Error dynamic evaluation and aging early warning method, device and equipment for intelligent electric energy meter

Abstract

The invention provides an intelligent electric energy meter error dynamic evaluation and aging early warning method, device and equipment, and relates to the technical field of electric power system metering. According to the method, a three-dimensional time sequence matrix of short-term, medium-term and long-term time sequence features is constructed, and a two-dimensional convolutional neural network CNN and a bidirectional long-short term memory network Bi-LSTM are combined to extract coupling features and time sequence dependency relationship features, so that multi-parameter collaborative modeling and time sequence dependency analysis are realized; the parameter weights of the errors are distributed in a self-adaptive mode so as to carry out error prediction evaluation; and then dynamically evaluating the aging state of the electricity meter based on an aging index calculation method of parameter drift. According to the method, the error prediction precision and the time sequence dynamic early warning capability can be remarkably improved, the accuracy is improved in a high-load and low-power-factor scene, the early warning delay is shortened from 72 hours to 4 hours, and the offline verification workload is reduced.

Description

Technical Field

[0001] This invention relates to the field of power system metering technology, and more specifically, to a method, device, and equipment for dynamic evaluation of errors and early warning of aging of smart energy meters. Background Technology

[0002] In smart grid metering scenarios, ordinary smart meters (non-environmental monitoring type) are limited by their functionality, only able to collect electrical parameters such as voltage, current, and power factor, but unable to acquire environmental data such as temperature, humidity, and air pressure. According to data from the State Grid Corporation of China, these types of meters account for over 70% of the existing meter inventory. However, their metering errors are significantly affected by various electrical factors such as current fluctuations, power factor changes, and voltage distortion. Currently, error evaluation methods for these meters mainly rely on offline verification or threshold judgment based on a single electrical parameter.

[0003] In existing technologies, the single-parameter threshold method is a common implementation approach, such as triggering an error warning when the current exceeds 120% of the rated value. However, this method ignores the interaction between multiple parameters. For example, when the power factor is simultaneously below 0.5, the error increase may reach 1.5 times the effect of a single factor, resulting in insufficient warning accuracy. Another existing technology is a static model based on electrical parameter statistics, which constructs an evaluation system by calculating indicators such as the mean voltage and the standard deviation of current. Although this method can reflect changes in electrical parameters to some extent, it ignores the time-series correlation between parameters. For example, the cumulative effect of error after three consecutive days of high-load operation is not fully considered. Furthermore, while traditional time-series models (such as ARIMA) can capture the historical patterns of the error itself, they cannot effectively correlate the immediate impact of changes in electrical parameters on the error, further limiting their predictive capabilities.

[0004] In view of the above, this application is hereby submitted. Summary of the Invention

[0005] This invention aims to provide a method, device, and equipment for dynamic evaluation of errors and early warning of aging in smart energy meters. It addresses the problems in existing technologies, such as the lack of multi-parameter coupling analysis, insufficient utilization of time-series dynamic features, and reliance on fixed-cycle verification for meter aging evaluation. By deeply mining the time-series characteristics and coupling relationships of electrical parameters such as voltage, current, and power factor, a high-precision metering error evaluation model is constructed.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: A method for dynamic error assessment and aging early warning of smart energy meters includes: S1: Real-time acquisition of electrical parameter time-series data and construction of short-term, medium-term and long-term time-series features to obtain electrical parameter time-series matrix; S2, based on the electrical parameter time series matrix, coupling features are extracted through a convolutional neural network, and time-dependent features are extracted through a bidirectional long short-term memory network based on the coupling features; S3. An attention mechanism is used to adaptively allocate parameter weights for error prediction by combining the coupling features and the temporal dependency. S4. Based on the electrical parameter time series matrix, calculate the parameter drift rate, and calculate the aging index by fitting the annual error growth rate obtained from several recent verification error prediction data through linear regression. This index is used to trigger an early warning when the aging index exceeds a set threshold.

[0007] Preferably, the electrical parameters include voltage, current, and power factor; S1 specifically includes: After arranging the electrical parameter time series data in chronological order, short-term time series features, medium-term time series features, and long-term time series features are constructed. The short-term time-series characteristics are: the maximum, minimum and fluctuation data of electrical parameters collected within a sliding window of hour length; The intermediate time series feature is: the cumulative deviation data of electrical parameters collected within a sliding window with a length of days; The long-term time series characteristics are: the standard deviation and trend data of electrical parameters collected within a sliding window of a month.

[0008] Preferably, the electrical parameter time series matrix is ​​used to extract coupling features through a convolutional neural network, specifically: The electrical parameter time series matrix, composed of short-term, medium-term, and long-term time series features, is processed by a two-dimensional convolutional neural network through sliding calculation. The correlation between voltage, current, and power factor is processed simultaneously to extract the features of the interaction of electrical parameters, i.e., coupling features.

[0009] Preferably, the bidirectional long short-term memory network consists of a forward LSTM and a backward LSTM, which respectively extract the forward hidden state features of the electrical parameter data. and backward hidden state features Forward hidden state features The calculation formula is as follows: ; in, This indicates the output of the input gate at the current moment; This represents the Sigmoid activation function; This represents the weight matrix of the input gate; This represents the input data for the input gate; This represents the input features at the current moment, i.e., the coupling features; Represents the hidden state features of the input gate; This represents the forward hidden state feature of the previous time step; The input gate weight matrix represents the hidden state features of the previous time step; This represents the bias vector of the input gate; This indicates the output of the forget gate at the current moment; The weight matrix representing the forget gate; The forget gate weight matrix represents the hidden state features of the previous time step; This represents the input data for the forget gate; The bias vector representing the forget gate; This indicates the output of the output gate at the current moment; This represents the weight matrix of the output gate; This represents the input data of the output gate; The output gate weight matrix represents the hidden state features of the previous time step; This represents the bias vector of the output gate; Indicates the current state of the cell; A weight matrix representing the state of candidate cells; The candidate cell state weight matrix represents the hidden state features of the previous time step; A bias vector representing the state of a candidate cell; This represents element-wise multiplication; Indicates the activation function; Similarly, the features of the backward hidden state are calculated. ; Final splicing output This is the temporal dependency feature, where ⊕ represents concatenation.

[0010] Preferably, the parameter importance weight vector is defined as follows: The parameter weights of the error Generate by calculating the correlation between the hidden state and the parameter features: ; ; in, The hidden state features of the bidirectional long short-term memory network at the current moment are the temporal dependency features. This represents the current electrical parameter vector; , , This is the corresponding weight matrix; This represents the attention parameter weights after softmax normalization; for The bias vector; The parameter weights represent the error at time k; This represents the total number of predicted moments; This represents the transpose symbol.

[0011] Preferably, the expression for error prediction evaluation is: Combining attention parameter weights With eigenvectors, error prediction values The expression is: ; in, This is the predicted error value; , These are the prediction layer weights and biases, respectively. The hidden state features of the bidirectional long short-term memory network at the current moment; This represents the Sigmoid activation function.

[0012] Preferably, the formula for calculating the aging index is: ; in, This is the aging index; the higher the value, the more severe the aging. The parameter drift rate is used to reflect the rate at which the mean of an electrical parameter deviates from its initial value. This is the annual error growth rate, used to reflect the annual growth trend of parameter error accumulation over time; This is a balancing coefficient used to adjust the contribution ratio of parameter drift rate to the annual error growth rate. Parameter drift rate The deviation rate between the long-term average value and the initial value of the electrical parameters is expressed as: ; in, Indicates the parameter drift rate; This represents the average value of electrical parameters over the past 12 months. Indicates the initial calibration value of the electrical parameters; Annual growth rate of error It is achieved by fitting the time series trend of recent verification error data through linear regression, and is expressed as: ; in, The regression coefficients are obtained by fitting the linear relationship between the error and the number of months using the least squares method. , The dates are the months of the first and last verification error data, respectively.

[0013] Preferably, it also includes updating model parameters through online iterative optimization and incremental learning mechanisms.

[0014] This invention also provides a device for dynamic evaluation of errors and early warning of aging in smart energy meters, comprising: The time-series feature construction unit is used to collect electrical parameter time-series data in real time and construct three types of time-series features: short-term, medium-term, and long-term, to obtain the electrical parameter time-series matrix. The feature extraction unit is used to extract coupling features based on the electrical parameter time series matrix through a convolutional neural network, and to extract time-series dependency features based on the coupling features through a bidirectional long short-term memory network. An error prediction unit is used to adaptively allocate parameter weights of the error by employing an attention mechanism, combining the coupling features and the temporal dependency, in order to perform error prediction evaluation; An aging warning unit is used to calculate the parameter drift rate based on the electrical parameter time series matrix, and to calculate the aging index by fitting the annual error growth rate obtained from several recent verification error prediction data through linear regression. The unit is used to trigger a warning when the aging index exceeds a set threshold.

[0015] The present invention also provides a smart energy meter error dynamic assessment and aging early warning device, including a processor and a memory. The memory stores a computer program, which can be executed by the processor to realize the smart energy meter error dynamic assessment and aging early warning method as described above.

[0016] The present invention also provides a computer-readable storage medium storing computer-readable instructions, which, when executed by a processor of the device on which the computer-readable storage medium is located, implement the above-described method for dynamic error assessment and aging early warning of a smart energy meter.

[0017] In summary, compared with the prior art, the present invention has the following beneficial effects: This invention significantly improves the accuracy of multi-parameter coupling analysis by constructing a three-dimensional time-series matrix, extracting multi-parameter interaction features using a two-dimensional CNN, and capturing temporal dependencies using Bi-LSTM. The root mean square error of the error prediction is reduced from 0.75% to 0.52%, with a 28% improvement in accuracy under high-load, low-power-factor scenarios. Furthermore, the dynamic prediction capability is enhanced, enabling prediction of error exceeding tolerance risks up to 48 hours in advance, reducing the warning delay from 72 hours to 4 hours. In particular, the aging assessment model based on parameter drift can replace traditional offline verification, reducing the verification workload by 30%, while the aging index AI achieves an 89% agreement rate with actual disassembly and inspection results.

[0018] In summary, this invention addresses the shortcomings of traditional methods in error prediction and aging assessment by introducing multi-parameter coupling analysis, time-series dynamic modeling, and aging assessment mechanisms. It significantly improves the metering accuracy and operation and maintenance efficiency of ordinary smart meters and has broad application prospects. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a method for dynamic evaluation of errors and early warning of aging in a smart energy meter, as provided in Example 1.

[0021] Figure 2 This is a schematic diagram of a smart energy meter error dynamic assessment and aging early warning device provided in Embodiment 2.

[0022] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention 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 the present invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0024] Example 1 Embodiment 1 of the present invention provides a method for dynamic evaluation of errors and aging warning of smart energy meters, which can be implemented by a device for dynamic evaluation of errors and aging warning of smart energy meters (hereinafter referred to as aging warning device), specifically, executed by one or more processors within the aging warning device.

[0025] In this embodiment, the aging warning device can be an electronic device equipped with a processor. The processor carries a computer program for the dynamic evaluation of smart meter error and aging warning method, and the computer program can be executed. For example, a computer, smartphone, smart tablet, workstation, etc. are not limited here.

[0026] like Figure 1 As shown, a method for dynamic evaluation of errors and early warning of aging of smart energy meters includes steps S1 to S4.

[0027] S1 collects electrical parameter time-series data in real time and constructs three types of time-series features: short-term, medium-term, and long-term, to obtain the electrical parameter time-series matrix.

[0028] Furthermore, the electrical parameters include (U), current (I), and power factor (cosφ). In practical applications, smart meters are typically deployed at the end of the power grid, responsible for collecting key electrical parameters such as voltage, current, and power factor, and transmitting the data to a backend server for processing via a built-in communication module. This invention first collects real-time meter operating data, including parameters such as voltage accuracy of 0.1V, current accuracy of 0.01A, power factor accuracy of 0.01, and frequency accuracy of 0.1Hz, with a sampling frequency set to 1 time / minute. Simultaneously, historical error verification data is also recorded; for example, the three most recent error values ​​will serve as important references for subsequent model training.

[0029] Based on this, the electrical parameter time-series data are arranged in chronological order to form a two-dimensional matrix of time × parameter (which can be understood as the time dimension expansion of a three-dimensional matrix). For example, with a time window of 1 day and sampling once per minute, the matrix dimension is 1440 × 3, and each row corresponds to the parameter combination at a certain moment. , , ].

[0030] For example, time 1: [220V, 5A, 0.9]; time 2: [218V, 6A, 0.85]; ... Next, short-term time series features, medium-term time series features, and long-term time series features are constructed.

[0031] Short-term time-series characteristics are: the maximum, minimum and fluctuation data of electrical parameters collected within a sliding window of hours (e.g., 1 hour) to capture transient parameter fluctuations (e.g., a sudden increase in current during peak load periods).

[0032] The medium-term time-series characteristics are: cumulative deviation data of electrical parameters collected within a sliding window of days, to analyze parameter coupling patterns within the daily cycle (such as the combination of high load and low power factor during the day). For example, analyzing the cumulative deviation of parameters during peak load periods (such as 18:00-22:00) within the daily cycle.

[0033] The long-term time series characteristics are: the standard deviation and trend data of electrical parameters (such as the monthly increase rate of current) collected within a sliding window of a month.

[0034] These temporal features form the basis of the three-dimensional temporal matrix, providing data support for subsequent multi-parameter coupling analysis.

[0035] S2, based on the electrical parameter time series matrix, coupling features are extracted through a convolutional neural network, and time-dependent features are extracted through a bidirectional long short-term memory network based on the coupling features.

[0036] The core of this step is to construct and process a three-dimensional time-series matrix, which arranges voltage U, current I, and power factor cosφ in time order to form a two-dimensional matrix of time × parameters.

[0037] To achieve multi-parameter collaborative modeling, a two-dimensional convolutional neural network (CNN) (e.g., using a 3×3 convolutional kernel with a stride of 1) is employed to perform sliding calculations on the matrix. This allows for the simultaneous processing of the correlation between voltage, current, and power factor, extracting the characteristics of the interactive influence of electrical parameters—i.e., coupling features—rather than analyzing individual parameters independently. For example, when a parameter combination of "current exceeding the rated value by 80% and power factor below 0.7" is detected, the CNN automatically identifies it as a "high load, low power factor" state and quantifies its amplification factor K for the error. For instance, in a real-world scenario, when voltage fluctuates by ±5% and current exceeds the rated value, K=1.2, meaning the error is 1.2 times the effect of a single factor.

[0038] Of course, a "parameter interaction coefficient K" can be generated through model training to characterize the impact of multiple factors on the error. For example, when the voltage fluctuates by ±5% and the current exceeds the rated value, K=1.2, meaning the error is 1.2 times the effect of a single factor, thus solving the problem of the inability to quantify the coupling effect of multiple parameters in existing technologies.

[0039] Existing technologies judge errors solely based on a single parameter threshold (such as current exceeding limits), without considering the synergistic effects between parameters. This solution utilizes a three-dimensional time-series matrix combined with a two-dimensional CNN. (1) Multi-parameter collaborative modeling: Simultaneously analyze the spatiotemporal correlation of U, I, and cosφ, which improves the error prediction accuracy by 25% in high-load and low-power-factor scenarios; (2) Dynamic feature recognition: Real-time capture of changes in parameter combinations (such as sudden error changes during load mode switching), providing early warning of error exceedance risk 4 hours earlier than traditional methods.

[0040] In addition, a bidirectional long short-term memory network is introduced to capture the temporal correlation of parameter changes. By combining the bidirectional long short-term memory network (Bi-LSTM) and the attention mechanism, the time-series dependency of electrical parameters (Bi-LSTM) and the dynamic allocation of parameter importance (attention layer) are processed simultaneously to achieve dynamic error prediction without environmental data.

[0041] The bidirectional long short-term memory network consists of a forward LSTM and a backward LSTM, which extract the forward hidden state features of the electrical parameter data. and backward hidden state features Forward hidden state features The calculation formula is as follows: ; in, This indicates the output of the input gate at the current moment; This represents the Sigmoid activation function; This represents the weight matrix of the input gate (which can be initialized by Xavier and then updated by stochastic gradient descent). This represents the input data for the input gate; This represents the input features at the current moment, i.e., the coupling features; Represents the hidden state features of the input gate; This represents the forward hidden state feature of the previous time step; The input gate weight matrix represents the hidden state features of the previous time step; This represents the bias vector of the input gate; This indicates the output of the forget gate at the current moment; The weight matrix representing the forget gate; The forget gate weight matrix represents the hidden state features of the previous time step; This represents the input data for the forget gate; The bias vector representing the forget gate; This indicates the output of the output gate at the current moment; This represents the weight matrix of the output gate; This represents the input data of the output gate; The output gate weight matrix represents the hidden state features of the previous time step; This represents the bias vector of the output gate; Indicates the current state of the cell; A weight matrix representing the state of candidate cells; The candidate cell state weight matrix represents the hidden state features of the previous time step; A bias vector representing the state of a candidate cell; This represents element-wise multiplication; Indicates the activation function; Similarly, the features of the backward hidden state are calculated. ; Final splicing output This is the temporal dependency feature, where ⊕ represents concatenation.

[0042] For example, after learning the pattern of "current exceeding the rated value for 3 consecutive days + power factor continuously decreasing", the possible growth rate of error in the next 24 hours can be predicted.

[0043] S3. An attention mechanism is used to adaptively allocate parameter weights for error prediction by combining the coupling features and the temporal dependency.

[0044] This step aims to further improve the accuracy and dynamic adaptability of error prediction. First, the electrical parameter time-series matrix is ​​input into a CNN layer to extract coupling features, and then a Bi-LSTM layer is used to capture temporal dependencies. Next, an attention layer is introduced to assign parameter weights, dynamically adjusting the weights based on the parameter's contribution to the error. For example, when voltage fluctuations exceed ±7%, the attention layer automatically increases the voltage feature weight to 35%, while the weight is 20% under normal conditions. By generating weight vectors through the correlation between hidden states and parameter features, the model can more accurately reflect the importance of different parameters in specific scenarios. Finally, the error prediction layer outputs the predicted error value, combining the attention weights and feature vectors to achieve dynamic error modeling. For example, when a "sudden current surge → sudden power factor drop" time-series pattern occurs for three consecutive days, the model can predict the error increase by 0.2% 48 hours in advance, reducing the prediction delay from 72 hours to 4 hours compared to the traditional ARIMA model.

[0045] Specifically, the parameter importance weight vector is defined as follows: The parameter weights of the error It is generated by calculating the correlation between the hidden state and the parameter features, using the following formula: ; ; in, The hidden state features of the bidirectional long short-term memory network at the current moment are the temporal dependency features. This represents the current electrical parameter vector; , , This is the corresponding weight matrix; This represents the attention parameter weights after softmax normalization; for The bias vector; The parameter weights represent the error at time k; This represents the total number of predicted moments; This represents the transpose symbol.

[0046] The expression for error prediction evaluation is: Combining attention parameter weights The expression for the error prediction value, along with the eigenvector, is: ; in, This is the predicted error value; , These are the prediction layer weights and biases, respectively. The hidden state features of the bidirectional long short-term memory network at the current moment; This represents the Sigmoid activation function.

[0047] S4. Based on the electrical parameter time series matrix, calculate the parameter drift rate, and calculate the aging index by fitting the annual error growth rate obtained from several recent verification error prediction data through linear regression. This index is used to trigger an early warning when the aging index exceeds a set threshold.

[0048] This step utilizes an aging index AI based on parameter drift to achieve online assessment of the aging status of electricity meters. First, the parameter drift rate is defined as the deviation rate between the long-term operating parameter mean and the initial value, and is calculated using the parameter drift rate calculation formula. By leveraging the degree of drift in the long-term time-series characteristics of electrical parameters, an aging index AI is constructed to replace traditional offline verification, enabling online assessment of the aging status of electricity meters.

[0049] The formula for calculating the aging index is as follows: ; in, This is the aging index; the higher the value, the more severe the aging. The parameter drift rate is used to reflect the rate at which the mean of an electrical parameter deviates from its initial value. This is the annual error growth rate, used to reflect the annual growth trend of parameter error accumulation over time; This is a balancing coefficient used to adjust the contribution ratio of parameter drift rate to the annual error growth rate. Parameter drift rate The deviation rate between the long-term average value and the initial value of the electrical parameters is expressed as: ; in, Indicates the parameter drift rate; This represents the average value of electrical parameters over the past 12 months. Indicates the initial calibration value of the electrical parameters; Annual growth rate of error It is achieved by fitting the time series trend of recent verification error data through linear regression, and is expressed as: ; in, The regression coefficients are obtained by fitting the linear relationship between the error and the number of months using the least squares method. , The dates are the months of the first and last verification error data, respectively.

[0050] For example, if a certain type of electricity meter drifts by 8% in its average current measurement after 5 years of operation, then the parameter drift rate is 8%. Simultaneously, the annual error growth rate can be calculated by fitting the time series trend of the error data from the last three calibrations using linear regression. For instance, if the errors of a certain electricity meter in the last three calibrations were 0.2%, 0.3%, and 0.4%, then the annual error growth rate is 0.1% / year. When the aging index AI is greater than 0.7, a replacement warning is triggered, corresponding to a remaining lifespan of less than 1 year. For example, if a certain electricity meter has a parameter drift rate of 8% and an annual error growth rate of 0.5% after 5 years of operation, then AI = 0.6 × 8% + 0.4 × 0.5% = 5.0%. Actual disassembly and inspection show that its error exceeds the tolerance risk by 70%, which traditional methods cannot identify in advance. However, this solution provides a warning 2 years in advance using the AI ​​value.

[0051] In another preferred embodiment, to ensure the model can continuously adapt to feature drift caused by meter aging, the method further includes updating model parameters through online iterative optimization and incremental learning mechanisms. 100,000 newly collected electrical parameter-error samples are automatically imported monthly, and the model parameters are updated using gradient descent. For example, after 100,000 new samples are added to a certain region, the weight matrices of the CNN and Bi-LSTM layers are automatically adjusted, significantly improving the model's adaptability to the new scenario. This online iterative optimization mechanism not only improves the model's robustness but also reduces the need for manual intervention.

[0052] The practical application scenarios of this invention cover a variety of complex operating conditions in smart grids. For example, in a pilot project of a provincial power grid company, this invention was successfully applied to error monitoring and aging assessment of 100,000 ordinary smart meters. Pilot results showed that the accuracy of multi-parameter coupled analysis was significantly improved, with the root mean square error of error prediction decreasing from 0.75% to 0.52%, and the accuracy improving by 28% under high load and low power factor scenarios. Simultaneously, the time-series dynamic prediction capability was enhanced, enabling prediction of error exceedance risks 48 hours in advance, reducing the early warning delay from 72 hours to 4 hours. Furthermore, the aging assessment model based on parameter drift achieved a consistency rate of 89%, reducing offline verification workload by 30% and significantly lowering operation and maintenance costs.

[0053] In summary, compared with the prior art, the present invention has the following beneficial effects: This invention addresses the shortcomings of traditional methods in error prediction and aging assessment by introducing multi-parameter coupling analysis, time-series dynamic modeling, and aging assessment mechanisms, thereby significantly improving the metering accuracy and operation and maintenance efficiency of ordinary smart energy meters.

[0054] Example 2 like Figure 2 As shown, the second embodiment of the present invention also provides a smart energy meter error dynamic assessment and aging early warning device, comprising: The time-series feature construction unit is used to collect electrical parameter time-series data in real time and construct three types of time-series features: short-term, medium-term, and long-term, to obtain the electrical parameter time-series matrix. The feature extraction unit is used to extract coupling features based on the electrical parameter time series matrix through a convolutional neural network, and to extract time-series dependency features based on the coupling features through a bidirectional long short-term memory network. An error prediction unit is used to adaptively allocate parameter weights of the error by employing an attention mechanism, combining the coupling features and the temporal dependency, in order to perform error prediction evaluation; An aging warning unit is used to calculate the parameter drift rate based on the electrical parameter time series matrix, and to calculate the aging index by fitting the annual error growth rate obtained from several recent verification error prediction data through linear regression. The unit is used to trigger a warning when the aging index exceeds a set threshold.

[0055] Example 3 The third embodiment of the present invention also provides a smart energy meter error dynamic assessment and aging early warning device, which includes a memory and a processor. The memory stores a computer program, which can be executed by the processor to realize the smart energy meter error dynamic assessment and aging early warning method as described above.

[0056] Example 4 The fourth embodiment of the present invention also provides a computer-readable storage medium storing computer-readable instructions. When the computer-readable instructions are executed by the processor of the device where the computer-readable storage medium is located, the computer-readable instructions implement the above-described method for dynamic evaluation of smart meter errors and aging warning.

[0057] In the several embodiments provided in this invention, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus and method embodiments described above are merely illustrative. For example, the flowcharts in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the invention. In this regard, each block in the flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0058] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0059] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. In the absence of further restrictions, an element defined by the phrase "comprising a..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0060] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0061] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0062] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0063] The use of "first" and "second" in the embodiments is merely to distinguish similar objects and does not represent a specific ordering of objects. It is understood that "first" and "second" can be interchanged in a specific order or sequence where permitted. It should be understood that the objects distinguished by "first" and "second" can be interchanged where appropriate so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0064] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for dynamic evaluation of errors and early warning of aging in smart energy meters, characterized in that, include: S1: Real-time acquisition of electrical parameter time-series data and construction of short-term, medium-term and long-term time-series features to obtain electrical parameter time-series matrix; S2, based on the electrical parameter time series matrix, coupling features are extracted through a convolutional neural network, and time-dependent features are extracted through a bidirectional long short-term memory network based on the coupling features; S3. An attention mechanism is used to adaptively allocate parameter weights for error prediction by combining the coupling features and the temporal dependency. S4. Based on the electrical parameter time series matrix, calculate the parameter drift rate, and calculate the aging index by fitting the annual error growth rate obtained from several recent verification error prediction data through linear regression. This index is used to trigger an early warning when the aging index exceeds a set threshold.

2. The method for dynamic evaluation of errors and aging early warning of smart energy meters according to claim 1, characterized in that... The electrical parameters include voltage, current, and power factor; S1 specifically refers to: After arranging the electrical parameter time series data in chronological order, short-term time series features, medium-term time series features, and long-term time series features are constructed. The short-term time-series characteristics are: the maximum, minimum and fluctuation data of electrical parameters collected within a sliding window of hour length; The intermediate time series feature is: the cumulative deviation data of electrical parameters collected within a sliding window with a length of days; The long-term time series characteristics are: the standard deviation and trend data of electrical parameters collected within a sliding window of a month.

3. The method for dynamic evaluation of errors and aging early warning of smart energy meters according to claim 2, characterized in that... The electrical parameter time-series matrix is ​​used to extract coupling features through a convolutional neural network, specifically: The electrical parameter time series matrix, composed of short-term, medium-term, and long-term time series features, is processed by a two-dimensional convolutional neural network through sliding calculation. The correlation between voltage, current, and power factor is processed simultaneously to extract the features of the interaction of electrical parameters, i.e., coupling features.

4. The method for dynamic evaluation of errors and aging early warning of smart energy meters according to claim 2, characterized in that... The bidirectional long short-term memory network consists of a forward LSTM and a backward LSTM, which extract the forward hidden state features of the electrical parameter data. and backward hidden state features Forward hidden state features The calculation formula is as follows: ； in, This indicates the output of the input gate at the current moment; This represents the Sigmoid activation function; This represents the weight matrix of the input gate; This represents the input data for the input gate; This represents the input features at the current moment, i.e., the coupling features; Represents the hidden state features of the input gate; This represents the forward hidden state feature of the previous time step; The input gate weight matrix represents the hidden state features of the previous time step; This represents the bias vector of the input gate; This indicates the output of the forget gate at the current moment; The weight matrix representing the forget gate; The forget gate weight matrix represents the hidden state features of the previous time step; This represents the input data for the forget gate; The bias vector representing the forget gate; This indicates the output of the output gate at the current moment; This represents the weight matrix of the output gate; This represents the input data of the output gate; The output gate weight matrix represents the hidden state features of the previous time step; This represents the bias vector of the output gate; Indicates the current state of the cell; A weight matrix representing the state of candidate cells; The candidate cell state weight matrix represents the features of the forward hidden state at the previous time step; A bias vector representing the state of a candidate cell; This represents element-wise multiplication; Indicates the activation function; Similarly, the features of the backward hidden state are calculated. ; The final concatenated output shows the temporal dependency features. : ； In this context, ⊕ represents splicing.

5. The method for dynamic evaluation of errors and aging early warning of smart energy meters according to claim 4, characterized in that... Define the parameter importance weight vector as follows: The parameter weights of the error Generate by calculating the correlation between the hidden state and the parameter features: ； ； in, The hidden state features of the bidirectional long short-term memory network at the current moment are the temporal dependency features. This represents the current electrical parameter vector; , , This is the corresponding weight matrix; This represents the attention parameter weights after softmax normalization; for The bias vector; The parameter weights represent the error at time k; This represents the total number of predicted moments; This represents the transpose symbol.

6. The method for dynamic evaluation of errors and aging early warning of smart energy meters according to claim 5, characterized in that... The expression for error prediction evaluation is: Combining attention parameter weights With eigenvectors, error prediction values The expression is: ； in, This is the predicted error value; , These are the prediction layer weights and biases, respectively. The hidden state features of the bidirectional long short-term memory network at the current moment; This represents the Sigmoid activation function.

7. The method for dynamic evaluation of errors and aging early warning of smart energy meters according to claim 1, characterized in that... The formula for calculating the aging index is as follows: ； in, This is the aging index; the higher the value, the more severe the aging. The parameter drift rate is used to reflect the rate at which the mean of an electrical parameter deviates from its initial value. This is the annual error growth rate, used to reflect the annual growth trend of parameter error accumulation over time; This is a balancing coefficient used to adjust the contribution ratio of parameter drift rate to the annual error growth rate. Parameter drift rate The deviation rate between the long-term average value and the initial value of the electrical parameters is expressed as: ； in, Indicates the parameter drift rate; This represents the average value of electrical parameters over the past 12 months. Indicates the initial calibration value of the electrical parameters; Annual growth rate of error It is achieved by fitting the time series trend of recent verification error data through linear regression, and is expressed as: ； in, The regression coefficients are obtained by fitting the linear relationship between the error and the number of months using the least squares method. , The dates are the months of the first and last verification error data, respectively.

8. The method for dynamic evaluation of errors and aging early warning of smart energy meters according to claim 1, characterized in that... It also includes updating model parameters through online iterative optimization and incremental learning mechanisms.

9. A device for dynamic evaluation of errors and early warning of aging in smart energy meters, characterized in that, include: The time-series feature construction unit is used to collect electrical parameter time-series data in real time and construct three types of time-series features: short-term, medium-term, and long-term, to obtain the electrical parameter time-series matrix. The feature extraction unit is used to extract coupling features based on the electrical parameter time series matrix through a convolutional neural network, and to extract time-series dependency features based on the coupling features through a bidirectional long short-term memory network. An error prediction unit is used to adaptively allocate parameter weights of the error by employing an attention mechanism, combining the coupling features and the temporal dependency, in order to perform error prediction evaluation; An aging warning unit is used to calculate the parameter drift rate based on the electrical parameter time series matrix, and to calculate the aging index by fitting the annual error growth rate obtained from several recent verification error prediction data through linear regression. The unit is used to trigger a warning when the aging index exceeds a set threshold.

10. A device for dynamic evaluation of errors and early warning of aging in smart energy meters, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that can be executed by the processor to implement a method for dynamic evaluation of errors and early warning of aging of a smart energy meter as described in any one of claims 1-8.

Images