Method and system for predicting single-phase electric energy meter burnout based on electric meter data

Abstract

The present invention discloses a method and system for predicting single-phase electric energy meter burnout based on electric meter data, which relates to the field of intelligent early warning technology for single-phase electric energy meters. The present invention not only obtains a three-dimensional array by analyzing the historical time series data of the single-phase electric energy meter, but also obtains a three-dimensional array by analyzing the historical time series data of the single-phase electric energy meter. X reshaped , and further extracts a variety of engineering features from a multi-dimensional perspective to obtain a multi-dimensional array X' eng This provides a more comprehensive and robust basis for risk assessment, helping to improve the model's generalization capabilities. Furthermore, the present invention provides a burnout prediction model that combines the advantages of deep learning and machine learning, fully exploiting the inherent patterns in electricity meter data to achieve highly accurate predictions of burnout risks. This invention addresses the limitations and ineffectiveness of existing single-phase electricity meter prediction technologies.

Description

Technical Field

[0001] The present invention relates to the field of intelligent early warning technology for single-phase electricity meters, and more specifically, to: 1. A method for predicting single-phase electricity meter burnouts based on electricity meter data; 2. A system for predicting single-phase electricity meter burnouts based on electricity meter data. The present invention utilizes machine learning methods to analyze time-series data such as voltage and current of single-phase electricity meters to predict the risk of single-phase electricity meter burnout. The present invention aims to improve the reliability and safety of single-phase electricity meter operation, reduce economic losses and safety hazards caused by electricity meter failures, and provide data support and decision-making basis for the intelligent operation and maintenance of power grids. Background Art

[0002] Electricity meters are essential devices used in power systems to measure electricity consumption. However, over long-term operation, various factors, such as current overloads, short circuits, loose connections, environmental factors (such as humidity and high temperatures), and manufacturing defects, can cause internal components (such as shunts, relays, and main control chips) to overheat and burn out. This can lead to meter failure, inaccurate measurement, and even serious safety accidents such as fires.

[0003] Single-phase energy meters are a common type of energy meter. Currently, troubleshooting and identifying burnout meters rely primarily on manual inspections, user reports, or simple threshold alarms. These methods have the following limitations:

[0004] 1. High hysteresis: Manual inspections are inefficient and cannot monitor in real time. Users usually report repairs after the meter has already failed. Repair responses are delayed, making it difficult to achieve early warning and preventive maintenance.

[0005] 2. Inefficiency: With a huge number of electricity meters, especially after the popularization of large-scale smart meters, it is unrealistic to rely entirely on manual inspections, which consumes a lot of manpower and material resources.

[0006] 3. False alarms and missed alarms: Simple current or voltage threshold alarms often fail to distinguish between normal fluctuations and abnormal trends, which can easily lead to false alarms and increase operation and maintenance costs. At the same time, some hidden faults or initial degradation phenomena may be missed, and potential meter burnout risks may not be discovered in time.

[0007] Although with the development of big data and artificial intelligence technologies in recent years, some studies have attempted to use machine learning methods to analyze single-phase electricity meter data to achieve fault diagnosis and prediction, the actual effects of these methods are not ideal. Summary of the Invention

[0008] Based on this, it is necessary to address the limitations and poor effects of existing single-phase electricity meter prediction technology, and provide a method and system for predicting single-phase electricity meter burnout based on meter data.

[0009] The present invention is achieved by adopting the following technical solutions:

[0010] In a first aspect, the present invention discloses a method for predicting a single-phase electric energy meter burnout based on electric meter data, comprising:

[0011] Step 1: Obtain historical time series data of the target single-phase electric energy meter;

[0012] Step 2: Preprocess historical time series data into a three-dimensional array X reshaped , based on X reshaped Perform multi-dimensional feature processing to obtain a multi-dimensional array X' eng ;

[0013] Step three, X reshaped 、 X' eng Input the trained burnout prediction model for processing to obtain the prediction result indicating whether the target single-phase electricity meter will burn out. R .

[0014] Among them, the burn table prediction model includes: LSTM branch, fully connected branch, and burn table prediction part.

[0015] LSTM branch is used for: Based on the three-level LSTM network X reshaped Processing to obtain a time series representation H a Among them, the second-level LSTM network is equipped with an AL layer for dynamically allocating time-step weights.

[0016] The fully connected branch is used for: X' eng Processed to obtain engineering representation H b .

[0017] Burning table prediction unit is used to: Combine H a 、 H b Process to obtain R .

[0018] This method for predicting single-phase electricity meter burnout based on electricity meter data implements the method or process according to an embodiment of the present disclosure.

[0019] In a second aspect, the present invention discloses a system for predicting single-phase electric energy meter burnout based on electric energy meter data, which uses the method for predicting single-phase electric energy meter burnout based on electric energy meter data disclosed in the first aspect.

[0020] The system for predicting the burning of a single-phase electric energy meter based on electric meter data includes: a data acquisition module, a data processing module, and a burning meter prediction module.

[0021] The data acquisition module is used to obtain the historical time series data of the target single-phase electric energy meter. The data processing module is used to pre-process the historical time series data into a three-dimensional array. X reshaped , based on X reshaped Perform multi-dimensional feature processing to obtain a multi-dimensional array X' eng The burn prediction module is used to: X reshaped 、 X' eng Input the trained burnout prediction model for processing to obtain the prediction result indicating whether the target single-phase electricity meter will burn out. R .

[0022] This system for predicting single-phase electricity meter burnout based on electricity meter data implements the method or process according to the embodiments of the present disclosure.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. The present invention not only obtains a three-dimensional array by analyzing the historical time series data of the single-phase electric energy meter X reshaped , and further extracts a variety of engineering features from a multi-dimensional perspective to obtain a multi-dimensional array X' eng , provides a more comprehensive and robust risk assessment basis for the model, which helps to improve the generalization ability of the model; and the present invention provides a method combining deep learning (LSTM branch processing X reshaped ) and machine learning (fully connected branch processing X' eng ) The meter burning prediction model with advantages can fully explore the inherent laws of electricity meter data and achieve high-precision prediction of meter burning risks.

[0025] 2. The model of the present invention introduces designs such as batch normalization, residual connection, attention mechanism and Dropout processing, which enhances the overall learning ability and noise resistance of the model, so that it can still maintain stable performance and strong robustness in complex and changing data environments.

[0026] 3. The present invention can achieve early prediction of potential meter burn risks, thereby providing power companies with sufficient warning time, implementing preventive maintenance, and avoiding passive handling after a fault occurs, with significant economic and social benefits.

[0027] 4. The entire prediction process of the present invention can be fully automated, which greatly improves the efficiency and intelligence level of electricity meter operation and maintenance and reduces labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0029] Figure 1 A flowchart of a method for predicting single-phase electric energy meter burnout based on electric meter data provided in Example 1 of the present invention;

[0030] Figure 2 for Figure 1 The structural diagram of the prediction model of the medium burning table;

[0031] Figure 3 for Figure 2 Structural diagram of the LSTM branch in

[15] ;

[0032] Figure 4 for Figure 3 Structural diagram of the middle AL layer;

[0033] Figure 5 for Figure 2 Structural diagram of the fully connected branches in

[15] . DETAILED DESCRIPTION

[0034] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0035] It should be noted that when a component is referred to as being "mounted on" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "set on" another component, it may be directly set on the other component or there may be a central component. When a component is considered to be "fixed to" another component, it may be directly fixed to the other component or there may be a central component.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used herein in the specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "or / and" as used herein includes any and all combinations of one or more of the associated listed items.

[0037] First, regarding the method mentioned in the background technology of using machine learning methods to analyze single-phase electricity meter data to achieve fault diagnosis and prediction, we analyzed it and found the reasons why it is not effective:

[0038] 1. Mainly focus on identifying faults that have already occurred, or are relatively simple in feature engineering (i.e., only focus on time series data of electrical parameters such as voltage, live line current, and neutral line current).

[0039] 2. The model architecture and design used when processing time series data are unreasonable, resulting in poor actual performance.

[0040] Based on the above reasons, the present invention aims to develop a method that can fully utilize the multi-dimensional time series data of single-phase electricity meters, combine the advantages of deep learning and traditional feature engineering, and achieve high-precision and early prediction of the burnout risk of single-phase electricity meters.

[0041] Example 1

[0042] See Figure 1 , which is a flow chart of a method for predicting single-phase electric energy meter burnout based on electric energy meter data provided in this embodiment 1. It should be noted that burnout refers to the phenomenon that the internal components of the electric energy meter are damaged due to current overload or line short circuit.

[0043] A major innovation of this method is that it not only considers historical time series data, but also extracts multi-dimensional engineering features based on historical time series data, providing a more comprehensive and robust basis for risk assessment; another major innovation is that it provides a method that combines deep learning (LSTM branch processing X reshaped ) and machine learning (fully connected branch processing X' eng ) The meter burning prediction model with advantages can fully explore the inherent laws of electricity meter data and achieve high-precision prediction of meter burning risks.

[0044] like Figure 1 As shown, the method includes the following steps:

[0045] Step 1: Obtain historical time series data of the target single-phase electricity meter.

[0046] Historical time series data includes: Δ T Every Δ tThe voltage collected (i.e., a series of time steps) t Corresponding voltage value U t , t ∈[1,Δ T / Δ t ]), live wire current (i.e., a series of time steps t Corresponding live wire current value I t ), neutral current (i.e., a series of time steps t Corresponding live wire current value Ⅱ t ).

[0047] Among them, Δ T is the preset continuous duration; Δ t is the preset collection period. Considering the balance between prediction performance and computing resources, in this embodiment 1, Δ T It is recommended to take 24h, Δ t It is recommended to take 15 minutes, so there are 96 time steps in total, and each time step contains 3 electrical parameter values ​​(i.e. the corresponding voltage value, live wire current value, and neutral wire current value). T , Δ t It can also be adjusted according to actual conditions, but sufficient prediction performance must be guaranteed.

[0048] Step 2: Preprocess historical time series data into a three-dimensional array X reshaped , based on X reshaped Perform multi-dimensional feature processing to obtain a multi-dimensional array X' eng .

[0049] It should be noted that since the target single-phase electricity meter itself does not have data processing capabilities and needs to upload data to the cloud or server, this process is prone to abnormal situations such as data loss and duplication. Therefore, it is necessary to preprocess the historical time series data. Generally, the preprocessing process includes:

[0050] S101: Perform data repair on historical time series data. For example, if missing values ​​occur, the median of adjacent data is used to fill them in; if duplicate values ​​occur, they are removed.

[0051] S102: Normalize the historical time series data processed in S101. Generally, it is recommended to use the MinMaxScaler method to classify the data into [0, 1] to eliminate the influence of different physical dimensions and numerical ranges.

[0052] S103, reshape the historical time series data processed by S102 into a three-dimensional format to obtainX reshaped .

[0053] in, , N Indicates the number of samples.

[0054] It should be noted that N During model training, it is the total number of samples contained in the sample data set, but in step 1, it is the number of target single-phase electricity meters, that is, 1.

[0055] After the above processing, X reshaped That is, it meets the input requirements of subsequent models.

[0056] As mentioned above, it is not enough to consider only time series data, so step 2 starts from X reshaped Further extract multi-dimensional engineering features - these features are related to the operating status of the electricity meter and the risk of burning the meter. Of course, not all of these engineering features need to be input into the subsequent model, but the ones with the greatest contribution are selected. M A feature - reducing computing consumption while ensuring data validity.

[0057] Specifically, the multi-dimensional feature processing method includes:

[0058] S201, based on X reshaped Extract multi-dimensional engineering features.

[0059] Among them, multi-dimensional engineering features include: impedance characteristics, power characteristics, statistical characteristics, correlation characteristics, and frequency domain characteristics.

[0060] ①. The process of extracting impedance characteristics is as follows:

[0061] Calculate time steps t Live wire impedance , neutral line impedance , and correspondingly form the live line impedance sequence and the neutral line impedance sequence;

[0062] Calculate relevant features based on the live wire impedance sequence, including but not limited to the live wire impedance mean , live line impedance standard deviation, live line impedance maximum, live line impedance minimum; calculate relevant features based on the neutral line impedance sequence, including but not limited to the neutral line impedance mean , zero line impedance standard deviation, zero line impedance maximum value, zero line impedance minimum value; and further calculate the impedance mean ratio , average difference .

[0063] ② The power feature extraction process is as follows:

[0064] Calculate time steps t FireWire power , neutral line power , and form the live wire power sequence and the neutral wire power sequence accordingly;

[0065] Calculate relevant features based on the live power sequence, including but not limited to the live power mean , live line power standard deviation, live line power maximum, live line power minimum; calculate relevant features based on the zero line power sequence, including but not limited to the zero line power mean , zero line power standard deviation, zero line power maximum value, zero line power minimum value; and further calculate the power mean ratio , average difference .

[0066] ③. The statistical feature extraction process is as follows:

[0067] calculate X reshaped Statistics such as mean, standard deviation, range, voltage interquartile range, skewness, kurtosis, root mean square (RMS), crest factor, and zero crossing rate (defined as the number of times the signal crosses the mean) of the medium voltage;

[0068] calculate X reshaped Statistics such as the mean, standard deviation, range, voltage interquartile range, skewness, kurtosis, root mean square, crest factor, and zero crossing rate of the live wire current;

[0069] calculate X reshaped Statistics such as the mean, standard deviation, range, voltage interquartile range, skewness, kurtosis, root mean square, crest factor, and zero crossing rate of the neutral current.

[0070] ④. The extraction process of relevant features is as follows:

[0071] calculate X reshaped Pearson correlation coefficients between medium voltage and live current, voltage and neutral current, and live current and neutral current;

[0072] calculate X reshaped The mean, standard deviation, and maximum absolute value of the difference between the live and neutral currents.

[0073] ⑤. The process of extracting frequency domain features is as follows:

[0074] right X reshapedThe medium voltage, live line current, and neutral line current are converted to the frequency domain (generally using fast Fourier transform), and the average value, standard deviation, and maximum value of their amplitude spectra are extracted respectively; then the power spectrum density is calculated (using the Welch method), and its average value and maximum value are extracted.

[0075] S202, select the front from the multi-dimensional engineering features M The features with higher contribution are used as reference engineering features X eng .

[0076] The contribution degree characterizes the correlation between the feature and the risk of burnout: the higher the contribution degree, the higher the risk of burnout.

[0077] In this embodiment 1, the ANOVA F-value method is combined with the SelectKBest method to select the top M The feature with higher contribution is selected to remove redundant and irrelevant features. In this embodiment 1, M is recommended to be 50.

[0078] S203, yes X eng Perform Z-score normalization to obtain X' eng .

[0079] in, , N Indicates the number of samples.

[0080] Obtained by Z-score standardization X' eng ——Its mean is 0 and standard deviation is 1, which helps improve the training efficiency and stability of subsequent models.

[0081] Similar to the above, N During model training, it is the total number of samples contained in the sample data set, but in step 2, it is the number of target single-phase electricity meters, that is, 1.

[0082] Step three, X reshaped 、 X' eng Input the trained burnout prediction model for processing to obtain the prediction result indicating whether the target single-phase electricity meter will burn out. R .

[0083] See Figure 2 ,The burn table prediction model includes : LSTM branch, fully connected branch, and burn table prediction part.

[0084] ①, LSTM branch is used for: based on the three-level LSTM network X reshapedProcessing to obtain a time series representation H a .

[0085] It should be emphasized that the second-level LSTM network is equipped with an AL layer for dynamically allocating time step weights, which can improve the processing effect of the LSTM branch.

[0086] like Figure 3 As shown in the figure, the LSTM branch includes: 1st to 3rd level LSTM networks; the output of the previous level LSTM network serves as the input of the next level LSTM network.

[0087] Among them, the structure of the first-level LSTM network is the same as that of the third-level LSTM network; the second-level LSTM network is equivalent to adding an AL layer on the basis of the first-level LSTM network.

[0088] Specifically:

[0089] I. The i-th level LSTM network includes: 1 LSTM layer, 1 BN layer, 1 Dense layer, 1 residual connection layer, and 1 Dropout layer; i∈{1, 3}.

[0090] In the i-th level LSTM network: the LSTM layer is used to i Input of the LSTM network in i The BN layer is used to perform batch normalization on the output of the LSTM layer; the Dense layer is used to in i Perform dimension adjustment; the residual connection layer is used to add the output of the Dense layer and the output of the BN layer; the Dropout layer is used to process the residual connection layer to prevent overfitting and obtain the output of the i-th level LSTM network out i .

[0091] Then we have: in 1 for X reshaped ; out 3 for H a .

[0092] Of course, the processing of the i-th level LSTM network can be expressed as:

[0093] ;

[0094] Where, LSTM (.) indicates LSTM layer; BN (.) indicates BN layer; Dense (.)express Dense layer; Dropout (.) indicates the Dropout layer.

[0095] II. The second-level LSTM network includes: 1 LSTM layer, 1 BN layer, 1 AL layer, 1 Dense layer, 1 residual connection layer, and 1 Dropout layer.

[0096] In the second-level LSTM network: The LSTM layer is used as the input to the second-level LSTM network in 2. Perform long-term dependency learning; the BN layer is used to perform batch normalization on the output of the LSTM layer and serve as the input of the AL layer Input ; AL layer is used based on Input Dynamically calculate time step weights a , and Input Weighted output of AL layer Output ; Dense layer is used to in 2 for dimension adjustment; the residual connection layer is used for the output of the Dense layer, Output Add; the Dropout layer is used to process the residual connection layer to prevent overfitting and obtain the output of the second-level LSTM network out 2.

[0097] It is important to note that the AL layer dynamically calculates the time-step attention score through a trainable weight matrix, enabling the model to focus on more important time points in the sequence. Figure 4 , including: 1 attention calculation layer, 1 Softmax layer, and 1 product calculation layer.

[0098] In the AL layer: the attention calculation layer is used to calculate the Input Calculate the time step attention score e ; The softmax layer is used to activate the softmax function e Processed into a ; The product calculation layer is used to convert a and Input Multiply to get Output .

[0099] in, e The calculation formula is: e = tanh ( W · Input + b ); tanh (.)express tanh function; W 、 b These are all learnable parameters.

[0100] Of course, the processing of the second-level LSTM network can be expressed as:

[0101] ;

[0102] ;

[0103] Where, LSTM (.) indicates LSTM layer; BN (.) indicates BN layer; AL (.) indicates AL layer; Dense (.)express Dense layer; Dropout (.) indicates the Dropout layer.

[0104] In addition, in the LSTM branch, the number of neurons in the three LSTM layers are 128, 64, and 32, respectively, and the dropout rates of the three dropout layers are 0.2, 0.2, and 0.1, respectively.

[0105] In summary, the processing of the LSTM branch can be expressed as follows:

[0106] ;

[0107] Where, LSTM 3 represents the third-level LSTM network; LSTM 2 represents the second-level LSTM network; LSTM 1 represents the first-level LSTM network.

[0108] ②、Fully connected branch is used for: Based on two-level fully connected network X' eng Processed to obtain engineering representation H b .

[0109] like Figure 5 As shown in the figure, the fully connected branch includes: the 1st to 2nd level fully connected networks; the output of the previous level fully connected network serves as the input of the next level fully connected network.

[0110] The j-th level fully connected network includes: 1 Dense layer, 1 BN layer, and 1 Dropout layer; j∈{1, 2}.

[0111] In the j-th level fully connected network: the Dense layer is used as the input to the j-th level fully connected network Come j The BN layer is used to perform batch normalization on the output of the Dense layer; the Dropout layer is used to process the BN layer to prevent overfitting and obtain the input of the j-th fully connected network. Goj ;

[0112] Then we have: Come 1 for X' eng ; Go 2 for H b .

[0113] Of course, the processing of the j-th level fully connected network can be expressed as:

[0114] ;

[0115] Where, BN (.) indicates BN layer; Dense (.)express Dense layer; Dropout (.) indicates the Dropout layer.

[0116] In summary, the processing of the fully connected branch can be expressed as follows:

[0117] ;

[0118] Where, FC 2(.) indicates the second-level fully connected network; FC 1(.) represents the first-level fully connected network.

[0119] In addition, in the fully connected branch, the number of neurons in the two Dense layers is 64 and 32 respectively, and the Dropout rate of the two Dropout layers is 0.3.

[0120] ③、The burn prediction unit is used to: combine H a 、 H b Process to obtain R .

[0121] like Figure 2 As shown in the figure, the burn table prediction part includes: 1 splicing layer, 2 Dense layers, and 1 Sigmoid layer.

[0122] In the burn table prediction section: the splicing layer is used to H a 、 H b The first Dense layer is used to weight and bias the output of the splicing layer; the second Dense layer is used to weight and bias the output of the first Dense layer; the Sigmoid layer is used to process the output of the second Dense layer through the Sigmoid activation function to obtain R.

[0123] Of course, the processing of the burn prediction unit can be expressed as follows:

[0124] ;

[0125] Where, Sigmoid (.) indicates Sigmoid layer; Dense (.)express Dense layer; Concat (.) indicates a concatenated layer.

[0126] It should be noted that R is the risk probability of burning the meter, R ∈[0,1]. The probability threshold is set to 0.5: if R >0.5, it means there is a risk of burning the meter; otherwise, there is no risk of burning the meter.

[0127] At this point, the prediction of single-phase electricity meter burnout based on meter data in the target substation has been realized.

[0128] In addition, in step 3, if R If the target single-phase energy meter is at risk of burning out, an alarm is triggered and / or a maintenance work order is generated. Otherwise, the process returns to step 1 or ends. This allows for decision-making based on the prediction results to optimize grid operations and maintenance.

[0129] Of course, this method uses a trained burn prediction model. The training method of the model is as follows:

[0130] Training is performed based on a sample dataset with true labels (i.e., whether the watch is burned) and a cross-entropy loss function with class weights is constructed. Loss , the model is trained using the Adam optimizer combined with the cosine annealing learning rate scheduler, and an early stopping strategy is applied to prevent overfitting until the training is completed.

[0131] Among them, the attenuation formula of the cosine annealing learning rate scheduler is:

[0132] ;

[0133] Where, represents the real-time learning rate; represents the initial learning rate; represents the minimum learning rate; K Indicates the total number of training times.

[0134] That is, the initial learning rate of training is 0.001 and the minimum learning rate is 0.0001. This scheduling strategy helps the model converge quickly in the early stages of training and fine-tune the weights in the later stages.

[0135] It should be noted that the purpose of adding category weights to the cross entropy loss function is to make the model pay more attention to the burned samples - because the number of burned samples is generally much smaller than that of normal samples, which will lead to category imbalance problems.

[0136] The formula for category weight is: ;

[0137] Where, ClassWeight k Indicates the k The category weights of the categories; TotalSamples Represents the total number of all samples in the training set (divided from the sample data set); NumClasses Indicates the total number of categories (2 in this method, i.e., unburned and burned tables); SamplesInClass k Indicates the k The number of samples for each category.

[0138] Then we have: ;

[0139] Where, y k Indicates the indicator of the kth category in the true label (i.e., if the sample belongs to the k If there are multiple categories, then it is 1; otherwise, it is 0); Indicates that the model predicts that the sample belongs to k The probability of a category.

[0140] Simulation Verification

[0141] In order to illustrate the effect of this method, this embodiment 1 carried out a model performance test on this method, and the final results are shown in Table 1 below.

[0142] Table 1 Model performance results

[0143]

[0144] As shown in Table 1, the overall accuracy of this model reached 71%, demonstrating good predictive accuracy. Specifically, this model performed well in identifying unburned watches, achieving a high recall rate. It also achieved a high prediction precision of 80% for burned watches. While maintaining this precision, it also achieved a recall rate of 57% for burned watch samples, providing a foundation for subsequent early warning and maintenance efforts.

[0145] Example 2

[0146] This embodiment 2 provides a system for predicting single-phase electricity meter burnout based on electricity meter data, which uses the method for predicting single-phase electricity meter burnout based on electricity meter data provided in embodiment 1.

[0147] The system for predicting the burning of a single-phase electric energy meter based on electric meter data includes: a data acquisition module, a data processing module, and a burning meter prediction module.

[0148] The data acquisition module is used to obtain historical time-series data from the target single-phase energy meter. It is recommended that the data acquisition module be installed at the energy meter end—it can be built into the target single-phase energy meter or connected to the target single-phase energy meter.

[0149] The data processing module is used to preprocess historical time series data into a three-dimensional array X reshaped , based on X reshaped Perform multi-dimensional feature processing to obtain a multi-dimensional array X' eng The data processing module is recommended to be set up in the cloud or server side, and to communicate with the data acquisition module, so as to process the historical time series data transmitted by the data acquisition module.

[0150] Burning table prediction module is used to: X reshaped 、 X' eng Input the trained burnout prediction model for processing to obtain the prediction result indicating whether the target single-phase electricity meter will burn out. R The burn prediction module is also recommended to be set up in the cloud or server side, so as to obtain the X reshaped 、 X' eng Make burn predictions.

[0151] Since this system uses the method for predicting single-phase electricity meter burnout based on electricity meter data in Example 1, it also has the same effect and will not be repeated here.

[0152] Example 3

[0153] This embodiment 3 discloses a computer device including a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the steps of the method for predicting single-phase electricity meter burnout based on meter data disclosed in embodiment 1 are implemented.

[0154] Computer devices can be either mobile or fixed terminals. Examples of the former include mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals). Examples of the latter include digital TVs and desktop computers.

[0155] This embodiment 3 also discloses a readable storage medium, in which computer program instructions are stored. When the computer program instructions are read and executed by a processor, the steps of the method for predicting single-phase electricity meter burnout based on meter data disclosed in embodiment 1 are executed.

[0156] Among them, the readable storage medium may include, but is not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0157] This embodiment 3 further discloses a computer program product, including a computer program. When the computer program is executed by a processor, the steps of the method for predicting a burnout of a single-phase electric energy meter based on electric meter data disclosed in embodiment 1 are implemented.

[0158] It should be noted that the computer program for executing the above-mentioned operations can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The above-mentioned computer program can be executed entirely on the user's computer, partially on the user's computer, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN).

[0159] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the patent for this invention shall be determined by the appended claims.

Claims

1. A method for predicting single-phase electric energy meter burnout based on electric meter data, characterized in that: include: Step 1: Obtain historical time series data of the target single-phase electric energy meter; Historical time series data includes: Δ T Every Δ t The collected voltage, live current, and neutral current; among them, Δ T is the preset continuous duration; Δ t is the preset collection period; Step 2: Preprocess historical time series data into a three-dimensional array X reshaped , based on X reshaped Perform multi-dimensional feature processing to obtain a multi-dimensional array X' eng ; The multi-dimensional feature processing method in step 2 includes: S201, based on X reshaped Extract multi-dimensional engineering features: Among them, multi-dimensional engineering features include: impedance features, power features, statistical features, correlation features, and frequency domain features; statistical features include: based on X reshaped Statistics calculated for medium voltage, live current, and neutral current respectively; S202, select the front from the multi-dimensional engineering features M The features with higher contribution are used as reference engineering features X eng ; S203, yes X eng Perform Z-score normalization to obtain X' eng ;in, , N Indicates the number of samples; Step three, X reshaped 、 X' eng Input the trained burnout prediction model for processing to obtain the prediction result indicating whether the target single-phase electricity meter will burn out. R ; Among them, the burn prediction model includes: LSTM branch, which is used for: based on the three-level LSTM network X reshaped Processing to obtain a time series representation H a The second-level LSTM network is equipped with an AL layer for dynamically allocating time-step weights. The AL layer includes: an attention calculation layer, a softmax layer, and a product calculation layer. The fully connected branch is used to: X' eng Processed to obtain engineering representation H b ;as well as Burning table prediction unit, which is used to: combine H a 、 H b Process to obtain R .

2. The method for predicting single-phase electric energy meter burnout based on electric meter data according to claim 1, characterized in that: The pre-processing methods in step 2 include: First, perform data repair on the historical time series data, then normalize it, and then reshape it into a three-dimensional format to obtain X reshaped ;in, , N Indicates the number of samples.

3. The method for predicting single-phase electric energy meter burnout based on electric meter data according to claim 1, characterized in that: The LSTM branch includes: 1st to 3rd level LSTM networks; the output of the previous level LSTM network serves as the input of the next level LSTM network; The i-th level LSTM network includes: 1 LSTM layer, 1 BN layer, 1 Dense layer, 1 residual connection layer, and 1 Dropout layer; i∈{1, 3}; In the i-th level LSTM network: the LSTM layer is used to i Input of the LSTM network in i The BN layer is used to perform batch normalization on the output of the LSTM layer; the Dense layer is used to in i Perform dimension adjustment; the residual connection layer is used to add the output of the Dense layer and the output of the BN layer; the Dropout layer is used to process the residual connection layer to prevent overfitting and obtain the output of the i-th level LSTM network out i ;in, in 1 for X reshaped ; out 3 for H a ; The second-level LSTM network includes: 1 LSTM layer, 1 BN layer, 1 AL layer, 1 Dense layer, 1 residual connection layer, and 1 Dropout layer; in the second-level LSTM network: the LSTM layer is used as the input to the second-level LSTM network in 2. Perform long-term dependency learning; the BN layer is used to perform batch normalization on the output of the LSTM layer and serve as the input of the AL layer Input ; AL layer is used based on Input Dynamically calculate time step weights a , and Input Weighted output of AL layer Output ; Dense layer is used to in 2 for dimension adjustment; the residual connection layer is used for the output of the Dense layer, Output Add; the Dropout layer is used to process the residual connection layer to prevent overfitting and obtain the output of the second-level LSTM network out 2.

4. The method for predicting single-phase electric energy meter burnout based on electric meter data according to claim 3, characterized in that: In the AL layer: the attention calculation layer is used to calculate the Input Calculate the time step attention score e ; The softmax layer is used to activate the softmax function e Processed into a ; The product calculation layer is used to convert a and Input Multiply to get Output ; in, e The calculation formula is: e = tanh ( W · Input + b ); tanh (.)express tanh function; W 、 b These are all learnable parameters.

5. The method for predicting single-phase electric energy meter burnout based on electric meter data according to claim 1, characterized in that: The fully connected branch includes: 1st to 2nd level fully connected networks; the output of the previous level fully connected network serves as the input of the next level fully connected network; The j-th level fully connected network includes: 1 Dense layer, 1 BN layer, and 1 Dropout layer; j∈{1, 2}; in the j-th level fully connected network: the Dense layer is used as the input to the j-th level fully connected network Come j The BN layer is used to perform batch normalization on the output of the Dense layer; the Dropout layer is used to process the output of the BN layer to prevent overfitting and obtain the output of the j-th fully connected network. Go j ; in, Come 1 for X' eng ; Go 2 for H b .

6. The method for predicting single-phase electric energy meter burnout based on electric meter data according to claim 1, characterized in that: The burn table prediction part includes: 1 splicing layer, 2 Dense layers, and 1 Sigmoid layer; In the burn table prediction section: the splicing layer is used to H a 、 H b The first Dense layer is used to weight and bias the output of the splicing layer; the second Dense layer is used to weight and bias the output of the first Dense layer; the Sigmoid layer is used to process the output of the second Dense layer through the Sigmoid activation function to obtain R .

7. The method for predicting single-phase electric energy meter burnout based on electric meter data according to claim 1, characterized in that: In step three, if R If the target single-phase energy meter is at risk of burning out, an alarm is triggered and / or a maintenance work order is generated; otherwise, the process returns to step 1 or ends.

8. The method for predicting single-phase electric energy meter burnout based on electric meter data according to claim 1, characterized in that: The training methods for the burn prediction model include: Training is performed based on a sample dataset with real labels, and a cross entropy loss function with class weights is constructed. Loss , the model is trained using the Adam optimizer combined with the cosine annealing learning rate scheduler, and an early stopping strategy is applied to prevent overfitting until the training is completed.

9. A system for predicting single-phase electric energy meter burnout based on electric meter data, characterized in that: It uses the method for predicting single-phase electric energy meter burnout based on electric meter data as described in any one of claims 1 to 8; The system for predicting single-phase electric energy meter burnout based on electric meter data includes: A data acquisition module is used to: acquire historical time series data of a target single-phase electric energy meter; Data processing module, which is used to preprocess historical time series data into a three-dimensional array X reshaped , based on X reshaped Perform multi-dimensional feature processing to obtain a multi-dimensional array X' eng ;as well as Burning table prediction module, which is used to: X reshaped 、 X' eng Input the trained burnout prediction model for processing to obtain the prediction result indicating whether the target single-phase electricity meter will burn out. R .

Images