A mode recognition method and system for three-phase imbalance of a commutation switch station area

Abstract

The application relates to the field of three-phase imbalance recognition, in particular to a commutation switch transformer area three-phase imbalance mode recognition method and system, which comprises the following steps: acquiring the features and corresponding labels of the commutation switch transformer area in a preset time window, and constructing a data set under different preset time windows; calculating the stability of the features to the labels and the redundancy of the features under different preset time windows according to the data set under the different preset time windows, calculating the importance of the features to the labels according to the stability of the features to the labels and the redundancy of the features, and screening out the best window corresponding to different labels according to the sum of the importance of the features to the labels under different preset time windows; and constructing a label recognition model according to the best window of the labels, which is used for real-time three-phase imbalance recognition after model training. The application has the effect of improving the three-phase imbalance recognition accuracy.

Description

Technical Field

[0001] This application relates to the field of three-phase imbalance identification, and in particular to a method and system for identifying three-phase imbalance patterns in a commutation switch area. Background Technology

[0002] In low-voltage distribution networks, electricity is transmitted to end users through a three-phase four-wire system. Ideally, the loads of phases A, B, and C should be evenly distributed. However, due to the spatial randomness and temporal fluctuations of single-phase loads (residential electricity, commercial electricity, etc.), a severe imbalance of three-phase currents is common in actual operation.

[0003] To address the aforementioned issues, automatic phase-switching switches have emerged. As an intelligent terminal, they can be installed at the user access point and, through power electronics or mechanical switches, remotely or automatically switch the user's load from the heavily loaded phase to the lightly loaded phase without interrupting power supply. However, phase-switching switches alone are insufficient. Accurately determining when to switch phases, which phase to switch to, and to which phase is crucial to the effectiveness of the system's management. Current phase-switching control systems mostly employ triggering mechanisms based on simple thresholds (such as a phase current exceeding a set value). This piecemeal approach may lead to frequent malfunctions due to temporary fluctuations, shortening equipment lifespan, and may also fail to recognize complex patterns such as alternation and fluctuations, resulting in poor or even worsened management effects. Therefore, how to accurately identify three-phase imbalances and improve the accuracy of three-phase imbalance identification is an urgent problem to be solved. Summary of the Invention

[0004] To address the problem of low accuracy in three-phase imbalance identification, this application provides a method and system for identifying three-phase imbalance patterns in a commutation switch area.

[0005] Firstly, this application provides a method for identifying three-phase imbalance patterns in a commutation switch area, employing the following technical solution: A method for identifying three-phase imbalance patterns in a commutation switch area includes the following steps: constructing an identification model based on historical data, acquiring real-time data of the commutation switch area, and obtaining the identification result of the three-phase imbalance pattern based on the identification model; The identification model is constructed as follows: the features and corresponding labels of the switching station area within a preset time window in historical data are obtained, and datasets under different preset time windows are constructed; the stability and feature redundancy of the features to the labels under different preset time windows are calculated based on the datasets under different preset time windows. The method for calculating feature redundancy is as follows: calculate the linear redundancy and nonlinear redundancy between the target feature and any other feature; calculate the noise of the feature; sum the linear redundancy of the target feature with other features based on the noise of the feature to obtain the linear redundancy of the target feature; sum the nonlinear redundancy of the target feature with other features based on the noise of the feature to obtain the nonlinear redundancy of the feature; calculate the probability that the distribution of the target feature conforms to a normal distribution; and sum the linear redundancy and nonlinear redundancy based on the probability to obtain the feature redundancy. The importance of features to labels is calculated based on the stability and redundancy of features. The optimal window for each label is selected based on the total importance of features to labels under different preset time windows. A label recognition model is then constructed based on the optimal window for each label.

[0006] Optionally, the characteristics of the commutation switch area include the mean, variance, rate of change, total harmonic distortion rate, harmonic content rate, and Hurst exponent of the three-phase imbalance.

[0007] Optionally, the stability calculation method is as follows: divide the feature and feature change into intervals, arbitrarily pair the feature intervals and feature change intervals to obtain multiple combinations, calculate the probability of the label appearing in different combinations, calculate the entropy value based on the probability, and normalize the opposite of the entropy value to obtain the stability.

[0008] Optionally, the recognition model is a BP neural network model.

[0009] Optionally, the linear redundancy can be calculated by: calculating the Pearson correlation coefficient or variance inflation factor between two features, taking the absolute value of the Pearson correlation coefficient to obtain the linear redundancy, or mapping the difference between the variance inflation factor and 1 through the tanh function to obtain the linear redundancy.

[0010] Optionally, the nonlinear redundancy can be calculated by: calculating the mutual information between two features, mapping the mutual information using the tanh function to obtain the nonlinear redundancy, or calculating the distance correlation coefficient between two features to obtain the nonlinear redundancy.

[0011] Optionally, the probability that the distribution of the target feature conforms to a normal distribution is obtained through a KS test.

[0012] Secondly, this application provides a three-phase unbalanced pattern recognition system for commutation switch areas, employing the following technical solution: A three-phase imbalance pattern recognition system for a commutation switch area includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the three-phase imbalance pattern recognition method for a commutation switch area as described above.

[0013] The beneficial effect is that the above-mentioned method for recognizing three-phase imbalance patterns in a phase-switching switch area generates a computer program and stores it in the memory so that it can be loaded and executed by the processor. Thus, a system can be built based on the memory and the processor, which is convenient to use.

[0014] This application has the following technical advantages: 1. Traditional identification methods often use fixed time windows for analysis, but different types of imbalance states (such as persistent overload and fluctuating imbalance) have different durations and change characteristics. This method calculates the importance of features to the label under different preset time windows, dynamically selecting the "optimal observation window" for each imbalance pattern. This allows the model to capture pattern features at the most appropriate scale, fundamentally improving the accuracy of identification.

[0015] 2. In addition to considering the stability between features and labels (reflecting the ability of features to distinguish patterns), this approach also introduces redundancy evaluation (including linear and nonlinear redundancy) and innovatively weights the impact of feature noise and data distribution. This multi-dimensional feature selection mechanism can select the most representative, informative, and highly independent feature combinations, effectively avoiding model performance degradation caused by feature redundancy or noise interference, thereby improving the model's robustness and generalization ability. Attached Figure Description

[0016] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily understood by reading the following detailed description with reference to the accompanying drawings. Several embodiments of this application are illustrated in the drawings by way of example and not limitation, and the same or corresponding reference numerals denote the same or corresponding parts.

[0017] Figure 1 This is a flowchart illustrating steps S1-S3 of a three-phase imbalance pattern recognition method for a phase-switching switch area according to this application.

[0018] Figure 2 This is a structural block diagram of a three-phase unbalanced pattern recognition system for a phase switching area according to this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] It should be understood that when the terms "first," "second," etc., are used in the claims, description, and drawings of this application, they are only used to distinguish different objects and not to describe a specific order. The terms "comprising" and "including" used in the description and claims of this application indicate the presence of the described features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof.

[0021] This application discloses a method for recognizing three-phase imbalance patterns in a commutation switch area, referring to... Figure 1 This includes the following steps: S1: Obtain the features and corresponding labels of the commutation switch area within a preset time window in historical data, and construct datasets under different preset time windows.

[0022] In one embodiment, sensors are used to collect three-phase current, three-phase voltage, and neutral current in the commutation switch area. A preset time window is used to calculate the unbalance of the three-phase current at each moment within the preset time window. Then, the mean, variance, rate of change, total harmonic distortion (THD), harmonic content, and Hurst exponent (the Hurst exponent is an important indicator for measuring the long-term memory and trend persistence of a time series) of the unbalance within the window are calculated. The rate of change of the unbalance within the window is calculated as the mean of the rate of change of the unbalance at each moment within the window. The harmonic content is calculated as follows: Fourier transform is performed on the collected current to obtain the amplitude of each harmonic; the effective value is obtained based on the amplitude; and the harmonic content is... The ratio of the effective value to the fundamental effective value is the harmonic content rate, and the total harmonic distortion rate is the ratio of the root mean square of the effective values ​​of all harmonic components to the fundamental effective value. Similarly, the mean, variance, and rate of change of the voltage imbalance within a preset time window of the three-phase voltage are obtained. Based on the characteristics calculated by those skilled in the art according to the collected data and combined with the actual conditions of the commutation switch area, each preset time window is labeled with a label, which is single-phase continuous heavy load / alternating imbalance / fluctuating imbalance / harmonic-dominated imbalance. Different preset time windows are obtained to obtain the characteristics and corresponding labels of the commutation switch area under different preset time windows. All the characteristics and labels under the same preset time window constitute the dataset under that preset time window.

[0023] S2: Calculate the stability of feature pairs and feature redundancy under different preset time windows based on the dataset under different preset time windows.

[0024] In one embodiment, the method for calculating feature redundancy is as follows: calculate the linear redundancy and nonlinear redundancy between the target feature and any feature; calculate the noise of the feature; perform a weighted summation of the linear redundancy between features based on the noise of the feature to obtain the linear redundancy of the feature; perform a weighted summation of the nonlinear redundancy between features based on the noise of the feature to obtain the nonlinear redundancy of the feature; calculate the probability that the distribution of the target feature conforms to a normal distribution; and perform a weighted summation of the linear redundancy and nonlinear redundancy based on the probability to obtain the feature redundancy.

[0025] In one embodiment, due to the different conditions under which different imbalance states occur, some imbalance states last for a long time, requiring a longer window to achieve good recognition results; while others last for a short time, achieving good recognition results within a shorter window. For example, fluctuating imbalances are caused by a large number of randomly starting and stopping loads and have a short duration; while continuous heavy loads on a single phase are caused by fixed, long-term uneven load distribution and have a long duration. Therefore, the importance of features to labels under different preset time windows is calculated based on the dataset under different windows. Specifically, the stability between individual features and labels is first calculated. When calculating stability, the stability between features and feature changes and labels is considered. Features and feature changes are divided into intervals. For example, the mean current imbalance degree is divided into 3 intervals, and the change in the mean current imbalance degree is divided into 3 intervals. Then, the intervals of the mean current imbalance degree and the intervals of the change in the mean current imbalance degree are combined to obtain 9 combinations. The degree of label disorder under the combination is calculated to obtain the stability. The specific calculation formula is as follows: in, Indicates the first j The first feature and the first i Stability between tags This represents the normalization function, the purpose of which is to normalize the stability to a range between 0 and 1. Indicates the first k In the combination, the first i The number of times each label appears. Indicates the first k In the combination, the first i The total number of times each tag appears. Represents a logarithmic function.

[0026] The closer the frequency of the labels in each combination, the less stable the relationship between the feature and the labels; the greater the difference in the frequency of the labels in each combination, the more concentrated the labels are under a certain feature combination. The more stable the relationship between the feature and the labels, the more important the feature is.

[0027] The importance of a feature depends not only on the stability between the feature and the label, but also on the redundancy between features. Higher feature redundancy means that a feature is more easily replaced by other features. Furthermore, high feature redundancy increases the risk of overfitting in the trained model and weakens its generalization ability. Therefore, it is necessary to comprehensively evaluate the importance of a feature based on both its stability and redundancy. The calculation process for feature redundancy is as follows: Since the relationships between features can be linear or non-linear, the redundancy between features needs to consider both linear and non-linear relationships. For linear relationships, the Pearson correlation coefficient or variance inflation factor is used to reflect the linear redundancy. Normalizing the Pearson correlation coefficient or variance inflation factor yields the linear redundancy. The absolute value of the Pearson correlation coefficient is then removed for normalization. The difference between the variance inflation factor and 1 is normalized using the tanh function. A larger normalized value indicates greater linear redundancy and lower feature importance. The non-linear redundancy between two features is calculated as follows: The mutual information or distance correlation coefficient between features is calculated, and normalized to obtain the non-linear redundancy. The mutual information is normalized using the tanh function. The distance correlation coefficient ranges from 0 to 1, therefore, further normalization is not required. The mean of the linear redundancy between each feature and all other features yields the linear redundancy of that feature. The mean of the non-linear redundancy between each feature and all other features yields the non-linear redundancy of that feature.

[0028] In another embodiment, the noise of the feature itself is considered when calculating the linear and nonlinear redundancy. The greater the noise, the lower the reliability of the calculated linear or nonlinear redundancy. The weighted linear or nonlinear redundancy is obtained by weighting according to the noise level of the feature itself. The noise level is calculated as follows: in, Indicates the noise level of the feature. Indicates The purpose of using an exponential function with base 0 is to normalize. The signal-to-noise ratio (SNR) represents the feature. The higher the SNR, the more useful information there is and the less noise reflects the feature; the lower the SNR, the less useful information there is and the more noise reflects the feature.

[0029] The linear or nonlinear redundancy is weighted according to the noise level of the feature. The specific calculation formula is as follows: in, Indicates the first j The linear redundancy of each feature Indicates the first j The noise level of each feature Indicates the first s The noise level of each feature Indicates the first j The first feature and the first s Linear redundancy among features The function is used to normalize the sum of the noise levels of two features. The greater the noise level, the lower the reliability of the linear redundancy between the two features, and the smaller the weight when calculating the linear redundancy of the target feature.

[0030] Similarly, the nonlinear redundancy of the features is also weighted using the same method, thus obtaining the linear and nonlinear redundancy of each feature.

[0031] The linear and nonlinear redundancy of a feature together reflect its overall redundancy. When calculating linear redundancy, the data distribution of the feature affects its reliability because the Pearson correlation coefficient and variance inflation factor are based on the assumption of a normal data distribution. Therefore, when obtaining feature redundancy from linear and nonlinear redundancy, it is necessary to assess the reliability of the linear redundancy. Specifically, the probability that the feature distribution conforms to a normal distribution is obtained through the KS test. The higher the probability of conforming to a normal distribution, the greater the weight of the linear redundancy. The specific formula for calculating feature redundancy is as follows: in, Indicates the first j Redundancy of each feature Indicates the first j The weights of linear redundancy of each feature. Indicates the first j The linear redundancy of each feature Indicates the first j The nonlinear redundancy of each feature. The sum of the weights of the linear redundancy and nonlinear redundancy of the feature is 1. If the weight of the linear redundancy is higher, the weight of the nonlinear redundancy is lower.

[0032] The weights for linear redundancy are calculated as follows: in, Indicates the first j The weights of linear redundancy of each feature. For normalization function, It is a constant. Indicates the first j The probability that the distribution of a feature conforms to a normal distribution, for example: The value is 10.

[0033] S3: Calculate the importance of features to labels based on the stability and redundancy of features, and select the best window for different labels based on the total importance of features to labels under different windows; construct the label recognition model based on the best window for the label.

[0034] In one embodiment, the importance of different features to different labels is calculated based on feature stability and feature redundancy. The specific calculation formula is as follows: in, Indicates the first j The feature is related to the first i The importance of each tag Indicates the first j The first feature and the first i Stability between tags Indicates the first j The redundancy of a feature. The greater the stability between a feature and its label, the more important the feature; the less redundancy a feature has, the more important it is.

[0035] Under different preset time windows, the importance of each feature to each label is calculated, the importance of all features to the target label is summed, and the preset time window with the largest summation value is selected as the best window for the target label.

[0036] In one embodiment, a backpropagation (BP) neural network is constructed to identify target labels. Step S2 obtains the optimal window for the target label and the importance of each feature within that window. The feature values, multiplied by their importance, are input into the neural network, which outputs the predicted value of the target label. The loss function of the neural network is cross-entropy loss. Gradient descent is used to update the model parameters. Training stops when the model's loss is less than a set threshold or the number of iterations reaches a preset maximum. The optimal model is obtained based on the model's accuracy as an evaluation metric. The model training is then complete. The same method is used to obtain the corresponding optimal model for different labels.

[0037] The system collects real-time data on the three-phase current, three-phase voltage, and neutral current of the commutation switch area. It calculates the characteristics of the commutation switch area under different optimal windows, multiplies the characteristics by the importance of the characteristics under the optimal window, inputs the resulting characteristic value into the optimal model under the optimal window, and outputs the prediction result of the label corresponding to the optimal model. The system performs the above operation on the characteristics under different optimal windows to obtain the prediction result of the label corresponding to the different optimal models. When a certain three-phase imbalance is detected, an alarm is issued to remind relevant personnel to handle the situation.

[0038] This application also discloses a three-phase imbalance pattern recognition system for a commutation switch area, referring to... Figure 2 It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a three-phase imbalance pattern recognition method for a commutation switch area according to this application.

[0039] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0040] In this application, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device.

[0041] While this specification has shown and described numerous embodiments of the present application, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will arise for those skilled in the art without departing from the spirit and intent of the present application. It should be understood that various alternatives to the embodiments of the present application described herein may be employed in the practice of this application.

[0042] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for recognizing three-phase imbalance patterns in a commutation switch area, characterized in that, The process includes the following steps: building an identification model based on historical data, obtaining real-time data of the commutation switch area, and obtaining the identification results of the three-phase imbalance mode based on the identification model; The identification model is constructed as follows: the features and corresponding labels of the switching station area within a preset time window in historical data are obtained, and datasets under different preset time windows are constructed; the stability and feature redundancy of the features to the labels under different preset time windows are calculated based on the datasets under different preset time windows. The method for calculating feature redundancy is as follows: calculate the linear redundancy and nonlinear redundancy between the target feature and any other feature; calculate the noise of the feature; sum the linear redundancy of the target feature with other features based on the noise of the feature to obtain the linear redundancy of the target feature; sum the nonlinear redundancy of the target feature with other features based on the noise of the feature to obtain the nonlinear redundancy of the feature; calculate the probability that the distribution of the target feature conforms to a normal distribution; and sum the linear redundancy and nonlinear redundancy based on the probability to obtain the feature redundancy. The importance of features to labels is calculated based on the stability and redundancy of features. The optimal window for each label is selected based on the total importance of features to labels under different preset time windows. A label recognition model is then constructed based on the optimal window for each label.

2. The method for recognizing three-phase imbalance patterns in a commutation switch area according to claim 1, characterized in that, The characteristics of the commutation switch area are the mean, variance, rate of change, total harmonic distortion, harmonic content, and Hurst exponent of the three-phase unbalance.

3. The method for recognizing three-phase imbalance patterns in a commutation switch area according to claim 1, characterized in that, The stability calculation method is as follows: divide the feature and feature change into intervals, arbitrarily pair the feature intervals and feature change intervals to obtain multiple combinations, calculate the probability of the label appearing in different combinations, calculate the entropy value based on the probability, and normalize the negative of the entropy value to obtain the stability.

4. The method for identifying three-phase imbalance patterns in a commutation switch area according to claim 1, characterized in that, The recognition model is a BP neural network model.

5. The method for identifying three-phase imbalance patterns in a commutation switch area according to claim 1, characterized in that, The linear redundancy is calculated as follows: calculate the Pearson correlation coefficient or variance inflation factor between two features, take the absolute value of the Pearson correlation coefficient to obtain the linear redundancy, or map the difference between the variance inflation factor and 1 through the tanh function to obtain the linear redundancy.

6. The method for recognizing three-phase imbalance patterns in a commutation switch area according to claim 1, characterized in that, The nonlinear redundancy is calculated as follows: the mutual information between two features is calculated, and the nonlinear redundancy is obtained by mapping the mutual information through the tanh function, or the nonlinear redundancy is obtained by calculating the distance correlation coefficient between two features.

7. The method for recognizing three-phase imbalance patterns in a commutation switch area according to claim 1, characterized in that, The probability that the distribution of the target feature conforms to a normal distribution is obtained through the KS test.

8. A three-phase imbalance pattern recognition system for a commutation switch area, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a three-phase imbalance pattern recognition method for a commutation switch area according to any one of claims 1-7.

Images