voltage and DC current at the output end is as follows Figure 1 , Figure 2 shown.
[0084] The improved harmony search algorithm is used to solve the AVDM parameter optimization model. In this embodiment, the algorithm parameters are set as follows: harmony memory sample capacity HM = 100, harmony memory consideration probability HMCR = 0.95, and pitch adjustment probability dynamic adjustment upper limit PAR max =0.9, pitch adjustment probability dynamic adjustment lower limit PAR min =0.5, the upper limit value bw of the dynamic adjustment bandwidth of the tone adjustment max = 1 and the lower limit value bw of the dynamic adjustment bandwidth of the tone adjustment min =0.01. The range of the modal number is 1 to 20, and the range of the quadratic penalty factor is 0 to 1000. The optimal quadratic penalty factor and modal decomposition number are obtained.
[0085] Among them, the optimization parameters of the DC voltage original signal AVDM are: The optimal number of modes is 6 and the quadratic penalty factor is 416.
[0086] Similarly, AVMD is performed on the original DC current signal, and the calculated optimization parameters are: The optimal number of modes is 10 and the quadratic penalty factor is 3.
[0087] According to the above parameter settings, the AVMD of the original DC voltage and DC current signals is realized, such as Figure 3 , Figure 4 shown.
[0088] according to Figure 3 , Figure 4 The maximum mutual correlation coefficient between each IMF and the original signal is calculated based on the results.
[0089] The maximum mutual information coefficient threshold is set to 0.8 to filter out the effective components and noise components.
[0090] Among them, the effective components after the DC voltage original signal AVMD are modes 1 to 4, and the noise components are modes 5 and 6; the effective components after the DC current original signal AVMD are modes 1, 3, 5, 7 to 10, and the noise components are modes 2, 4, and 6.
[0091] The improved harmony search algorithm is further used to solve the WTD parameter optimization model to achieve optimal filtering and denoising of noise components.
[0092] Among them, the wavelet basis functions include haar, db, bior, coif, and sym; The upper limit of decomposition levels is set to 20, and the lower limit of decomposition levels is set to 2; The threshold function includes a soft threshold function and a hard threshold function; Threshold criteria include rigrsure criterion, sqtwolog criterion, heursure criterion, and minimaxi criterion.
[0093] Based on the optimized parameter combination, AWTD is performed on the noise component, and the filtered and denoised components are superimposed on the effective components to obtain the denoised DC voltage and current data, such as Figure 5 , Figure 6 shown.
[0094] In order to further study the effectiveness of the proposed method, the solution of this embodiment considers the following four methods for comparison.
[0095] Method 1: Classic WTD. When using WTD for filtering and noise reduction, db4 is selected as the wavelet basis function, the number of decomposition layers is 4, and the soft threshold function and heursure criterion are used; Method 2: Classic VMD and AWTD. When using classic VMD to perform modal decomposition on the original signals of DC voltage and DC current, the quadratic penalty factor and modal decomposition number are set to 1000 and 5 respectively. Method 3: AVMD and classic WTD. When using WTD for filtering and denoising, db4 is selected as the wavelet basis function, the number of decomposition layers is 4, and the soft threshold function and heursure criterion are used. Method 4: AVMD-AWTD, which is the method adopted in this embodiment.
[0096] According to the above parameter settings, the signal-to-noise ratio and root mean square error indicators of various methods for reducing noise signals are calculated as shown in Tables 1 and 2.
[0097] Table 1. Signal-to-noise ratio and root mean square error indicators of noise reduction signals method Signal-to-Noise Ratio Root mean square error Method 1 17.215 21.242 Method 2 20.682 19.672 Method 3 25.344 18.215 Method 4 28.526 16.983 Table 2. SNR and RMS error of noise-reduced signals method Signal-to-Noise Ratio Root mean square error Method 1 21.532 5.958 Method 2 23.667 4.482 Method 3 26.728 3.017 Method 4 30.945 1.842 Comparing the results of Method 2, Method 3 and Method 1, it can be seen that the application of VMD can improve the filtering and denoising effect of WTD.
[0098] By comparing the results of method 4 with those of method 2 and method 3, it can be seen that after the introduction of the AVMD mechanism, the average improvement values of the signal-to-noise ratio and the root mean square error are 34.34% and 36.29%; after the introduction of the AWTD mechanism, the average improvement values of the signal-to-noise ratio and the root mean square error are 14.17% and 22.86%. It can be seen that by introducing the parameter optimization mechanism to improve the classic VMD and WTD, the filtering and noise reduction effect of the DC metering signal can be significantly improved.
[0099] The scheme of this embodiment establishes a parameter optimization model for establishing the variational modal decomposition of the DC voltage and current signals at the output end, and combines mutual information analysis to achieve adaptive distinction between the effective modal components and the noise modal components of the original signal. On this basis, a wavelet threshold denoising parameter optimization model with optimal composite evaluation indicators of signal-to-noise ratio, root mean square error, smoothness, and correlation coefficient is established to achieve optimal filtering and denoising of the noise modal components. The effectiveness of the proposed method is verified by calculation and analysis of measured data.
[0100] In summary, the present invention aims at the problem that the parameter setting of the existing DC power load output terminal voltage and current signal filtering and noise reduction method is not optimized and the filtering and noise reduction effect needs to be improved, and proposes a DC power meter measurement signal noise reduction method based on AVMD-AWTD. From the example analysis, it can be concluded that the method proposed in the present invention can further improve the filtering and noise reduction effect of the DC power meter measurement signal, thereby ensuring the accuracy of DC power measurement.
[0101] It should be understood that the embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope limited by the appended claims of the application.
Claims
1. A method for reducing noise of a DC electric energy meter measurement signal, characterized in that: The following steps are involved: S1: Establish an adaptive variational mode decomposition parameter optimization model for the output DC voltage / current signal with the goal of minimizing the local signal envelope entropy; S2: Calculate and solve the model, and distinguish the effective component from the noise component according to the maximum mutual information coefficient between each modal component and the original sampling data sequence obtained by the solution; S3: For the noise component, an adaptive wavelet threshold denoising parameter optimization model is established with composite index evaluation to filter and reduce the noise component; S4: The improved harmony search algorithm is used to solve the adaptive variational mode decomposition parameter optimization model and the adaptive wavelet threshold denoising parameter optimization model, and the optimal parameter combination scheme is obtained by solving the model.
2. A method for reducing noise of a DC electric energy meter measurement signal according to claim 1, characterized in that: The modal components after modal decomposition of the original sampling data sequence of the DC voltage / current at the output end are subjected to Hilbert transformation to obtain a new modal component sequence; Dividing each new modal component sequence by the sum of all new modal component sequences obtains a proportional coefficient of each data point in the new modal component sequence after Hilbert transformation; The signal envelope entropy value of the modal component of the original sampling data sequence of the DC voltage / current at the output end after the variational modal decomposition is calculated according to the proportional coefficient of all data points.
3. A method for reducing noise of a DC electric energy meter measurement signal according to claim 1 or 2, characterized in that: Set the constraints of the adaptive variational mode decomposition parameter optimization model. The quadratic penalty factor and the number of mode decompositions in the model need to meet their respective upper and lower limit constraints.
4. A method for reducing noise of a DC electric energy meter measurement signal according to claim 1 or 2, characterized in that: When the maximum mutual information coefficient is greater than or equal to the maximum mutual information coefficient threshold, the corresponding modal component is a valid component; when the maximum mutual information coefficient is less than the maximum mutual information coefficient threshold, the corresponding modal component is a noise component.
5. A method for reducing noise of a DC electric energy meter measurement signal according to claim 4, characterized in that: Maximum mutual information coefficient γ MIC (X i ; O) The calculation expression is: Where N is the number of sample points of each decomposed modal sequence and the original signal; p(x i , o) is the joint probability density of the i-th decomposition mode sequence and the original signal; p(x i ) is the marginal probability density of the i-th decomposition modal sequence; p(o) is the marginal probability density of the i-th original signal; The grid method is used to calculate the marginal probability density and joint probability density of each decomposed modal sequence and the original signal; a and b are the number of rows and columns of the grid, respectively.
6. A method for reducing noise of a DC electric energy meter measurement signal according to claim 1 or 5, characterized in that: The adaptive wavelet threshold denoising parameter optimization model uses wavelet basis function, decomposition layer number, threshold function and threshold criterion as decision variables; The composite index evaluation is obtained by weighted summation of signal-to-noise ratio, root mean square error, smoothness and correlation coefficient.
7. A method for reducing noise of a DC electric energy meter measurement signal according to claim 6, characterized in that: The constraints for setting the adaptive wavelet threshold denoising parameter optimization model include: The small basis wave function is in the set of small basis wave functions; The number of decomposition levels is between the upper and lower limits of the set decomposition levels; The threshold function is in the threshold function set; Threshold criteria are in the threshold criteria set.
8. A method for reducing noise of a DC electric energy meter measurement signal according to claim 1 or 7, characterized in that: The improved harmony search algorithm comprises: Establishing an initial harmony memory library, wherein each harmony group in the initial harmony memory library includes an initial decomposition parameter group and its corresponding local signal envelope entropy value; According to a preset dynamic adjustment strategy, the harmony in the initial harmony memory library is updated; Iterate the optimization to determine whether the preset convergence condition group is met. When the preset convergence condition group is met, output the quadratic penalty factor and modal decomposition number corresponding to the optimal harmony.
9. A method for reducing noise of a DC electric energy meter measurement signal according to claim 8, characterized in that: The dynamic adjustment strategy includes: Generate a random number. If the random number is greater than the probability considered by the harmony memory library, regenerate a decomposition parameter group that meets the preset conditions from the initial harmony memory library, obtain each decomposition mode through variational mode decomposition, calculate the local envelope entropy of the corresponding decomposition mode, and use it as the first harmony; otherwise, select a harmony from the initial harmony memory library as the first harmony; If the first harmony comes from the initial harmony memory library, a random number is regenerated. If the random number is less than the pitch adjustment probability, the secondary penalty factor and the modal decomposition number in the first harmony are adjusted according to the pitch adjustment bandwidth. Otherwise, no adjustment is made to form a second harmony. According to the quadratic penalty factor of the second harmony and the modal decomposition number, each decomposition mode is obtained through variational modal decomposition; The local envelope entropy of the corresponding decomposition mode is calculated to obtain the objective function value corresponding to the second harmony. If the corresponding objective function value is better than the worst harmony in the initial harmony memory bank, the second harmony replaces the worst harmony to update the initial harmony memory bank.
10. A method for reducing noise of a DC electric energy meter measurement signal according to claim 9, characterized in that: The expression for adjusting the secondary penalty factor and the modal decomposition number in the first harmony according to the pitch adjustment bandwidth is: K2=K1+bw(g) α2=α1+bw(g) In the formula, PAR min and PAR max Dynamically adjust the lower and upper limits for the pitch adjustment probability, respectively; bw min and bw max Dynamically adjust lower and upper limits for the tone adjustment bandwidth, respectively; g is the current iteration number; g max is the maximum number of iterations; K1 and K2 are the modal decomposition numbers of the first and second harmony respectively; α1 and α2 are the quadratic penalty factors for the first and second harmony, respectively.
Citation Information
Patent Citations
Direct-current electric energy meter and electric energy metering method
CN117388570A
DC electric energy metering noise reduction method, device, storage medium and system
CN116840770A
Wavelet threshold denoising method and device based on parameter optimization
CN117200748A
Cited By
Method and system for measuring trace ammonia by optical means
CN119985446A
