Phase identification method and system for connecting a single-phase electric energy meter to a measuring switch

Abstract

The present invention provides a phase identification method and system for a single-phase electric energy meter connected to a measuring switch, and relates to the technical field of electric power phase identification. The method includes data collection and sorting, voltage fluctuation value calculation, difference calculation and absolute value processing, effective characteristic voltage data screening, and phase determination. The three phases A, B, and C of the intelligent measuring switch are used as references, and the effective value of the user's electric energy meter voltage and the effective value of the measuring switch voltage are collected once a minute through the RS-485 bus. After caching the data of two adjacent moments in the measurement period, the voltage fluctuations Δv of two adjacent moments are calculated and compared, and the phases with obvious characteristics Δv are screened out and the corresponding phase results are accumulated. After time accumulation, the phase with the largest cumulative value is screened out as the phase where the electric energy meter is located. The present invention uses a variety of data processing methods to improve the accuracy and reliability of phase identification.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power phase identification, and in particular to a phase identification method and system for connecting a single-phase electric energy meter to a measuring switch. Background Art

[0002] During the operation of the power system, determining the phase information of the electric energy meter has always been an important and challenging task. Traditional phase determination methods are mainly based on manual on-site surveys and simple electrical measuring instruments, such as using a phase meter to perform point-by-point measurements on the power line. This method not only consumes a lot of manpower, material resources and time costs, but also due to the complexity of the measurement environment and the error factors of manual operation, it is difficult to ensure the high accuracy and consistency of the measurement results. In addition, the data update cycle of the traditional method is long, and it is impossible to reflect the dynamic changes in the operating status of the power system in real time. It is difficult to effectively deal with common scenarios in modern power systems such as rapid fluctuations in power load and access to distributed energy.

[0003] With the rise of smart grid technology, smart power devices such as smart measuring switches and single-phase energy meters are becoming increasingly popular. These devices have powerful data collection and transmission capabilities, making it possible to achieve automated, high-precision phase determination. However, how to fully tap the information behind the massive voltage data collected by these devices and develop an efficient and reliable phase determination algorithm and system has become an urgent problem to be solved in the current power technology field.

[0004] As an edge computing node, the intelligent measuring switch has multiple advanced IoT sensing functions such as distribution network topology identification, meter box management, electricity theft warning, power outage event reporting, line loss analysis and judgment, line impedance calculation, automatic fault diagnosis, etc., which can effectively maintain and monitor the energy meter. The above functions are based on the phase relationship between the energy meter and the measuring switch. However, in the meter acquisition system based on RS-485 bus, the user phase is difficult to identify. The reason is that the amount of voltage and current data collected every day is small, mostly one collection point every 15 minutes or one collection point every day, which cannot support the phase analysis algorithm of the measuring switch. At the same time, the data read from the energy meter through RS-485 is not synchronized with the measuring switch, and the storage capacity of the measuring switch is limited. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a phase identification method and system for connecting a single-phase electric energy meter to a measuring switch, so as to solve the problems of the prior art described in the background technology.

[0006] To solve the above technical problems, the embodiments of the present invention provide the following technical solutions:

[0007] A phase identification method for a single-phase electric energy meter connected to a measuring switch comprises the following steps:

[0008] The intelligent measuring switch reads the voltage effective value of a single-phase electric energy meter every minute and records the three-phase voltage effective value of the intelligent measuring switch at the same time. The collector collects the voltage effective value data of n electric energy meters every minute, and records the voltage effective value of the mth electric energy meter at the tth moment as , where m∈(1, 2…n), the effective value of the voltage of the measuring switch at time t is recorded as , , , calculate the difference ΔV between two adjacent voltage data;

[0009] Will arrive The time is recorded as the time of the i-th adjacent data of the electric energy meter, and the m-th block of electric energy meter is calculated arrive The voltage fluctuation value between adjacent sampling times is obtained , … , m∈(1,2……n), and the other energy meters can be calculated in the same way arrive The voltage fluctuation value between adjacent sampling times; the measurement switch arrive The voltage fluctuation values ​​at each moment are calculated and the difference is obtained to obtain the measured switch voltage fluctuation value;

[0010] Will arrive The voltage fluctuation value data matrix of the n electric meters at the moment is respectively connected to the A, B, and C of the measuring switch. arrive The voltage fluctuation values ​​at adjacent moments before and after the moment are calculated and the absolute values ​​are taken to obtain the absolute value fluctuation data matrix sequence corresponding to the A, B, and C phases in turn. , , ;in, arrive The voltage fluctuation value data matrix of the n electric meters at the moment is:

[0011] ;

[0012] For the absolute value fluctuation data matrix sequence , , The absolute value fluctuation data corresponding to the same electric energy meter at different times are processed by the sliding window averaging algorithm. After the sliding window averaging algorithm is processed, the minimum value of the absolute value fluctuation data of the three-phase voltage at the corresponding time of the corresponding electric energy meter in the data matrix sequence is found by sorting by the sorting algorithm. If the absolute value of the difference between the minimum value and the other two phase values ​​is greater than the dynamic threshold, the data at the corresponding time is the effective characteristic voltage data, and the effective characteristic voltage data is accumulated into the corresponding phase result SumPhaseX, X∈(A,B,C). Otherwise, the data is discarded. After the screening is completed, Time data coverage Time data;

[0013] in:

[0014] = ;

[0015] =

[0016] =

[0017] , , Respectively represent the A, B, and C phases of the measurement switch arrive Voltage fluctuation value at adjacent moments before and after the moment;

[0018] When the cumulative sum of effective characteristic voltage data (SumPhaseA+SumPhaseB+SumPhase)>n, and the ratio of the cumulative data of each phase calculated in sequence to the total exceeds 50%, the corresponding phase is determined as the phase result of the electric energy meter in combination with the confidence assessment algorithm. If the confidence is higher than the preset threshold, the phase is taken as the phase result of the electric energy meter. If the confidence is lower than the preset threshold, the data backtracking and re-analysis process is started.

[0019] Optionally, the voltage fluctuation of the electric energy meter at adjacent moments satisfies ,in represents the effective value of the voltage at the i-th moment of the m-th electric energy meter; , , , , , They respectively represent the effective value of the voltage of the measuring switch at the i-th moment; i is a positive integer greater than 2.

[0020] Optionally, in the screening of effective characteristic voltage data, an artificial intelligence deep learning model is also used to learn and train historical effective characteristic voltage data and corresponding phase results to establish a correlation model between data characteristics and phases.

[0021] Optionally, when determining the phase, a multi-dimensional comprehensive evaluation method is also used, including the accumulation and proportion of effective characteristic voltage data, the real-time power factor of the power system, and the current imbalance.

[0022] Optionally, an adaptive filtering algorithm is introduced before calculating the voltage fluctuation value, and filtering parameters are dynamically adjusted according to the real-time frequency characteristics of the power system.

[0023] Optionally, the artificial intelligence deep learning model is used to learn and train the historical effective characteristic voltage data and the corresponding phase results, and a correlation model between data characteristics and phases is established to provide additional judgment basis for the screening of effective characteristic voltage data, specifically:

[0024] Construct a convolutional neural network (CNN) as an artificial intelligence deep learning model;

[0025] In the data preprocessing stage, the historical effective characteristic voltage data is standardized, and the corresponding phase results are one-hot encoded, and the A, B, and C phases are encoded as [1,0,0], [0,1,0], and [0,0,1] respectively;

[0026] Divide the preprocessed historical data into a training set, a validation set, and a test set according to a predetermined ratio;

[0027] During the training process, the effective characteristic voltage data of the training set is input into the convolutional neural network CNN, the data features are extracted through the convolution layer, the feature dimension is reduced through the pooling layer, and the data is integrated and mapped through the fully connected layer;

[0028] After the model training is completed, when new voltage data is input, the phase to which the new voltage data belongs is predicted based on the learned relationship pattern between historical data features and phases, providing additional judgment basis for the screening of valid characteristic voltage data.

[0029] Optionally, the confidence assessment algorithm is specifically:

[0030] Data preprocessing and feature extraction: obtaining the phase determination case data of the electric energy meter from the power system database, including the cumulative sum of effective characteristic voltage data, the ratio of each phase cumulative data to the total sum, power factor, current imbalance, and the corresponding real phase results and working condition information during determination;

[0031] Data cleaning and organization, removing missing values ​​and outliers from the data, and performing feature extraction and conversion on the original data;

[0032] A random forest classification model is constructed, and the processed historical data is divided into a training set and a test set. The selected model is trained using the training set, and the model performance is optimized by adjusting the hyperparameters of the number of decision trees, the depth of the tree, and the selection criteria for splitting nodes. During the training process, the model learns the relationship between different feature combinations and phase determination results.

[0033] Use the trained model to predict the phase of the electric energy meter data to be evaluated and obtain the predicted phase result;

[0034] The probability P of the predicted phase is calculated based on the voting results of each decision tree in the random forest. It is defined that there are Q decision trees in the random forest, of which T decision trees predict the current phase, then P=Q / T;

[0035] According to the working condition information of the current electric energy meter, the phase determination cases under the corresponding working conditions are searched in the historical data, and the correct proportion of predictions in these corresponding cases is calculated. ;

[0036] Calculate the confidence C, the expression is ,in , is the weight coefficient.

[0037] Optionally, the adaptive filtering algorithm is specifically:

[0038] Continuously collect voltage sample values ​​at set time intervals , and store it in a data buffer, the buffer size is related to the filter order M;

[0039] At each sampling moment , calculate the output of the filter , the calculation formula is

[0040]

[0041] in is the i-th coefficient of the filter at the i-th moment, is the voltage sample value at the corresponding moment in the input data buffer;

[0042] Calculate expected output The actual filter output The error between , using the error to update the filter coefficients , the update formula is: ,in is the input data vector, is the update step size;

[0043] The mean square error at the current moment is calculated based on the calculated error, and the calculated mean square error is compared with the preset minimum mean square error threshold. If it is less than the preset minimum mean square error threshold, the filter converges and the coefficient update is stopped.

[0044] Optionally, in the screening of the effective characteristic voltage data, the adjustment formula of the dynamic threshold is:

[0045]

[0046] in is the new threshold, is the original threshold, is the threshold adjustment calculated based on the load change. It is the threshold adjustment amount calculated according to the line fluctuation amplitude, a is the weight of load change on threshold adjustment, and b is the weight of line fluctuation on threshold adjustment.

[0047] The present invention further proposes a system based on the phase identification method of the above-mentioned single-phase electric energy meter connected to the measuring switch, comprising:

[0048] The data collection and sorting module reads the voltage effective value of a round of hanging single-phase electric energy meters every minute and records the three-phase voltage effective value of the intelligent measurement switch at the same time. The voltage effective value data of n electric energy meters are collected every minute, and the voltage effective value of the mth electric energy meter at the tth moment is recorded as , where m∈(1, 2…n), the effective value of the voltage of the measuring switch at time t is recorded as , , , calculate the difference ΔV between two adjacent voltage data;

[0049] The voltage fluctuation value calculation module will arrive The time is recorded as the time of the i-th adjacent data of the electric energy meter, and the m-th block of electric energy meter is calculated arrive The voltage fluctuation value between adjacent sampling times is obtained , … , m∈(1,2……n), and the other energy meters can be calculated in the same way arrive The voltage fluctuation value between adjacent sampling times; the measurement switch arrive The voltage fluctuation values ​​at each moment are calculated and the difference is obtained to obtain the measured switch voltage fluctuation value;

[0050] The difference calculation and absolute value processing module will arrive The voltage fluctuation value data matrix of the n electric meters at the moment is respectively connected to the A, B, and C of the measuring switch. arrive The voltage fluctuation values ​​at adjacent moments before and after the moment are calculated and the absolute values ​​are taken to obtain the absolute value fluctuation data matrix sequence corresponding to the A, B, and C phases in turn. , , ;in, arrive The voltage fluctuation value data matrix of the n electric meters at the moment is:

[0051] ;

[0052] The effective characteristic voltage data screening module is used to screen the absolute value fluctuation data matrix sequence. , , The absolute value fluctuation data corresponding to the same electric energy meter at different times are processed by the sliding window averaging algorithm. After the sliding window averaging algorithm is processed, the minimum value of the absolute value fluctuation data of the three-phase voltage at the corresponding time of the corresponding electric energy meter in the data matrix sequence is found by sorting by the sorting algorithm. If the absolute value of the difference between the minimum value and the other two phase values ​​is greater than the dynamic threshold, the data at the corresponding time is the effective characteristic voltage data, and the effective characteristic voltage data is accumulated into the corresponding phase result SumPhaseX, X∈(A,B,C). Otherwise, the data is discarded. After the screening is completed, Time data coverage Time data;

[0053] in:

[0054] = ;

[0055] = ;

[0056] = ;

[0057] , , Respectively represent the A, B, and C phases of the measurement switch arrive Voltage fluctuation value at adjacent moments before and after the moment;

[0058] Phase determination module, when the cumulative sum of effective characteristic voltage data (SumPhaseA+SumPhaseB+SumPhase)>n, and the ratio of the cumulative data of each phase calculated in sequence to the total sum exceeds 50%, the corresponding phase is determined as the phase result of the electric energy meter in combination with the confidence assessment algorithm. If the confidence is higher than the preset threshold, the phase is taken as the phase result of the electric energy meter. If the confidence is lower than the preset threshold, the data backtracking and re-analysis process is started.

[0059] The beneficial effects of the above technical solution of the present invention are as follows:

[0060] 1. The present invention uses the three phases A, B, and C of the intelligent measuring switch as a reference, collects the effective value of the user's electric energy meter voltage and the effective value of the measuring switch voltage once a minute through the RS-485 bus, and after caching the data of two adjacent moments in the measurement period, calculates the voltage fluctuations Δv of the two adjacent times for comparison, selects the phases with obvious characteristics Δv and the corresponding results for accumulation, and after a certain period of accumulation, selects the phase with the largest cumulative value as the phase where the electric energy meter is located. This solves the problem that the traditional phase identification method has a long data update cycle and cannot reflect the dynamic changes of the operating status of the power system in real time.

[0061] 2. The optimized voltage data obtained by the present invention after being processed by the adaptive filtering algorithm can reduce the errors introduced by inaccurate data in the subsequent steps of calculating the voltage fluctuation value, performing difference calculation and absolute value processing. Accurate voltage fluctuation value calculation is an important prerequisite for effective characteristic voltage data screening and phase determination. The adaptive filtering algorithm provides a solid guarantee for the accuracy of the entire phase identification process, which helps to improve the reliability of the final phase identification result.

[0062] 3. When screening effective characteristic voltage data, a dynamic threshold adjustment mechanism is introduced to adjust the threshold in real time according to the actual operation of the power system, so that the screening process is more in line with the real characteristics of the voltage fluctuation data. Under different working conditions, the real effective characteristic voltage data can be accurately identified, the interference of invalid data can be reduced, and the accuracy of the accumulation of effective characteristic voltage data can be improved, providing a more reliable data basis for subsequent phase determination.

[0063] 4. In addition, in the screening of effective characteristic voltage data, an artificial intelligence deep learning model is used to learn and train the historical effective characteristic voltage data and the corresponding phase results, establish a correlation model between data features and phases, and assist in the phase judgment based on the current screening data, which can improve the screening accuracy.

[0064] 5. When determining the phase, a confidence assessment algorithm is used, and a multi-dimensional comprehensive assessment method is also used, including the accumulation and proportion of effective characteristic voltage data, the real-time power factor of the power system, and the current imbalance, to improve the accuracy and reliability of phase identification of single-phase electricity meters connected to the measurement switch. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A flow chart of a phase identification method for connecting a single-phase electric energy meter to a measuring switch according to the present invention;

[0066] Figure 2 The schematic diagram of the phase identification system of the single-phase electric energy meter connected to the measuring switch of the present invention is shown in FIG. DETAILED DESCRIPTION

[0067] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0068] like Figure 1 As shown, the present invention proposes a phase identification method for connecting a single-phase electric energy meter to a measuring switch, comprising the following steps:

[0069] S1. Data collection and collation: The intelligent measuring switch reads the voltage effective value of a round of single-phase electric energy meters every minute and records the three-phase voltage effective value of the intelligent measuring switch at the same time. The collector collects the voltage effective value data of n electric energy meters every minute, and records the voltage effective value of the mth electric energy meter at the tth moment as , where m∈(1, 2…n), the effective value of the voltage of the measuring switch at time t is recorded as , , , calculate the difference ΔV between two adjacent voltage data;

[0070] S2. Calculation of voltage fluctuation value: arrive The time is recorded as the time of the i-th adjacent data of the electric energy meter, and the m-th block of electric energy meter is calculated arrive The voltage fluctuation value between adjacent sampling times is obtained , … , m∈(1,2……n), and the other energy meters can be calculated in the same way arrive The voltage fluctuation value between adjacent sampling times; the measurement switch arrive The voltage fluctuation values ​​at each moment are calculated and the difference is obtained to obtain the measured switch voltage fluctuation value;

[0071] S3. Difference calculation and absolute value processing: arrive The voltage fluctuation value data matrix of the n electric meters at the moment is respectively connected to the A, B, and C of the measuring switch. arrive The voltage fluctuation values ​​at adjacent moments before and after the moment are calculated and the absolute values ​​are taken to obtain the absolute value fluctuation data matrix sequence corresponding to the A, B, and C phases in turn. , , ;in, arrive The voltage fluctuation value data matrix of the n electric meters at the moment is:

[0072] ;

[0073] S4. Screening of effective characteristic voltage data: Screening of the absolute value fluctuation data matrix sequence , , The absolute value fluctuation data corresponding to the same electric energy meter at different times are processed by the sliding window averaging algorithm. After the sliding window averaging algorithm is processed, the minimum value of the absolute value fluctuation data of the three-phase voltage at the corresponding time of the corresponding electric energy meter in the data matrix sequence is found by sorting by the sorting algorithm. If the absolute value of the difference between the minimum value and the other two phase values ​​is greater than the dynamic threshold, the data at the corresponding time is the effective characteristic voltage data, and the effective characteristic voltage data is accumulated into the corresponding phase result SumPhaseX, X∈(A,B,C). Otherwise, the data is discarded. After the screening is completed, Time data coverage Time data;

[0074] in:

[0075] =

[0076] =

[0077] =

[0078] , , Respectively represent the A, B, and C phases of the measurement switch arrive Voltage fluctuation value at adjacent moments before and after the moment;

[0079] S5. Phase determination: when the cumulative sum of effective characteristic voltage data (SumPhaseA+SumPhaseB+SumPhase)>n, and the ratio of the cumulative data of each phase calculated in sequence to the total sum exceeds 50%, the corresponding phase is determined as the phase result of the electric energy meter in combination with the confidence assessment algorithm. If the confidence is higher than the preset threshold, the phase is taken as the phase result of the electric energy meter. If the confidence is lower than the preset threshold, the data backtracking and re-analysis process is started.

[0080] Among them, in step S2, the voltage fluctuation of the electric energy meter at adjacent moments satisfies ,in represents the effective value of the voltage at the i-th moment of the m-th electric energy meter; , , , , , They respectively represent the effective value of the voltage of the measuring switch at the i-th moment; i is a positive integer greater than 2.

[0081] In step S2, an adaptive filtering algorithm is introduced before calculating the voltage fluctuation value, and the filtering parameters are dynamically adjusted according to the real-time frequency characteristics of the power system. The adaptive filtering algorithm is specifically:

[0082] Continuously collect voltage sample values ​​at set time intervals , and store it in a data buffer, the buffer size is related to the filter order M;

[0083] At each sampling moment , calculate the output of the filter , the calculation formula is

[0084]

[0085] in is the i-th coefficient of the filter at the i-th moment, is the voltage sample value at the corresponding moment in the input data buffer;

[0086] Calculate expected output The actual filter output The error between , using the error to update the filter coefficients , the update formula is: ,in is the input data vector, is the update step size;

[0087] The mean square error at the current moment is calculated based on the calculated error, and the calculated mean square error is compared with the preset minimum mean square error threshold. If it is less than the preset minimum mean square error threshold, the filter converges and the coefficient update is stopped.

[0088] There are various noises in the power system, such as electromagnetic interference, which can cause fluctuations in the collected voltage data and affect the accuracy of the subsequent voltage fluctuation value calculation. During the operation of the power system, the frequency is not fixed. Factors such as load changes and power grid failures can cause frequency fluctuations, which in turn affect phase recognition. The adaptive filtering algorithm can optimize the filtering parameters in a timely manner according to these real-time changing frequency conditions to ensure that the processing of the voltage sample value is always maintained in the best state, so that the processed data is closer to the real voltage signal characteristics. The optimized voltage data obtained after processing by the adaptive filtering algorithm can reduce the errors introduced by inaccurate data in the subsequent calculation of voltage fluctuation values, difference calculations and absolute value processing. Accurate voltage fluctuation value calculation is an important prerequisite for effective characteristic voltage data screening and phase determination. The adaptive filtering algorithm provides a solid guarantee for the accuracy of the entire phase recognition process and helps to improve the reliability of the final phase recognition results.

[0089] In step S4, during the screening process, the absolute value fluctuation data matrix sequence , , The absolute value fluctuation data corresponding to the same electric energy meter at different times are processed by the sliding window averaging algorithm. In the actual operation of the power system, due to various factors (such as measurement errors, short-term interference, etc.), the collected absolute value fluctuation data may have certain random fluctuations. The sliding window averaging algorithm can effectively smooth these random fluctuations by averaging adjacent data points, making the data more stable and reliable. For example, for an electric energy meter with abnormal voltage fluctuation values ​​caused by short-term interference at several consecutive moments, after sliding window averaging processing, a smooth value that can better reflect its true trend can be obtained, reducing the interference of abnormal fluctuations on subsequent analysis.

[0090] Then, the minimum value of the three-phase voltage absolute value fluctuation data at the corresponding time of the corresponding electric energy meter in the data matrix sequence is found through the sorting algorithm, and the absolute value of the difference between the minimum value and the other two phase values ​​is greater than the dynamic threshold, then the data at the corresponding time is the effective characteristic voltage data. The adjustment formula of the dynamic threshold is:

[0091]

[0092] in is the new threshold, is the original threshold, is the threshold adjustment calculated based on the load change. It is the threshold adjustment amount calculated according to the line fluctuation amplitude, a is the weight of load change on threshold adjustment, and b is the weight of line fluctuation on threshold adjustment.

[0093] Dynamic adjustment principle: For example, through the formula The threshold adjustment amount is calculated according to the load change and added to the original threshold to achieve dynamic adjustment of the threshold. The value of increases accordingly. If a is set reasonably, the new threshold will increase or decrease accordingly to adapt to the voltage fluctuation caused by load changes. This can ensure the accuracy of effective characteristic voltage data screening under different load conditions and avoid misjudgment or missed judgment due to load changes. The principle of adjustment based on line fluctuation amplitude is the same.

[0094] The dynamic threshold adjustment mechanism can adjust the threshold in real time according to the actual operation of the power system, so that the screening process is more in line with the real characteristics of the voltage fluctuation data. Under different working conditions, it can accurately identify the real effective characteristic voltage data, reduce the interference of invalid data, improve the accuracy of the accumulation of effective characteristic voltage data, and provide a more reliable data basis for subsequent phase determination.

[0095] In addition, in the screening of effective characteristic voltage data, an artificial intelligence deep learning model is also used to train the historical effective characteristic voltage data and the corresponding phase results, establish a correlation model between data features and phases, and assist in the phase determination based on the current screening data. Specifically:

[0096] Construct a convolutional neural network (CNN) as an artificial intelligence deep learning model;

[0097] In the data preprocessing stage, the historical effective characteristic voltage data is standardized, and the corresponding phase results are one-hot encoded, and the A, B, and C phases are encoded as [1,0,0], [0,1,0], and [0,0,1] respectively;

[0098] The preprocessed historical data is divided into training set, validation set and test set according to the predetermined ratio. The training set is used for model training to allow the model to learn the characteristic patterns and phase relationships in the historical data; the validation set is used to evaluate the performance of the model during the training process, adjust the hyperparameters of the model, and prevent overfitting; the test set is used to finally evaluate the generalization ability of the model;

[0099] During the training process, the effective characteristic voltage data of the training set is input into the convolutional neural network CNN. The features of the data are extracted through the convolution layer, the feature dimension is reduced through the pooling layer, and the data is integrated and mapped through the fully connected layer. Specifically:

[0100] Convolution layer extracts features: During the training process, the effective feature voltage data of the training set is input into the convolution layer of the CNN model. The convolution layer performs convolution operations by sliding multiple convolution kernels on the data, which can automatically extract local features in the voltage data, such as specific patterns and frequency characteristics of voltage fluctuations. These local features are important for determining the phase to which the electric energy meter belongs. For example, some specific voltage fluctuation patterns may be associated with specific phases.

[0101] Pooling layer dimensionality reduction: After the convolution layer extracts features, the data dimension may be large. The pooling layer reduces the data dimension by performing dimensionality reduction processing, such as maximum pooling or average pooling, to reduce the amount of data computation and storage requirements while retaining key feature information. This helps improve the training efficiency of the model and prevent overfitting, allowing the model to focus more on learning important feature patterns.

[0102] Fully connected layer integration mapping: The fully connected layer integrates the features output by the pooling layer and maps them to the phase prediction result space. It learns how to combine the extracted features to accurately predict the phase of the electric energy meter. By continuously adjusting the weights and biases of the fully connected layer, the model gradually optimizes the ability to predict the phase, making the prediction results closer to the actual phase results.

[0103] Finally, the trained model can be used to assist in the judgment of the current screening data. When new voltage data is input, the model can quickly predict the phase to which it may belong based on the learned relationship between the historical data features and the phase, providing additional judgment basis for the screening of effective characteristic voltage data. This helps to improve the accuracy of screening, reduce the occurrence of misjudgments and missed judgments, and further improve the performance of the entire phase recognition method.

[0104] In the data screening process of phase identification of single-phase electric energy meters connected to measuring switches, the use of artificial intelligence deep learning models to learn historical effective characteristic voltage data and corresponding phase results can improve the accuracy of screening. The data relationship in the power system is complex, and traditional screening methods alone may not be able to fully capture the intrinsic relationship between effective characteristic voltage data and phase. The deep learning model has a strong feature learning ability and can automatically mine the correlation pattern between deep, nonlinear data features and phase from a large amount of historical data. For example, it can find the correspondence between subtle changes in effective characteristic voltage data and phase under different load conditions and different power grid states, thereby providing a more accurate basis for judgment for the current screening data. Secondly, the operating conditions of the power system are complex and changeable, and the expression form and relationship with the phase of the effective characteristic voltage data under different conditions may be different. By learning and training historical data containing various conditions, the deep learning model can learn the characteristic patterns under different conditions, so that when facing the screening data of different conditions, it can more accurately identify the effective characteristic voltage data and improve the accuracy and adaptability of screening.

[0105] In the phase judgment of step S5, the confidence evaluation algorithm is specifically:

[0106] Data preprocessing and feature extraction: obtaining the phase determination case data of the electric energy meter from the power system database, including the cumulative sum of effective characteristic voltage data, the ratio of each phase cumulative data to the total sum, power factor, current imbalance, and the corresponding real phase results and working condition information during determination;

[0107] Data cleaning and organization, removing missing values ​​and outliers from the data, and performing feature extraction and conversion on the original data;

[0108] A random forest classification model is constructed, and the processed historical data is divided into a training set and a test set. The selected model is trained using the training set, and the model performance is optimized by adjusting the hyperparameters of the number of decision trees, the depth of the tree, and the selection criteria for splitting nodes. During the training process, the model learns the relationship between different feature combinations and phase determination results.

[0109] Use the trained model to predict the phase of the electric energy meter data to be evaluated and obtain the predicted phase result;

[0110] The probability P of the predicted phase is calculated based on the voting results of each decision tree in the random forest. It is defined that there are Q decision trees in the random forest, of which T decision trees predict the current phase, then P=Q / T;

[0111] According to the working condition information of the current electric energy meter, the phase determination cases under the corresponding working conditions are searched in the historical data, and the correct proportion of predictions in these corresponding cases is calculated. ;

[0112] Calculate the confidence C, the expression is ,in , is the weight coefficient.

[0113] In addition, in addition to the above confidence assessment algorithm, a multi-dimensional comprehensive assessment method is also used for phase determination, including the accumulation and proportion of effective characteristic voltage data, the real-time power factor of the power system, and the current imbalance, to improve the accuracy and reliability of phase identification of the single-phase energy meter connected to the measurement switch. The specific performance is as follows:

[0114] (1) Effective characteristic voltage data accumulation and proportion

[0115] Reflecting the degree of phase correlation: The cumulative sum and proportion of effective characteristic voltage data is one of the important bases for judging the phase to which the electric energy meter belongs. The voltage fluctuation characteristics of electric energy meters on different phases will be different during operation. By calculating the cumulative sum of effective characteristic voltage data in each phase and the proportional relationship between them, the degree of correlation between the electric energy meter and each phase can be intuitively reflected. For example, if the cumulative sum of effective characteristic voltage data of a certain electric energy meter on phase A accounts for a high proportion of the total cumulative sum and exceeds a certain threshold (such as 50%), it is preliminarily indicated that the electric energy meter is more likely to belong to phase A.

[0116] Auxiliary phase preliminary judgment: In the initial stage of phase judgment, the accumulation and ratio of effective characteristic voltage data can provide a preliminary judgment direction. When the ratio is obviously prominent on a certain phase, it can be used as an important reference for phase judgment, narrowing the possible phase range and laying the foundation for subsequent more accurate judgment. At the same time, combined with the evaluation of other dimensions, this preliminary judgment result can be further verified and refined.

[0117] (2) Real-time power factor of power system

[0118] Reflecting the operating status of the system: The real-time power factor of the power system is an important indicator reflecting the operating status of the system. The power factor is closely related to the ratio of active power to reactive power in the power system, and the load characteristics on different phases will affect the size of the power factor. For example, a phase with more inductive loads may result in a lower power factor, and a phase with more capacitive loads may result in a higher power factor. Therefore, the real-time power factor can provide additional information for phase determination and help distinguish the load characteristics of different phases.

[0119] Supplementary basis for phase judgment: In the process of phase judgment, taking the real-time power factor into consideration can judge the phase of the electric energy meter from the perspective of the overall operation of the power system. When the accumulation and proportion of effective characteristic voltage data have a certain tendency on a certain phase, combined with the real-time power factor of the phase, the rationality of the phase judgment can be further confirmed. If the relationship between the two conforms to the basic principles and laws of the power system, the confidence in the phase judgment result can be enhanced; otherwise, it may be necessary to review the phase judgment result to avoid misjudgment due to a single factor.

[0120] (3) Current imbalance

[0121] Reflecting the three-phase imbalance: Current imbalance is an indicator to measure the degree of imbalance of the three-phase current in the power system. In a three-phase power system, ideally the three-phase current should be balanced, but in actual operation, current imbalance may occur due to uneven load distribution. The current imbalance on different phases will be different, and this difference can provide clues for phase determination. For example, if the current imbalance of one phase is significantly higher than that of the other two phases, and there is a certain correlation with the accumulation and proportion of the effective characteristic voltage data of a certain electric energy meter, it can assist in determining whether the electric energy meter belongs to this phase.

[0122] Comprehensive judgment improves accuracy: Combining the current imbalance with the accumulation and proportion of effective characteristic voltage data, real-time power factor and other factors for comprehensive evaluation can provide a more comprehensive understanding of the operating status of the power system and the relationship between the electric energy meter and each phase. Through multi-dimensional information interaction verification, the accuracy of phase determination can be improved, and the erroneous results caused by inaccurate judgment of a single factor can be reduced. In a complex power system environment, this multi-dimensional comprehensive evaluation method can better adapt to various operating conditions and provide strong support for accurate phase identification.

[0123] like Figure 2 As shown, the present invention also proposes a system based on the phase identification method of the above-mentioned single-phase electric energy meter connected to the measuring switch, comprising:

[0124] The data collection and sorting module 101 reads the voltage effective value of a round of hanging single-phase electric energy meters every minute and records the three-phase voltage effective value of the intelligent measuring switch at the same time. The voltage effective value data of the electric energy meters of n number are collected every minute, and the voltage effective value of the electric energy meter of the mth block at the tth moment is recorded as , where m∈(1, 2…n), the effective value of the voltage of the measuring switch at time t is recorded as , , , calculate the difference ΔV between two adjacent voltage data;

[0125] The voltage fluctuation value calculation module 102 calculates arrive The time is recorded as the time of the i-th adjacent data of the electric energy meter, and the m-th block of electric energy meter is calculated arrive The voltage fluctuation value between adjacent sampling times is obtained , … , m∈(1,2……n), and the other energy meters can be calculated in the same way arrive The voltage fluctuation value between adjacent sampling times; the measurement switch arrive The voltage fluctuation values ​​at each moment are calculated and the difference is obtained to obtain the measured switch voltage fluctuation value;

[0126] The difference calculation and absolute value processing module 103 converts arrive The voltage fluctuation value data matrix of the n electric meters at the moment is respectively connected to the A, B, and C of the measuring switch. arrive The voltage fluctuation values ​​at adjacent moments before and after the moment are calculated and the absolute values ​​are taken to obtain the absolute value fluctuation data matrix sequence corresponding to the A, B, and C phases in turn. , , ;in, arrive The voltage fluctuation value data matrix of the n electric meters at the moment is:

[0127] ;

[0128] The effective characteristic voltage data screening module 104 selects the absolute value fluctuation data matrix sequence , , The absolute value fluctuation data corresponding to the same electric energy meter at different times are processed by the sliding window averaging algorithm. After the sliding window averaging algorithm is processed, the minimum value of the absolute value fluctuation data of the three-phase voltage at the corresponding time of the corresponding electric energy meter in the data matrix sequence is found by sorting by the sorting algorithm. If the absolute value of the difference between the minimum value and the other two phase values ​​is greater than the dynamic threshold, the data at the corresponding time is the effective characteristic voltage data, and the effective characteristic voltage data is accumulated into the corresponding phase result SumPhaseX, X∈(A,B,C). Otherwise, the data is discarded. After the screening is completed, Time data coverage Time data;

[0129] in:

[0130] =

[0131] =

[0132] =

[0133] , , Respectively represent the A, B, and C phases of the measurement switch arrive Voltage fluctuation value at adjacent moments before and after the moment;

[0134] Phase determination module 105, when the cumulative sum of effective characteristic voltage data (SumPhaseA+SumPhaseB+SumPhase)>n, and the ratio of the cumulative data of each phase calculated in sequence to the total sum exceeds 50%, the corresponding phase is determined as the phase result of the electric energy meter in combination with the confidence evaluation algorithm. If the confidence is higher than the preset threshold, the phase is taken as the phase result of the electric energy meter. If the confidence is lower than the preset threshold, the data backtracking and re-analysis process is started.

[0135] The specific methods for implementing each module of the above system will not be repeated here.

[0136] In summary, the present invention uses the three phases A, B, and C of the intelligent measuring switch as a reference, collects the effective value of the user's electric energy meter voltage and the effective value of the measuring switch voltage once a minute through the RS-485 bus, and after caching the data of two adjacent moments in the measurement period, respectively calculates the voltage fluctuation Δv of the two adjacent times for comparison, selects the phase results with obvious characteristics Δv and corresponding to the accumulation, and after a certain period of accumulation, selects the phase with the largest cumulative value as the phase where the electric energy meter is located. This solves the problem that the traditional phase identification method has a long data update cycle and cannot reflect the dynamic changes of the power system operation status in real time.

[0137] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A phase identification method for a single-phase electric energy meter connected to a measuring switch, characterized in that: The following steps are involved: The intelligent measuring switch reads the voltage effective value of a single-phase electric energy meter every minute and records the three-phase voltage effective value of the intelligent measuring switch at the same time. The collector collects the voltage effective value data of n electric energy meters every minute, and records the voltage effective value of the mth electric energy meter at the tth moment as , where m∈(1, 2…n), the effective value of the voltage of the measuring switch at time t is recorded as , , , calculate the difference ΔV between two adjacent voltage data; Will arrive The time is recorded as the time of the i-th adjacent data of the electric energy meter, and the m-th block of electric energy meter is calculated arrive The voltage fluctuation value between adjacent sampling times is obtained , … , m∈(1,2……n), and the other energy meters can be calculated in the same way arrive The voltage fluctuation value between adjacent sampling times; the measurement switch arrive The voltage fluctuation values ​​at each moment are calculated and the difference is obtained to obtain the measured switch voltage fluctuation value; Will arrive The voltage fluctuation value data matrix of n electric energy meters at the moment is respectively related to the A, B, and C of the measuring switch. arrive The voltage fluctuation values ​​at adjacent moments before and after the moment are calculated and the absolute values ​​are taken to obtain the absolute value fluctuation data matrix sequence corresponding to the A, B, and C phases in turn. , , ;in, arrive The voltage fluctuation value data matrix of the n electric energy meters at the moment is: ； For the absolute value fluctuation data matrix sequence , , The absolute value fluctuation data corresponding to the same electric energy meter at different times are processed by the sliding window averaging algorithm. After the sliding window averaging algorithm is processed, the minimum value of the absolute value fluctuation data of the three-phase voltage at the corresponding time of the corresponding electric energy meter in the data matrix sequence is found by sorting by the sorting algorithm. If the absolute value of the difference between the minimum value and the other two phase values ​​is greater than the dynamic threshold, the data at the corresponding time is the effective characteristic voltage data, and the effective characteristic voltage data is accumulated into the corresponding phase result SumPhaseX, X∈(A,B,C). Otherwise, the data is discarded. After the screening is completed, Time data coverage Time data; in: = ； = ； = ； , , Respectively represent the A, B, and C phases of the measurement switch arrive Voltage fluctuation value at adjacent moments before and after the moment; When the cumulative sum of effective characteristic voltage data (SumPhaseA+SumPhaseB+SumPhase)>n, and the ratio of the cumulative data of each phase calculated in sequence to the total exceeds 50%, the corresponding phase is determined as the phase result of the electric energy meter in combination with the confidence assessment algorithm. If the confidence is higher than the preset threshold, the phase is taken as the phase result of the electric energy meter. If the confidence is lower than the preset threshold, the data backtracking and re-analysis process is started.

2. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 1, characterized in that: The voltage fluctuation of the electric energy meter at adjacent moments satisfies ,in represents the effective value of the voltage at the i-th moment of the m-th electric energy meter; , , , , , They respectively represent the effective value of the voltage of the measuring switch at the i-th moment; i is a positive integer greater than 2.

3. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 1, characterized in that: In the screening of effective characteristic voltage data, an artificial intelligence deep learning model is also used to learn and train historical effective characteristic voltage data and corresponding phase results to establish a correlation model between data characteristics and phase.

4. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 1, characterized in that: When determining the phase, a multi-dimensional comprehensive evaluation method is also used, including the accumulation and proportion of effective characteristic voltage data, the real-time power factor of the power system, and the current imbalance.

5. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 1, characterized in that: An adaptive filtering algorithm is introduced before calculating the voltage fluctuation value, and the filtering parameters are dynamically adjusted according to the real-time frequency characteristics of the power system.

6. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 3, characterized in that: The artificial intelligence deep learning model is used to learn and train the historical effective characteristic voltage data and the corresponding phase results, and a correlation model between data characteristics and phases is established to provide additional judgment basis for the screening of effective characteristic voltage data, specifically: Construct a convolutional neural network (CNN) as an artificial intelligence deep learning model; In the data preprocessing stage, the historical effective characteristic voltage data is standardized, and the corresponding phase results are one-hot encoded, and the A, B, and C phases are encoded as [1,0,0], [0,1,0], and [0,0,1] respectively; Divide the preprocessed historical data into a training set, a validation set, and a test set according to a predetermined ratio; During the training process, the effective characteristic voltage data of the training set is input into the convolutional neural network CNN, the data features are extracted through the convolution layer, the feature dimension is reduced through the pooling layer, and the data is integrated and mapped through the fully connected layer; After the model training is completed, when new voltage data is input, the phase to which the new voltage data belongs is predicted based on the learned relationship pattern between historical data features and phases, providing additional judgment basis for the screening of valid characteristic voltage data.

7. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 1, characterized in that: The confidence evaluation algorithm is specifically: Data preprocessing and feature extraction: obtaining the phase determination case data of the electric energy meter from the power system database, including the cumulative sum of effective characteristic voltage data, the ratio of each phase cumulative data to the total sum, power factor, current imbalance, and the corresponding real phase results and working condition information during determination; Data cleaning and organization, removing missing values ​​and outliers from the data, and performing feature extraction and conversion on the original data; A random forest classification model is constructed, and the processed historical data is divided into a training set and a test set. The selected model is trained using the training set, and the model performance is optimized by adjusting the hyperparameters of the number of decision trees, the depth of the tree, and the selection criteria for splitting nodes. During the training process, the model learns the relationship between different feature combinations and phase determination results. Use the trained model to predict the phase of the electric energy meter data to be evaluated and obtain the predicted phase result; The probability P of the predicted phase is calculated based on the voting results of each decision tree in the random forest. It is defined that there are Q decision trees in the random forest, of which T decision trees predict the current phase, then P=Q / T; According to the working condition information of the current electric energy meter, the phase determination cases under the corresponding working conditions are searched in the historical data, and the proportion of correct predictions in these corresponding cases is calculated. ; Calculate the confidence C, the expression is ,in , is the weight coefficient.

8. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 5, characterized in that: The adaptive filtering algorithm is specifically: Continuously collect voltage sample values ​​at set time intervals , and store it in a data buffer, the buffer size is related to the filter order M; At each sampling moment , calculate the output of the filter , the calculation formula is ； in is the i-th coefficient of the filter at the i-th moment, is the voltage sample value at the corresponding moment in the input data buffer; Calculate expected output The actual filter output The error between , using the error to update the filter coefficients , the update formula is: ,in is the input data vector, is the update step size; The mean square error at the current moment is calculated based on the calculated error, and the calculated mean square error is compared with the preset minimum mean square error threshold. If it is less than the preset minimum mean square error threshold, the filter converges and the coefficient update is stopped.

9. The phase identification method for connecting a single-phase electric energy meter to a measuring switch according to claim 1, characterized in that: In the screening of the effective characteristic voltage data, the adjustment formula of the dynamic threshold is: ； in is the new threshold, is the original threshold, is the threshold adjustment calculated based on the load change. It is the threshold adjustment amount calculated according to the line fluctuation amplitude, a is the weight of load change on threshold adjustment, and b is the weight of line fluctuation on threshold adjustment.

10. A system based on the phase identification method of a single-phase electric energy meter connected to a measuring switch according to any one of claims 1 to 9, characterized in that: include: The data collection and sorting module reads the voltage effective value of a round of hanging single-phase electric energy meters every minute and records the three-phase voltage effective value of the intelligent measurement switch at the same time. The voltage effective value data of n electric energy meters are collected every minute, and the voltage effective value of the mth electric energy meter at the tth moment is recorded as , where m∈(1, 2…n), the effective value of the voltage of the measuring switch at time t is recorded as , , , calculate the difference ΔV between two adjacent voltage data; The voltage fluctuation value calculation module will arrive The time is recorded as the time of the i-th adjacent data of the electric energy meter, and the m-th block of electric energy meter is calculated arrive The voltage fluctuation value between adjacent sampling times is obtained , … , m∈(1,2……n), and the other energy meters can be calculated in the same way arrive The voltage fluctuation value between adjacent sampling times; the measurement switch arrive The voltage fluctuation values ​​at each moment are calculated and the difference is obtained to obtain the measured switch voltage fluctuation value; The difference calculation and absolute value processing module will arrive The voltage fluctuation value data matrix of n electric energy meters at the moment is respectively related to the A, B, and C of the measuring switch. arrive The voltage fluctuation values ​​at adjacent moments before and after the moment are calculated and the absolute values ​​are taken to obtain the absolute value fluctuation data matrix sequence corresponding to the A, B, and C phases in turn. , , ;in, arrive The voltage fluctuation value data matrix of the n electric energy meters at the moment is: ； The effective characteristic voltage data screening module is used to screen the absolute value fluctuation data matrix sequence. , , The absolute value fluctuation data corresponding to the same electric energy meter at different times are processed by the sliding window averaging algorithm. After the sliding window averaging algorithm is processed, the minimum value of the absolute value fluctuation data of the three-phase voltage at the corresponding time of the corresponding electric energy meter in the data matrix sequence is found by sorting by the sorting algorithm. If the absolute value of the difference between the minimum value and the other two phase values ​​is greater than the dynamic threshold, the data at the corresponding time is the effective characteristic voltage data, and the effective characteristic voltage data is accumulated into the corresponding phase result SumPhaseX, X∈(A,B,C). Otherwise, the data is discarded. After the screening is completed, Time data coverage Time data; in: = ； = ； = ； , , Respectively represent the A, B, and C phases of the measurement switch arrive Voltage fluctuation value at adjacent moments before and after the moment; Phase determination module, when the cumulative sum of effective characteristic voltage data (SumPhaseA+SumPhaseB+SumPhase)>n, and the ratio of the cumulative data of each phase calculated in sequence to the total sum exceeds 50%, the corresponding phase is determined as the phase result of the electric energy meter in combination with the confidence assessment algorithm. If the confidence is higher than the preset threshold, the phase is taken as the phase result of the electric energy meter. If the confidence is lower than the preset threshold, the data backtracking and re-analysis process is started.

Images