Self-adaptive filtering method and system for intelligent electric energy meter

Abstract

The invention discloses a self-adaptive filtering method and system for an intelligent electric energy meter, and relates to the technical field of power grid signal processing, and the method comprises the steps: obtaining a power grid voltage and current signal set collected by a target intelligent electric energy meter historically, and obtaining a signal noise spectrum data set and a noise level evaluation result through spectrum analysis and noise level evaluation; if the result exceeds a preset noise threshold value, identifying an interference type and building a noise adaptive filtering multi-channel, integrating and deploying the noise adaptive filtering multi-channel into a target intelligent electric meter to complete signal filtering test and effect evaluation, and dynamically optimizing and generating an electric energy meter adaptive filtering multi-channel based on the obtained multi-channel filtering effect index parameter; and finally, signal filtering closed-loop control is realized. The technical problem that a traditional electric energy meter is weak in filtering pertinence and cannot adapt to dynamic noise of a power grid in electric power related scenes with complex dynamic interference is solved, and the technical effects that the electric variable measurement accuracy of the electric energy meter is improved, and the adaptive capacity of the electric energy meter to different interference scenes is enhanced are achieved.

Description

Technical Field

[0001] This invention relates to the field of power grid signal processing technology, and in particular to an adaptive filtering method and system for smart meters. Background Technology

[0002] Accurate measurement of voltage and current by smart meters is crucial for electricity metering and the efficient and stable operation of the power grid, and is of great significance for electricity consumption accounting and system management. Current signal filtering for electricity variable measurements mostly employs traditional single-filter schemes, which have played a certain role in stable power scenarios. However, with the increasing types and dynamic characteristics of power grid interference, traditional filtering methods lack targeted adaptability and cannot accurately handle complex and diverse noise interference, resulting in insufficient accuracy and reliability of electricity variable measurement data, making it difficult to meet the actual needs of accurate assessment and efficient management of electricity metering. Summary of the Invention

[0003] This application provides an adaptive filtering method and system for smart meters, which solves the technical problems of traditional meter filtering being weak in targeting and unable to adapt to dynamic noise of the power grid in power-related scenarios with complex dynamic interference.

[0004] The first aspect of this application provides an adaptive filtering method for smart meters. The method includes: acquiring a historically collected set of grid voltage and current signals from a target smart meter; performing spectrum analysis and noise level assessment on the grid voltage and current signal set to obtain a signal-noise spectrum dataset and a noise level assessment result; if the noise level assessment result exceeds a preset noise threshold, identifying the interference type of the signal-noise spectrum dataset to obtain the noise signal interference type, and constructing a noise adaptive filtering multi-channel based on the noise signal interference type; integrating and deploying the noise adaptive filtering multi-channel within the target smart meter for signal filtering testing and effect evaluation to obtain multi-channel filtering effect index parameters; dynamically optimizing the noise adaptive filtering multi-channel based on the multi-channel filtering effect index parameters to generate a smart meter adaptive filtering multi-channel, and performing closed-loop signal filtering control through the smart meter adaptive filtering multi-channel.

[0005] A second aspect of this application provides an adaptive filtering system for smart meters, the system comprising: a voltage and current signal set acquisition module, used to acquire historically collected grid voltage and current signal sets from the target smart meter, perform spectrum analysis and noise level assessment on the grid voltage and current signal sets to obtain a signal-noise spectrum dataset and a noise level assessment result; a multi-channel filtering construction module, used to identify interference types in the signal-noise spectrum dataset if the noise level assessment result exceeds a preset noise threshold, obtain noise signal interference types, and build a noise adaptive filtering multi-channel based on the noise signal interference types; a filtering parameter acquisition module, used to integrate and deploy the noise adaptive filtering multi-channel within the target smart meter for signal filtering testing and effect evaluation to obtain multi-channel filtering effect index parameters; and a filtering control execution module, used to dynamically optimize the noise adaptive filtering multi-channel based on the multi-channel filtering effect index parameters, generate a smart meter adaptive filtering multi-channel, and perform closed-loop signal filtering control through the smart meter adaptive filtering multi-channel.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application acquires historical electrical variable signals from smart energy meters, and constructs an adapted multi-channel filtering system through spectrum analysis, noise assessment, and interference type identification. It then dynamically optimizes the multi-channel filtering performance parameters to form an adaptive filtering closed-loop control. This accurately handles various interferences affecting electrical variable measurement, making the electrical variable measurement results of smart energy meters more accurate and stable. This achieves the technical effect of improving the accuracy of electrical variable measurement and enhancing its adaptability to different interference scenarios. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a flowchart illustrating the adaptive filtering method for smart energy meters provided in an embodiment of this application.

[0009] Figure 2 This is a schematic diagram of the structure of an adaptive filtering system for smart energy meters provided in an embodiment of this application.

[0010] Figure labeling: Voltage and current signal set acquisition module 1, filtering multi-channel construction module 2, filtering parameter acquisition module 3, filtering control execution module 4. Detailed Implementation

[0011] This application provides an adaptive filtering method and system for smart meters, which solves the technical problems of traditional meter filtering being weak in targeting and unable to adapt to dynamic noise of the power grid in power-related scenarios with complex dynamic interference.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, an adaptive filtering method for smart energy meters includes: The system acquires a set of historically collected grid voltage and current signals from the target smart meter, performs spectrum analysis and noise level assessment on the grid voltage and current signal set, and obtains a signal noise spectrum dataset and noise level assessment results.

[0015] Specifically, firstly, the historical voltage and current signal sets of the target smart meter are acquired through its built-in storage module or communication interface. These signals are in digital form and contain voltage and current sampling data from different time periods. The sampling frequency follows the conventional standards for power grid monitoring to ensure that the signals can fully reflect the actual operating status of the power grid.

[0016] Next, the power grid voltage and current signal set is standardized and time-series labeled to form a standard voltage and current sequence signal set. Based on the data characteristics and data analysis accuracy requirements of this standard signal set, spectral analysis parameters including window functions and FFT points are set. Subsequently, spectral analysis is carried out according to the set parameters to obtain a signal-noise spectral dataset. Finally, by defining a noise level assessment index, the dataset is evaluated to obtain the final noise level assessment result. This step will be explained in detail later.

[0017] If the noise level assessment result exceeds the preset noise threshold, the interference type is identified in the signal noise spectrum dataset to obtain the noise signal interference type, and a noise adaptive filtering multi-channel is built according to the noise signal interference type.

[0018] Optionally, if the noise level assessment result exceeds a preset noise threshold, firstly, feature parameters are extracted from the signal noise spectrum dataset to obtain a set of signal noise spectrum feature parameters including peak frequency, bandwidth, amplitude distribution, and harmonic components. Then, based on the electricity meter signal application standards, an electricity meter interference signal pattern library is constructed. Subsequently, a similarity analysis is performed on the feature parameter set based on this pattern library to generate a signal interference pattern similarity set. Finally, according to this similarity set, the interference type is identified and integrated in the electricity meter interference signal pattern library to ultimately obtain the noise signal interference type. This step will be explained in detail later.

[0019] The following is the process for building the noise adaptive filtering multi-channel: First, based on the noise signal interference type, design a filtering multi-channel architecture that corresponds one-to-one between each filtering channel and the signal interference type. Then, select an appropriate interference type-noise filtering algorithm based on the noise signal interference type. Next, train the filter gain based on this filtering algorithm to construct multiple interference type signal filters. Finally, map and embed these interference type signal filters into the aforementioned filtering multi-channel architecture to complete the construction of the noise adaptive filtering multi-channel. This step will be explained in detail later.

[0020] The noise adaptive filtering multi-channel was integrated and deployed in the target smart meter for signal filtering testing and effect evaluation, and the multi-channel filtering effect index parameters were obtained.

[0021] In one embodiment of this application, the noise adaptive filtering multi-channel is first integrated and deployed within the target smart meter. Using an embedded system integration method, the hardware interface and computing resources of the target smart meter are first adapted. The hardware driver for the noise adaptive filtering multi-channel is ported to the embedded processor of the meter, ensuring compatibility between the driver and the meter's hardware circuitry. The signal input interface of this filtering multi-channel is connected to the voltage and current signal acquisition module within the meter via a standardized data bus, and the output interface is connected to the signal receiving end of the meter's metering chip, thus achieving a seamless link between signal acquisition, filtering, and metering analysis. At the software level, the control logic module for the noise adaptive filtering multi-channel is embedded into the meter's main control program. The module's operating parameters are configured to be consistent with the meter's sampling frequency and data transmission rate. After integration and deployment, a power-on self-test is performed to verify whether the noise adaptive filtering multi-channel can respond normally to the meter's control commands and whether the signal transmission is stable.

[0022] Next, signal filtering tests were conducted. A combination of simulated interference signal testing and actual operating condition signal testing was employed. First, a signal generator produced simulated test signals including various types of harmonic interference and pulse interference. These signals were injected into the signal acquisition terminal of the target smart meter according to different interference intensity levels, ensuring that the test signals covered the interference scenarios that the meter might encounter in actual operation. Simultaneously, the actual voltage and current signals of the target smart meter under actual power grid conditions were collected as actual operating condition test data. The meter's signal acquisition and filtering functions were then activated. The multi-channel filtering system automatically matched the corresponding filtering channel and filter according to the type of interference in the input signal. The data acquisition equipment synchronously recorded the signal data before and after filtering, including the time-domain waveforms and frequency-domain spectrum data of the original input signal and the filtered output signal, ensuring the integrity and synchronization of the test data.

[0023] Then, the filtering effect is evaluated and the index parameters are calculated. The noise amplitude exceeding the standard rate, average noise amplitude, signal distortion, and signal-to-noise ratio (SNR) before and after filtering are selected as core evaluation indicators. The noise amplitude exceeding the standard rate and average noise amplitude are calculated using the previously set calculation method, respectively, by statistically analyzing the percentage of frequency points with exceeding the standard amplitude in the filtered signal noise spectrum data and the arithmetic mean of the amplitudes of all noise frequency points. Signal distortion is calculated using the total harmonic distortion rate (THD) method, by comparing the harmonic components of the filtered signal with those of a standard clean signal. The SNR is obtained by calculating the ratio of the effective amplitude of the filtered signal to the residual noise amplitude. All indicators are quantified using the meter's built-in data processing module or an external computer, and finally integrated to form multi-channel filtering effect index parameters.

[0024] Based on the multi-channel filtering effect index parameters, the noise adaptive filtering multi-channel is dynamically optimized to generate the energy meter adaptive filtering multi-channel, and the signal filtering closed-loop control is performed through the energy meter adaptive filtering multi-channel.

[0025] Specifically, the generation process of the adaptive filtering multi-channel for the energy meter is as follows: First, based on the multi-channel filtering effect index parameters, the noise adaptive filtering multi-channel is dynamically optimized and analyzed to determine the multi-channel optimization direction parameter threshold; then, the target multi-channel optimization direction parameter is obtained within the parameter threshold range; finally, the noise adaptive filtering multi-channel is optimized and updated based on the target parameter to generate the energy meter adaptive filtering multi-channel. This step will be explained in detail in the following content.

[0026] Then, the signal filtering closed-loop control process is initiated. The target smart meter continuously collects grid voltage and current signals and inputs them into the adaptive filtering multi-channel in real time. The multi-channel automatically matches the corresponding filtering channel and the optimized filter for precise filtering based on the real-time changes in the type of interference in the signal. At the same time, the filtered signal data is collected in real time, and core performance indicators such as noise amplitude exceedance rate and signal-to-noise ratio are calculated. These are continuously compared with preset qualified standards. If the indicators deviate, a new round of parameter optimization and multi-channel configuration update is triggered. Through the iterative cycle of acquisition-filtering-evaluation-optimization, dynamic closed-loop control of the signal filtering effect is achieved, ensuring that the energy meter maintains stable filtering performance and metering accuracy under complex interference scenarios.

[0027] Furthermore, the method provided in this application embodiment includes: The power grid voltage and current signal set is standardized and time-series identified to obtain a standard voltage and current sequence signal set. Based on the data characteristics and data analysis accuracy requirements of the standard voltage and current sequence signal set, spectrum analysis parameters are set, including window functions and FFT points. Spectrum analysis is performed on the standard voltage and current sequence signal set according to the spectrum analysis parameters to obtain a signal noise spectrum dataset. A noise level assessment index is defined, and the signal noise spectrum dataset is evaluated based on the noise level assessment index to obtain a noise level assessment result.

[0028] Specifically, firstly, the power grid voltage and current signal sets obtained in the aforementioned steps are standardized and time-series labeled. Standardization employs a maximum-minimum normalization method, first statistically analyzing the maximum and minimum values ​​of voltage and current data in the signal set, then transforming all data to the range of 0 to 1 using a linear mapping formula to eliminate the influence of different units and ensure data comparability. Time-series labeling involves marking each sampled data point with a corresponding acquisition timestamp according to the order of signal acquisition, forming a standard voltage and current sequence signal set, facilitating subsequent tracing of the signal's temporal characteristics.

[0029] Then, based on the data characteristics and data analysis accuracy requirements of the standard voltage and current sequence signal set, the corresponding spectrum analysis parameters are set. The Hanning window is selected as the window function, which effectively balances spectral leakage suppression and frequency resolution while maintaining simple logic. The FFT points are set to 1024, a power of 2, which is compatible with the butterfly operation structure of the Fast Fourier Transform, improving computational efficiency while meeting the conventional accuracy requirements for smart meter signal analysis.

[0030] Subsequently, according to the set spectrum analysis parameters, the standard voltage and current sequence signal set is subjected to spectrum calculation to generate a voltage and current signal spectrum diagram. Then, a normal spectrum diagram of the electricity meter signal is obtained, and a normal feature threshold is determined by feature extraction. Following this normal feature threshold, anomaly comparison and identification are performed on the voltage and current signal spectrum diagram, marking the abnormal signal frequency point set. Finally, based on the abnormal signal frequency point set, noise data in the voltage and current signal spectrum diagram is filtered and extracted to obtain a signal noise spectrum dataset. This step will be explained in detail later.

[0031] Next, noise level assessment indicators are defined and assessments are conducted. The noise amplitude exceedance rate and average noise amplitude are selected as core assessment indicators. The noise amplitude exceedance rate is the ratio of the number of frequency points in the signal noise spectrum data whose amplitude exceeds the preset qualified threshold to the total number of noise frequency points. The average noise amplitude is the arithmetic mean of the amplitudes of all noise frequency points in the data set. The preset qualified threshold can be selected as 1.1-1.3 times the maximum amplitude of the normal signal or 2-3 times the standard deviation of the noise mean.

[0032] Finally, the calculated values ​​of the two indicators are compared with their corresponding preset acceptable thresholds. Then, a weighted summation is used to generate a comprehensive quantitative noise level assessment value. The weights are set according to the metering accuracy requirements, with the core principle being that indicators with a greater impact on metering error have higher weights. The specific setting method is as follows: First, the correlation between the two indicators and the metering error is calibrated experimentally. In a standard laboratory environment, scenarios with different intensities of noise amplitude exceeding the standard and average noise amplitude exceeding the standard are simulated. The corresponding metering error values ​​are recorded, and the Pearson correlation coefficient between the two indicators and the metering error is calculated. Indicators with larger absolute values ​​of correlation coefficients are assigned higher weights. For example, if the noise amplitude exceeding rate is 0.6 and the average noise amplitude is 0.4, the calculation method is: Comprehensive quantitative value = Noise amplitude exceeding rate × 0.6 + (Average noise amplitude / Preset amplitude benchmark value) × 0.4. The preset amplitude benchmark value is 1.2 times the maximum amplitude of the normal signal. This comprehensive quantitative value is the final noise level assessment result.

[0033] Through a coherent process of signal acquisition, standardization, parameter setting, spectrum analysis, and noise assessment, combined with signal processing methods, noise spectrum data in power grid voltage and current signals were accurately extracted and the noise level was quantitatively assessed, providing a reliable basis for subsequent interference identification and filtering scheme customization.

[0034] Furthermore, the method provided in this application embodiment includes: The standard voltage and current sequence signal set is subjected to spectral calculation according to the spectral analysis parameters to generate a voltage and current signal spectrum; a normal spectrum of the electricity meter signal is obtained, and features are extracted from the normal spectrum of the electricity meter signal to determine the normal signal feature threshold; based on the normal signal feature threshold, anomaly comparison and identification are performed on the voltage and current signal spectrum to mark the abnormal signal frequency point set; noise data is filtered and extracted from the voltage and current signal spectrum according to the abnormal signal frequency point set to obtain a signal noise spectrum dataset.

[0035] Optionally, firstly, according to the spectrum analysis parameters determined in the aforementioned steps, the standard voltage and current sequence signal set is subjected to spectrum calculation to generate a voltage and current signal spectrum diagram. Specifically, the Fast Fourier Transform (FFT) method is used. First, the standard voltage and current sequence signal set is divided into frames of fixed length to ensure that each frame contains complete periodic features. Then, each frame is multiplied by a Hanning window, which effectively balances spectral leakage suppression and frequency resolution. Subsequently, a 1024-point FFT operation is performed on each processed frame. This number of points is adapted to the butterfly operation structure of the FFT, which can improve computational efficiency while meeting the signal analysis accuracy requirements, converting the time-domain signal into a frequency-domain signal to obtain the amplitude data corresponding to each frequency point. Finally, the voltage and current signal spectrum diagram is plotted with frequency as the x-axis and amplitude as the y-axis.

[0036] Next, the normal spectrum of the electricity meter signal was obtained, and the normal characteristic threshold of the signal was determined. This spectrum was obtained as follows: an electricity meter of the same model and specifications as the target smart meter was selected, and pure grid voltage and current signals were collected in a standard laboratory environment free from any interference. Then, the same spectrum analysis parameters, including the Hanning window and 1024-point FFT, were used to perform spectrum calculations to obtain the normal spectrum of the electricity meter signal. Feature extraction employed peak detection. The normal spectrum of the electricity meter signal was traversed to identify the characteristic frequency points corresponding to the power frequency and each harmonic. The amplitude range of these characteristic frequency points was recorded, and 1.2 times the upper limit of this range was taken as the normal characteristic threshold of the signal to ensure that normal signal fluctuations would not be misjudged as abnormal.

[0037] Then, based on the normal signal characteristic threshold, anomaly comparison and identification are performed on the voltage and current signal spectrum. A point-by-point comparison method is adopted, and the amplitude of each frequency point in the voltage and current signal spectrum is checked in turn. The amplitude is compared with the normal signal characteristic threshold. If the amplitude of a certain frequency point exceeds the normal signal characteristic threshold, the frequency point is determined to be an abnormal signal frequency point. At the same time, the specific frequency value and corresponding amplitude data of the frequency point are recorded. All the identified abnormal frequency points and their related data are sorted and summarized to form an abnormal signal frequency point set.

[0038] Finally, noise data is filtered and extracted from the voltage and current signal spectrum based on the abnormal signal frequency point set. In the voltage and current signal spectrum, the corresponding frequency domain data is accurately located based on the frequency values ​​recorded in the abnormal signal frequency point set. Frequency domain data corresponding to normal signals such as power frequency and harmonics are excluded, and only the frequency domain data corresponding to the abnormal signal frequency points are extracted. The extracted abnormal frequency domain data is then sorted and organized in ascending order of frequency to obtain the final signal noise spectrum dataset.

[0039] Through the above-described sequential process, combined with the specific spectral analysis parameters of the Hanning window and FFT points, the noise spectrum data in the power grid voltage and current signals was accurately extracted, providing accurate and complete data support for subsequent noise interference type identification.

[0040] Furthermore, the method provided in this application embodiment includes: Feature parameters are extracted from the signal and noise spectrum dataset to obtain a signal and noise spectrum feature parameter set, which includes peak frequency, bandwidth, amplitude distribution, and harmonic components. An energy meter interference signal pattern library is constructed based on the energy meter signal application standard. Similarity analysis is performed on the signal and noise spectrum feature parameter set based on the energy meter interference signal pattern library to obtain a signal interference pattern similarity set. Interference type identification and integration are performed on the energy meter interference signal pattern library according to the signal interference pattern similarity set to obtain the noise signal interference type.

[0041] Specifically, the first step is to obtain the preset noise threshold: Similar to the previous steps, a device of the same model and specifications as the target smart energy meter is selected. Pure grid voltage and current signals are collected in a standard, interference-free laboratory environment. The spectrum is calculated using a Hanning window combined with 1024-point FFT spectral analysis parameters to obtain a baseline normal signal spectrum. The characteristic frequency points of the power frequency and each harmonic are extracted using the peak detection method. The amplitude range is recorded, and the upper limit is taken. Referring to the grid interference allowable limits specified in GB / T 14549 power quality standard, the smaller value between the upper limit of the normal signal amplitude and the standard allowable limit is taken as the benchmark. A safety factor of 1.2 is added to finally determine the preset noise threshold. After obtaining this threshold, the previously obtained noise level assessment results are compared with the preset noise threshold. When the assessment result exceeds the preset noise threshold, it indicates that the current noise's impact on the energy meter's metering accuracy has exceeded the acceptable range, and the subsequent noise signal interference type identification process is initiated.

[0042] Next, feature parameters are extracted from the signal-noise spectrum dataset. The extraction process is as follows: Peak frequency: By traversing the signal-noise spectrum dataset, the peak detection algorithm is used to identify the amplitude corresponding to each frequency point, and the frequency point with the largest amplitude is selected as the peak frequency; Bandwidth: The half-width at half-maximum (WHM) method is used to calculate the bandwidth. First, the maximum amplitude corresponding to the peak frequency is determined, and then two frequency points corresponding to half the maximum amplitude are found. The difference between the two is the bandwidth; Amplitude distribution: The amplitude of all frequency points in the signal-noise spectrum dataset is statistically analyzed to generate an amplitude interval distribution histogram. The peak interval and the dispersion of the histogram are used as amplitude distribution features; Harmonic components: By comparing the signal-noise spectrum dataset with the normal spectrum of the electricity meter signal, the harmonic frequencies other than the fundamental frequency are identified, and the frequency values ​​and corresponding amplitudes of these harmonic frequencies are recorded as harmonic component features.

[0043] Subsequently, an interference signal pattern library for smart meters was constructed based on the application standards for smart meter signals. The construction process combined power industry standards with accumulated experimental data, and the specific steps were as follows: First, common noise interference types of smart meters were identified, including harmonic interference, impulse interference, power frequency drift interference, and electromagnetic radiation interference. Referring to relevant industry standards such as GB / T 14549 "Power Quality - Harmonics in Public Power Grids," the typical characteristic range of each type of interference was clarified. In a standard laboratory environment, various typical interference signals were simulated using a signal generator. These interference signals were then superimposed with clean power grid signals to obtain test signals containing a single interference type. The same spectral analysis parameters, including Hanning window and 1024-point FFT, were used to process each type of test signal, extracting characteristic parameters such as peak frequency, bandwidth, amplitude distribution, and harmonic components to form a standard characteristic parameter template for each type of interference. The label information of all interference types and their corresponding standard characteristic parameter templates were compiled and summarized. A structured smart meter interference signal pattern library was constructed according to the mapping relationship between interference type and standard characteristic parameter set. The library also reserves an expansion interface for subsequent addition of feature templates for new interference types.

[0044] Then, the Euclidean distance similarity calculation method is adopted. The specific process is as follows: First, all parameters of the signal-noise spectrum feature parameter set and each standard feature parameter template in the energy meter interference signal pattern library are normalized to eliminate the influence of differences in the dimensions of different parameters. For each interference type feature template in the energy meter interference signal pattern library, the Euclidean distance between the normalized feature parameter set to be identified and the feature parameter set of the template is calculated. The smaller the distance value, the higher the similarity between the two. The Euclidean distance calculation results between the feature parameter set to be identified and all interference type templates in the pattern library are sorted according to the interference type label to form a signal interference pattern similarity set.

[0045] Finally, a similarity threshold was set, which was calibrated experimentally; an example value of 0.8 was used, corresponding to the critical value of Euclidean distance. The similarity set of signal interference patterns was traversed, and interference type templates with similarity higher than the threshold were selected. If only one interference type template with similarity higher than the threshold existed, the interference type corresponding to that template was directly used as the preliminary identification result. If multiple interference type templates with similarity higher than the threshold existed, the interference type corresponding to the template with the highest similarity was selected as the primary interference type, and the others that met the criteria were designated as secondary interference types. If no template with similarity higher than the threshold existed, the interference was marked as unknown interference. The above identification results were integrated to clarify the primary and secondary interference types, ultimately yielding a complete list of noise signal interference types.

[0046] Through the above steps, combined with signal processing, standard references and similarity calculation methods, the accurate identification and integration of interference signal types exceeding the preset noise threshold is achieved, providing an accurate basis for the subsequent construction of a highly adaptable adaptive filtering multi-channel.

[0047] Furthermore, the method provided in this application embodiment includes: Based on the noise signal interference type, a multi-channel filtering architecture is designed, where each filtering channel in the multi-channel filtering architecture corresponds one-to-one with the signal interference type; an interference type-noise filtering algorithm is selected according to the noise signal interference type; filter gain training is performed based on the interference type-noise filtering algorithm to construct multiple interference type signal filters; the multiple interference type signal filters are mapped and embedded into the multi-channel filtering architecture to build a noise adaptive filtering multi-channel.

[0048] In this embodiment, the interference type-noise filtering algorithm is a set of adaptable algorithms that match different noise signal interference types such as harmonic interference, pulse interference, and power frequency drift interference, respectively, and include corresponding dedicated filtering algorithms such as notch filtering, amplitude limiting-median composite filtering, and adaptive notch filtering.

[0049] Specifically, firstly, based on the identified noise signal interference types, a multi-channel filtering architecture is designed to ensure a one-to-one correspondence between each filtering channel and the signal interference type. The specific types of noise signal interference are first identified, such as harmonic interference, impulse interference, and power frequency drift interference, and the quantity and characteristic differences of each type are statistically analyzed. A parallel multi-channel architecture is adopted, which includes a signal input module, a multi-channel filtering module, and a signal output fusion module. The number of channels in the multi-channel filtering module matches the number of identified interference types, and each channel is allocated independent signal processing resources to avoid mutual interference between different types of interference during filtering. To ensure accurate matching between channels and interference types, a mapping table of interference type-channel identifiers is pre-set in the architecture, clearly defining the interference processing target corresponding to each channel, ensuring that subsequent filter embedding can directly and specifically handle the corresponding type of interference.

[0050] Next, based on the type of noise interference, a suitable interference type-noise filtering algorithm is selected. Algorithms with stable filtering effects are chosen for the characteristics of different interference types: For harmonic interference, a notch filter algorithm is selected, which can accurately filter out harmonic components at specific frequencies with low computational complexity, making it compatible with the hardware processing capabilities of smart meters; for pulse interference, a composite algorithm combining amplitude limiting filtering and median filtering is selected. First, amplitude limiting filtering removes pulse signals exceeding a reasonable amplitude range, and then median filtering smooths signal fluctuations, avoiding signal distortion caused by a single algorithm; for power frequency drift interference, an adaptive notch filter algorithm is selected, which can dynamically track minute drifts in the power frequency, ensuring the stability of the filtering effect; for electromagnetic radiation interference, a low-pass filter algorithm is selected to filter out high-frequency radiation interference components. The initial values ​​of the core parameters for each algorithm are set according to the characteristics of the corresponding interference type. For example, the initial value of the center frequency of the notch filter is set to the identified harmonic peak frequency.

[0051] Then, based on the interference type-noise filtering algorithm, the interference type-noise filter is initialized. A clean voltage and current signal set is collected, and multi-scale noise is injected into it according to the identified noise signal interference type to obtain a voltage and current noise signal set. Subsequently, the clean voltage and current signal set and the voltage and current noise signal set are associated and identified to obtain a voltage and current signal training dataset. Finally, based on this training dataset, the initialized interference type-noise filter is trained with gain filtering to complete the construction of filters for multiple interference types. This step will be explained in detail later.

[0052] Finally, based on the pre-defined mapping table of interference type and channel identifier in the multi-channel filtering architecture, each trained interference type signal filter is connected to its corresponding filtering channel, achieving precise mapping between filters and channels. The interfaces between filters and channels are adapted to unify the data format and transmission rate of input and output signals, ensuring smooth signal transmission within the channels. After embedding, a connectivity test is performed on the entire multi-channel filtering architecture. Analog signals containing various interference types are input to verify whether the filters in each channel can start normally and process the corresponding interference. Simultaneously, it is checked whether the signal output fusion module can effectively integrate the processed signals from each channel to form a stable output signal, ultimately completing the construction of the noise adaptive filtering multi-channel.

[0053] Through the processes of architecture design, algorithm matching, gain training, and mapping embedding, combined with filter architecture design, filter algorithm, and parameter training methods, a noise adaptive filtering multi-channel system that is precisely adapted to the type of interference was built, providing reliable hardware architecture support for subsequent targeted noise filtering and improving the measurement accuracy of electrical variables in smart energy meters.

[0054] Furthermore, the method provided in this application embodiment includes: According to the interference type-noise filtering algorithm, initialize the interference type-noise filter; collect a pure voltage and current signal set, and inject multi-scale noise into the pure voltage and current signal set according to the noise signal interference type to obtain a voltage and current noise signal set; associate and identify the pure voltage and current signal set with the voltage and current noise signal set to obtain a voltage and current signal training dataset; perform gain filtering training on the interference type-noise filter based on the voltage and current signal training dataset to construct multiple interference type signal filters.

[0055] Specifically, firstly, the interference type-noise filtering algorithm determined in the preceding steps is used to initialize the interference type-noise filter. For different interference types and their corresponding filtering algorithms, an appropriate parameter initialization strategy is adopted: For the notch filter corresponding to harmonic interference, the initialization parameters include center frequency, quality factor Q, and initial gain. The initial center frequency is set to the peak harmonic frequency identified in the previous step of obtaining the noise signal interference type; the quality factor Q is set to 10 to balance filter selectivity and response speed; and the initial gain is set to 1. For the amplitude-limiting-median composite filter corresponding to impulse interference, the initial amplitude-limiting threshold is set to 1.5 times the maximum amplitude of the normal signal; the median filter window length is set to 5; odd-numbered windows can avoid signal offset. For the adaptive notch filter corresponding to power frequency drift interference, the initial center frequency is set to 50Hz, and the frequency tracking step size is set to 0.1Hz. For the low-pass filter corresponding to electromagnetic radiation interference, the initial cutoff frequency is set to 1kHz, and the initial gain is set to 1. The initialization of all filters is completed by writing parameters into the hardware configuration register, using a filter parameter initialization method to ensure that the initial state is stable and reusable.

[0056] Next, a pure voltage and current signal set is collected. The method for collecting this pure voltage and current signal set is the same as the method for obtaining the normal spectrum diagram of the electricity meter signal in the previous steps: select equipment of the same model and specifications as the target smart electricity meter, and collect grid voltage and current signals under different operating conditions, such as light load and full load, in a standard laboratory environment free from any interference. The sampling frequency is consistent with that of the target electricity meter, and the collection time is no less than 1 hour to ensure that the signal covers the typical state of normal grid operation, thus forming a pure voltage and current signal set. Multi-scale noise injection adopts the signal superposition method. For each type of interference, a standard interference signal of different intensity levels is generated by a signal generator: for example, for harmonic interference, characteristic harmonic signals of the 3rd, 5th, and 7th orders with an amplitude of 5%-20% of the fundamental amplitude of the pure signal are generated; for pulse interference, pulse signals with a pulse width of 10μs-1ms and an amplitude of 1.2-2 times the maximum amplitude of the pure signal are generated. Each type of interference corresponds to three intensity levels of interference signals, which are superimposed on different segments of the clean signal set. During superposition, the timing of the interference signals and the clean signals is synchronized, ultimately generating a voltage and current noise signal set containing multiple interference types and different intensities.

[0057] Then, using a time-series matching + label mapping association method, each signal segment in the pure voltage and current signal set is matched with its corresponding noise signal segment (i.e., the signal after interference is superimposed during the same time period) according to the time sequence of signal acquisition, ensuring the temporal consistency of the two. Next, an association label is added to each matched signal pair, including the interference type, interference intensity level, and signal acquisition condition (e.g., harmonic interference -10% - full load). The labels are in digital encoding format for easy retrieval and recognition during subsequent training. All labeled signal pairs are organized in an input-expected output format, with the voltage and current noise signal as the training input and the corresponding pure voltage and current signal as the expected output, ultimately forming a structured voltage and current signal training dataset.

[0058] Finally, based on the obtained voltage and current signal training dataset, the interference type-noise filter is trained using gain filtering. The training employs the minimum mean square error (MSE) algorithm, which has a fast convergence speed and is well-suited to the embedded hardware computing capabilities of smart meters. The specific training process is as follows: the voltage and current noise signals from the training dataset are input segment by segment into the corresponding initialization filter, and the mean square error between the filter's output signal and the desired output (pure voltage and current signals) is calculated. Based on the MSE value, the filter's gain parameters are iteratively adjusted using the gradient descent formula of the minimum mean square error algorithm, with an iteration step size of 0.01. The MSE is recalculated after each iteration. When the MSE is less than a preset convergence threshold, such as 0.001, the filter is successfully trained. When the iteration stops, the current gain parameters are saved. For each type of interference filter, the above training process is repeated until all filters meet the convergence condition, ultimately constructing multiple interference type signal filters adapted to different interference types.

[0059] By combining signal acquisition, superposition, and filtering training methods, the accurate construction of filters for different types of interference was achieved, ensuring that each filter can adapt to the characteristics of the corresponding interference, and providing a reliable core filtering unit for the subsequent construction of multi-channel filtering.

[0060] Furthermore, the method provided in this application embodiment includes: The front end of the noise adaptive filtering multi-channel is connected to a signal interference type identifier, which consists of a signal feature extractor and a signal interference classification head connected in series. The signal feature extractor is used to extract spectral features from the energy meter's collected signals to obtain spectral feature extraction parameters. The signal interference classification head is used to classify the spectral feature extraction parameters into interference types according to the energy meter's interference signal pattern library.

[0061] In one embodiment, the overall structure of the signal interference type identifier is first constructed. The specifications of the signal input and output interfaces of the identifier are determined first. The input interface is adapted to the transmission format of the signals collected by the smart energy meter, and the output interface is matched with the noise adaptive filtering multi-channel signal input module to ensure the compatibility and stability of signal transmission. The core of the signal interference type identifier consists of two functional modules: a signal feature extractor and a signal interference classification head. These two modules are constructed in series, meaning the output of the signal feature extractor is directly connected to the input of the signal interference classification head, forming a continuous processing chain from signal input to feature extraction to interference classification.

[0062] Next, a signal feature extractor is constructed. Its core function is to extract spectral features from the energy meter's acquired signal to obtain spectral feature extraction parameters. This functional module integrates a spectrum analysis unit and a feature parameter calculation unit. The spectrum analysis unit uses the Hanning window combined with 1024-point FFT spectral analysis parameters to perform frequency domain transformation on the input energy meter acquired signal, obtaining signal spectrum data. The feature parameter calculation unit, for the transformed signal spectrum data, also extracts the peak frequency using a peak detection algorithm, calculates the bandwidth using the half-width at half-maximum (WHM) method, obtains the amplitude distribution characteristics by statistically analyzing the amplitude distribution interval, and identifies harmonic components by combining the normal spectrum diagram of the energy meter signal obtained in the previous steps. Finally, these parameters are integrated to form the spectral feature extraction parameters, which are then output to the subsequent signal interference classification head through a standardized interface.

[0063] Then, a signal interference classification head is constructed to classify interference types based on the spectral feature extraction parameters according to the electricity meter interference signal pattern library. The classification head is built using a template matching module. First, the electricity meter interference signal pattern library constructed in the previous steps is imported into the storage unit of the classification head, establishing a mapping index between standard feature templates and interference types. When the spectral feature extraction parameters output by the signal feature extractor are received, the classification head first normalizes the parameters to eliminate dimensional differences. Then, using the Euclidean distance similarity calculation method, the normalized feature parameters are compared with each standard feature template in the pattern library. The standard template with the highest similarity is selected, and the corresponding interference type is determined based on the mapping index, forming the interference type classification result.

[0064] Finally, the signal feature extractor and the signal interference classification head were connected and adapted in series. The interfaces of these two functional modules were standardized, unifying data transmission rates and formats. A synchronous clock signal was used to achieve timing synchronization between the two, ensuring that the spectral feature extraction parameters output by the signal feature extractor could be transmitted to the signal interference classification head in real time and accurately. After the series connection was completed, the entire signal interference type identifier was functionally verified. Input signals from energy meters containing different interference types were used to verify its ability to accurately output the corresponding spectral feature extraction parameters and interference type classification results, ensuring the stable and reliable performance of the signal interference type identifier and its ability to properly interface with the front-end requirements of multi-channel noise adaptive filtering.

[0065] Furthermore, the method provided in this application embodiment includes: Based on the multi-channel filtering effect index parameters, the noise adaptive filtering multi-channel is dynamically optimized and analyzed to determine the multi-channel optimization direction parameter threshold; within the multi-channel optimization direction parameter threshold, the target multi-channel optimization direction parameter is obtained, and based on the target multi-channel optimization direction parameter, the noise adaptive filtering multi-channel is optimized and updated to generate the energy meter adaptive filtering multi-channel.

[0066] Optionally, firstly, dynamic optimization analysis of multi-channel noise adaptive filtering is performed based on multi-channel filtering effect index parameters. The core evaluation criteria for filtering effect are first clarified. Referring to the requirements for signal processing accuracy of electricity meters in GB / T 17215.211 "General Requirements, Tests and Test Conditions for AC Measuring Equipment", four basic qualification standards are set based on actual filtering needs: for example, signal-to-noise ratio not less than 45dB, total harmonic distortion rate (corresponding to the core indicator of signal distortion) not exceeding 0.5%, noise amplitude exceeding the standard rate not exceeding 5%, and average noise amplitude not exceeding 10% of the maximum amplitude of the normal signal. The multi-channel filtering effect index parameters obtained from the test, specifically including noise amplitude exceeding the standard rate, average noise amplitude, signal distortion, and signal-to-noise ratio, are compared one by one with the above four basic qualification standards to accurately locate the optimization direction. If the signal-to-noise ratio (SNR) is below standard, the optimization direction is to improve the accuracy of filter gain adjustment; if the total harmonic distortion (THD) exceeds the standard, the optimization direction is to refine the center frequency parameter of the notch filter; if the noise amplitude exceeds the standard, the optimization direction is to adjust the threshold judgment logic of the filter; if the average noise amplitude exceeds the standard, the optimization direction is to enhance the filter's attenuation capability for low-to-medium intensity noise. For each optimization direction, combined with the hardware computing power of the target smart meter and the resource constraints of the embedded system, parameter threshold ranges are determined through parameter sensitivity analysis: for example, the optimization threshold range for filter gain is set to 0.8-1.2, the optimization threshold range for the center frequency of the notch filter is set to the target frequency ±0.5Hz, the optimization range for the filter threshold judgment logic is set to 1.1-1.3 times the maximum amplitude of the normal signal, and the optimization range for the attenuation coefficient of low-to-medium intensity noise is set to 0.3-0.7, ensuring that the optimization parameters do not exceed the hardware operating limits, forming a set of parameter thresholds for multiple optimization directions.

[0067] Next, the multi-channel optimization parameter particle space is initialized according to the multi-channel optimization direction parameter threshold; then, the particle space is iteratively compared and optimized according to the multi-target filtering of the energy meter until the preset termination condition is met, and the optimal parameter particle containing the target multi-channel optimization direction parameter is obtained. This step will be explained in detail in the following content.

[0068] Finally, the optimization configuration update adopts a combination of register writing and software configuration file update. The target multi-channel optimization direction parameters are written one by one into the configuration registers of each filtering module of the noise adaptive filtering multi-channel, while the parameter configuration file in the meter main control program is updated to ensure persistent storage of parameters. This completes the optimization configuration of the noise adaptive filtering multi-channel and generates the electricity meter adaptive filtering multi-channel.

[0069] By combining index comparison, grid search optimization, and embedded system parameter configuration, precise optimization of multiple channels for noise adaptive filtering is achieved. The generated multi-channel adaptive filtering for electricity meters can better adapt to the actual operating conditions of the target smart meter, ensuring the stability of filtering effect and metering accuracy.

[0070] Furthermore, the method provided in this application embodiment includes: Based on the threshold of the multi-channel optimization direction parameter, the multi-channel optimization parameter particle space is initialized; the multi-channel optimization parameter particle space is iteratively compared and optimized according to the multi-target of the energy meter filter until the preset termination condition is met, and the optimal parameter particle is obtained. The optimal parameter particle includes the target multi-channel optimization direction parameter.

[0071] In one embodiment, firstly, a particle swarm optimization algorithm is used for basic parameter initialization. The physical meaning of each particle is defined, with each particle corresponding to a complete set of multi-channel optimization parameters. The particle dimension is consistent with the number of multi-channel optimization directions. For example, if the optimization directions include filter gain, notch center frequency, and median filter window length, the particle dimension is set to 3. Based on the determined multi-channel optimization direction parameter thresholds, the value range of each dimension of the particle is set, meaning that the parameters of each dimension are restricted within the parameter threshold of the corresponding optimization direction. The particle population size is initialized to 50, and an initial particle swarm is generated through uniform random sampling. The initial position of each particle is randomly determined within the parameter threshold range of its corresponding dimension. The initial velocity is set to 0.1 to ensure stable initial particle motion. Simultaneously, the individual optimal position and the global optimal position of the particle swarm are initialized. The individual optimal position is initialized to the initial position of each particle itself, and the global optimal position is initialized to the position of the particle with the best fitness in the initial particle swarm, thus completing the initialization of the multi-channel optimization parameter particle space.

[0072] Next, the multi-channel optimization parameters are iteratively compared and optimized according to the multi-objectives of the energy meter filtering. The multi-objectives are set as maximizing the signal-to-noise ratio (SNR), minimizing the total harmonic distortion (THD), and minimizing the noise amplitude exceeding the limit. Based on these multi-objectives, a fitness function is constructed, and a weighted summation method is used to transform the multi-objectives into single-objective optimization. The weights for SNR and THD are 0.4, THD and noise amplitude exceeding the limit are 0.3, and the fitness function value is calculated as follows: Fitness Function = (1 - Noise Amplitude Exceedance Rate) × 0.3 + (1 - THD) × 0.3 + (SNR / Maximum Reference SNR) × 0.4. A larger function value indicates a better filtering effect from the parameter combination. In each iteration, the parameter combination corresponding to each particle is configured into the noise adaptive filtering multi-channel, signal filtering tests are performed, and the corresponding filtering effect index parameters are calculated. These parameters are then substituted into the fitness function to obtain the fitness value for each particle.

[0073] Next, the current fitness value of each particle is compared with its own historical best fitness value. If the current value is better, the individual optimal position is updated to the current particle position. The individual optimal fitness values ​​of all particles are compared with the global optimal fitness value. If a better value exists, the global optimal position is updated to the corresponding individual optimal position. Then, the velocity of each particle is adjusted according to the velocity update formula of the particle swarm optimization algorithm. The velocity update takes into account the guiding role of the individual optimal position and the global optimal position, and the velocity update coefficient is set to 0.5. The particle position is then adjusted according to the updated velocity to ensure that the new particle position is still within the threshold range of the multi-channel optimization direction parameters. The preset termination condition is that the number of iterations reaches 100 or the change in the global optimal fitness value is less than 0.001 after 10 consecutive iterations. When either termination condition is met, the iteration stops. At this time, the particle corresponding to the global optimal position is the optimal parameter particle, and the parameters contained in this particle are the target multi-channel optimization direction parameters.

[0074] By using the initialization and iterative optimization process of the particle swarm optimization algorithm, combined with the design of the multi-objective fitness function for electricity meter filtering, the target multi-channel optimization direction parameters are accurately obtained under parameter threshold constraints. This ensures that the parameter combination can maximize the filtering effect and provides a scientific and reasonable parameter basis for subsequent multi-channel optimization configuration.

[0075] In summary, the adaptive filtering method for smart energy meters provided in this application has the following technical effects: This application assesses the noise level and identifies the interference type of the signals collected by smart energy meters, builds an adaptive multi-channel filter, and integrates and deploys it after parameter training and optimization. This accurately filters out various types of interference, improves the filtering effect and the accuracy of electrical variable measurement, and makes the energy meter more reliable. It achieves the technical effect of improving the accuracy of electrical variable measurement of energy meters and enhancing their adaptability to different interference scenarios.

[0076] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides an adaptive filtering system for smart energy meters, the system comprising: Voltage and current signal set acquisition module 1 is used to acquire the grid voltage and current signal set historically collected by the target smart meter, perform spectrum analysis and noise level assessment on the grid voltage and current signal set, and obtain signal noise spectrum dataset and noise level assessment results.

[0077] The filtering multi-channel construction module 2, if the noise level assessment result exceeds the preset noise threshold, performs interference type identification on the signal noise spectrum dataset to obtain the noise signal interference type, and builds a noise adaptive filtering multi-channel according to the noise signal interference type.

[0078] The filtering parameter acquisition module 3 is used to integrate and deploy the noise adaptive filtering multi-channel within the target smart meter to perform signal filtering tests and effect evaluations, and obtain multi-channel filtering effect index parameters.

[0079] The filtering control execution module 4 dynamically optimizes the noise adaptive filtering multi-channel based on the multi-channel filtering effect index parameters, generates the energy meter adaptive filtering multi-channel, and performs signal filtering closed-loop control through the energy meter adaptive filtering multi-channel.

[0080] Furthermore, the voltage and current signal acquisition module 1 is used to perform the following steps: The power grid voltage and current signal set is standardized and time-series identified to obtain a standard voltage and current sequence signal set. Based on the data characteristics and data analysis accuracy requirements of the standard voltage and current sequence signal set, spectrum analysis parameters are set, including window functions and FFT points. Spectrum analysis is performed on the standard voltage and current sequence signal set according to the spectrum analysis parameters to obtain a signal noise spectrum dataset. A noise level assessment index is defined, and the signal noise spectrum dataset is evaluated based on the noise level assessment index to obtain a noise level assessment result.

[0081] Furthermore, the voltage and current signal acquisition module 1 is used to perform the following steps: The standard voltage and current sequence signal set is subjected to spectral calculation according to the spectral analysis parameters to generate a voltage and current signal spectrum; a normal spectrum of the electricity meter signal is obtained, and features are extracted from the normal spectrum of the electricity meter signal to determine the normal signal feature threshold; based on the normal signal feature threshold, anomaly comparison and identification are performed on the voltage and current signal spectrum to mark the abnormal signal frequency point set; noise data is filtered and extracted from the voltage and current signal spectrum according to the abnormal signal frequency point set to obtain a signal noise spectrum dataset.

[0082] Furthermore, the filtering multi-channel construction module 2 is used to perform the following steps: Feature parameters are extracted from the signal and noise spectrum dataset to obtain a signal and noise spectrum feature parameter set, which includes peak frequency, bandwidth, amplitude distribution, and harmonic components. An energy meter interference signal pattern library is constructed based on the energy meter signal application standard. Similarity analysis is performed on the signal and noise spectrum feature parameter set based on the energy meter interference signal pattern library to obtain a signal interference pattern similarity set. Interference type identification and integration are performed on the energy meter interference signal pattern library according to the signal interference pattern similarity set to obtain the noise signal interference type.

[0083] Furthermore, the filtering multi-channel construction module 2 is used to perform the following steps: Based on the noise signal interference type, a multi-channel filtering architecture is designed, where each filtering channel in the multi-channel filtering architecture corresponds one-to-one with the signal interference type; an interference type-noise filtering algorithm is selected according to the noise signal interference type; filter gain training is performed based on the interference type-noise filtering algorithm to construct multiple interference type signal filters; the multiple interference type signal filters are mapped and embedded into the multi-channel filtering architecture to build a noise adaptive filtering multi-channel.

[0084] Furthermore, the filtering multi-channel construction module 2 is used to perform the following steps: According to the interference type-noise filtering algorithm, initialize the interference type-noise filter; collect a pure voltage and current signal set, and inject multi-scale noise into the pure voltage and current signal set according to the noise signal interference type to obtain a voltage and current noise signal set; associate and identify the pure voltage and current signal set with the voltage and current noise signal set to obtain a voltage and current signal training dataset; perform gain filtering training on the interference type-noise filter based on the voltage and current signal training dataset to construct multiple interference type signal filters.

[0085] Furthermore, the filtering multi-channel construction module 2 is used to perform the following steps: The front end of the noise adaptive filtering multi-channel is connected to a signal interference type identifier, which consists of a signal feature extractor and a signal interference classification head connected in series. The signal feature extractor is used to extract spectral features from the energy meter's collected signals to obtain spectral feature extraction parameters. The signal interference classification head is used to classify the spectral feature extraction parameters into interference types according to the energy meter's interference signal pattern library.

[0086] Furthermore, the filter control execution module 4 is used to perform the following steps: Based on the multi-channel filtering effect index parameters, the noise adaptive filtering multi-channel is dynamically optimized and analyzed to determine the multi-channel optimization direction parameter threshold; within the multi-channel optimization direction parameter threshold, the target multi-channel optimization direction parameter is obtained, and based on the target multi-channel optimization direction parameter, the noise adaptive filtering multi-channel is optimized and updated to generate the energy meter adaptive filtering multi-channel.

[0087] Furthermore, the filter control execution module 4 is used to perform the following steps: Based on the threshold of the multi-channel optimization direction parameter, the multi-channel optimization parameter particle space is initialized; the multi-channel optimization parameter particle space is iteratively compared and optimized according to the multi-target of the energy meter filter until the preset termination condition is met, and the optimal parameter particle is obtained. The optimal parameter particle includes the target multi-channel optimization direction parameter.

[0088] The adaptive filtering system for smart meters provided in this embodiment of the invention can execute the adaptive filtering method for smart meters provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0089] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0090] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An adaptive filtering method for smart energy meters, characterized in that, The method includes: The system acquires a set of historically collected grid voltage and current signals from the target smart meter, performs spectrum analysis and noise level assessment on the grid voltage and current signal set, and obtains a signal noise spectrum dataset and noise level assessment results. If the noise level assessment result exceeds the preset noise threshold, the interference type is identified in the signal noise spectrum dataset to obtain the noise signal interference type, and a noise adaptive filtering multi-channel is built according to the noise signal interference type. The noise adaptive filtering multi-channel is integrated and deployed in the target smart meter for signal filtering test and effect evaluation to obtain multi-channel filtering effect index parameters. Based on the multi-channel filtering effect index parameters, the noise adaptive filtering multi-channel is dynamically optimized to generate the energy meter adaptive filtering multi-channel, and the signal filtering closed-loop control is performed through the energy meter adaptive filtering multi-channel.

2. The adaptive filtering method for smart energy meters as described in claim 1, characterized in that, The signal-noise spectrum dataset and noise level assessment results are obtained, including: The power grid voltage and current signal set is standardized and time-series labeled to obtain a standard voltage and current sequence signal set. Based on the data characteristics and data analysis accuracy requirements of the standard voltage and current sequence signal set, spectrum analysis parameters are set, including window functions and FFT points. Perform spectral analysis on the standard voltage and current sequence signal set according to the spectral analysis parameters to obtain a signal noise spectral dataset. Define a noise level assessment index, and evaluate the signal noise spectrum dataset based on the noise level assessment index to obtain the noise level assessment result.

3. The adaptive filtering method for smart energy meters as described in claim 2, characterized in that, Obtain the signal-noise spectrum dataset, including: The standard voltage and current sequence signal set is subjected to spectral calculation according to the spectral analysis parameters to generate a voltage and current signal spectrum diagram. Obtain the normal spectrum diagram of the electricity meter signal, extract features from the normal spectrum diagram of the electricity meter signal, and determine the normal feature threshold of the signal; Based on the normal feature threshold of the signal, anomaly comparison and identification are performed on the voltage and current signal spectrum, and the abnormal signal frequency point set is marked. Based on the abnormal signal frequency point set, noise data is filtered and extracted from the voltage and current signal spectrum to obtain a signal noise spectrum dataset.

4. The adaptive filtering method for smart energy meters as described in claim 1, characterized in that, The types of noise signal interference are obtained, including: Feature parameters are extracted from the signal-noise spectrum dataset to obtain a signal-noise spectrum feature parameter set, which includes peak frequency, bandwidth, amplitude distribution, and harmonic components. Based on the application standards for electricity meter signals, a library of electricity meter interference signal patterns is constructed. Based on the energy meter interference signal pattern library, a similarity analysis is performed on the signal noise spectrum feature parameter set to obtain a signal interference pattern similarity set; The interference type of the electricity meter interference signal pattern library is identified and integrated according to the signal interference pattern similarity set to obtain the noise signal interference type.

5. The adaptive filtering method for smart energy meters as described in claim 1, characterized in that, Build a multi-channel adaptive noise filter, including: Based on the noise signal interference type, a multi-channel filtering architecture is designed, and each filtering channel in the multi-channel filtering architecture corresponds one-to-one with the signal interference type. Based on the type of noise signal interference, select the interference type-noise filtering algorithm; Based on the aforementioned interference type-noise filtering algorithm, filter gain training is performed to construct multiple interference type signal filters; The multiple interference type signal filters are mapped and embedded into the filtering multi-channel architecture to build a noise adaptive filtering multi-channel.

6. The adaptive filtering method for smart energy meters as described in claim 5, characterized in that, Construct filters for multiple types of interference signals, including: Initialize the interference type-noise filter according to the interference type-noise filtering algorithm; Collect a pure voltage and current signal set, and inject multi-scale noise into the pure voltage and current signal set according to the noise signal interference type to obtain a voltage and current noise signal set; The pure voltage and current signal set is associated and identified with the voltage and current noise signal set to obtain a voltage and current signal training dataset. Based on the voltage and current signal training dataset, the gain filtering training of the interference type-noise filter is performed to construct multiple interference type signal filters.

7. The adaptive filtering method for smart energy meters as described in claim 4, characterized in that, The front end of the noise adaptive filtering multi-channel is connected to a signal interference type identifier, which consists of a signal feature extractor and a signal interference classification head connected in series. The signal feature extractor is used to extract spectral features from the energy meter's collected signals to obtain spectral feature extraction parameters. The signal interference classification head is used to classify the spectral feature extraction parameters into interference types according to the energy meter's interference signal pattern library.

8. The adaptive filtering method for smart energy meters as described in claim 1, characterized in that, Generate an adaptive filtering multi-channel for the electricity meter, including: Based on the multi-channel filtering effect index parameters, the noise adaptive filtering multi-channel is dynamically optimized and analyzed to determine the multi-channel optimization direction parameter threshold. The target multi-channel optimization direction parameters are obtained within the threshold of the multi-channel optimization direction parameters, and the noise adaptive filtering multi-channel is optimized and updated based on the target multi-channel optimization direction parameters to generate the energy meter adaptive filtering multi-channel.

9. The adaptive filtering method for smart energy meters as described in claim 8, characterized in that, The target multi-channel optimization direction parameters are obtained by optimizing within the threshold range of the multi-channel optimization direction parameters, including: Initialize the multi-channel optimization parameter particle space based on the multi-channel optimization direction parameter threshold. The multi-channel optimization parameter particle space is iteratively compared and optimized according to the multi-objective filtering of the electricity meter until a preset termination condition is met, and the optimal parameter particle is obtained. The optimal parameter particle includes the target multi-channel optimization direction parameter.

10. An adaptive filtering system for smart energy meters, characterized in that, The system is used to implement the adaptive filtering method for smart energy meters according to any one of claims 1-9, the system comprising: The voltage and current signal set acquisition module is used to acquire the grid voltage and current signal set historically collected by the target smart meter, perform spectrum analysis and noise level assessment on the grid voltage and current signal set, and obtain the signal noise spectrum dataset and noise level assessment results. The filtering multi-channel construction module, if the noise level assessment result exceeds the preset noise threshold, performs interference type identification on the signal noise spectrum dataset to obtain the noise signal interference type, and builds a noise adaptive filtering multi-channel based on the noise signal interference type; The filtering parameter acquisition module is used to integrate and deploy the noise adaptive filtering multi-channel within the target smart meter to perform signal filtering tests and effect evaluations, and obtain multi-channel filtering effect index parameters. The filtering control execution module dynamically optimizes the noise adaptive filtering multi-channel based on the multi-channel filtering effect index parameters, generates the energy meter adaptive filtering multi-channel, and performs signal filtering closed-loop control through the energy meter adaptive filtering multi-channel.

Images