Low-voltage charging pile abnormal power consumption identification method and system based on XGBoost algorithm

Abstract

The invention discloses a low-voltage charging pile abnormal power consumption identification method and system based on an XGBoost algorithm. The method comprises the following steps: acquiring ledgers, time-sharing point currents and continuous multi-day power data of charging pile users and non-charging pile users; preprocessing the data to eliminate noise; extracting statistical dimension and business dimension data features, and marking to form a training sample set; constructing an XGBoost algorithm model and training the XGBoost algorithm model; for all charging piles in districts and counties, the model is used for identifying charging pile users with abnormal electricity consumption through current data; non-electric vehicle charging loads are identified and filtered based on a filtering algorithm through continuous multi-day power data, the non-charging load proportion is calculated, and detailed list of charging pile users using electricity abnormally is obtained; and feeding back the list of the charging pile users with abnormal electricity consumption to the city, and calculating the model precision ratio through the on-site checking condition of a marketing team. According to the invention, the efficiency and accuracy of abnormal power consumption identification of the low-voltage charging pile are obviously improved.

Description

Technical Field

[0001] The present invention belongs to the field of power systems, and in particular relates to a method and system for identifying abnormal power consumption of a low-voltage charging pile based on an XGBoost algorithm. Background Art

[0002] The penetration rate of new energy electric vehicles is rising sharply, and the construction of corresponding charging facilities is also being vigorously promoted.

[0003] At present, electric vehicle charging piles are charged by independent meters. In order to avoid being restricted by the tiered electricity prices for residents, some customers will connect other electrical equipment to the low-voltage charging pile meter by means of lead wires and plug strips, which can easily cause the overall load to exceed the rated capacity of the charging pile, bringing about major safety hazards and even fire in serious cases. Therefore, it is urgent to carry out targeted investigation and management of the mixed electricity consumption of charging piles.

[0004] Traditional verification methods mainly rely on manual experience to check power current and power curves, which have problems such as inability to conduct full verification, low verification efficiency and inability to detect anomalies in a timely manner. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a method and system for identifying abnormal power consumption of low-voltage charging piles based on the XGBoost algorithm. The method uses digital means to realize the automatic identification of abnormal power consumption of low-voltage charging piles.

[0006] Technical means: The present invention is based on XGBoost The algorithm-based method for identifying abnormal power consumption of low-voltage charging piles includes: Measurement data acquisition: Obtain the records of charging pile users and non-charging pile users, time-divided current, and time-divided power data for multiple consecutive days from the electricity consumption information collection system; Measurement data preprocessing: preprocess the acquired data, use the isnull function to identify missing values, and use the Lagrange polynomial interpolation method to fill in the identified missing values; use the average value of the absolute value of the first-order difference of the current data to pre-screen the users of charging piles in the ledger and screen out users of charging piles with abnormal current consumption; Feature extraction: The current curve of charging pile users is used as the current data of normal charging pile power consumption, and is labeled as normal power consumption; the current curve of non-charging pile users is used as the current data of abnormal charging pile power consumption, and is labeled as abnormal power consumption; multiple statistical dimension data features and business dimension data features are extracted from the pre-processed current data of charging pile users and non-charging pile users, and combined with the normal power consumption and abnormal power consumption labels to form a training sample set; Model building and training: selection XGBoost Algorithm model, set according to sample information XGBoostThe initial parameter range of the algorithm model builds the basic XGBoost Algorithm model; import the feature data of the training sample set into the basic XGBoost The algorithm model is used for model training. XGBoost In the process of algorithm modeling, the grid search method is used to adjust the parameters within the range, and the model is selected to achieve the best prediction accuracy on the training sample set, and the prediction result based on XGBoost An algorithm-based model for identifying abnormal power consumption in low-voltage charging piles; Model prediction: Obtain the current data of all charging piles in the district and county, and use XGBoost The algorithm's low-voltage charging pile abnormal power consumption recognition model is used to identify and obtain charging pile users with abnormal power consumption; Identification of non-charging loads: For the charging pile users with abnormal power consumption, a filtering algorithm based on the power first-order difference threshold is used to perform multiple smoothing filters on their time-sharing power data for multiple consecutive days. At the same time, combined with the historical average power peak value of charging electric vehicles, the non-charging load is separated, the non-electric vehicle charging load is identified and filtered, the non-charging load ratio is calculated, and finally a list of charging pile users with abnormal power consumption is obtained; Algorithm testing: Feedback the list of charging pile users with abnormal electricity consumption to the prefecture-level cities, obtain the on-site verification results of the marketing team, and calculate the XGBoost The precision rate of the algorithm's low-voltage charging pile abnormal power consumption identification model.

[0007] Furthermore, the acquired data is preprocessed, the missing values ​​are identified using the isnull function, and the identified missing values ​​are filled using the Lagrange polynomial interpolation method; the average value of the absolute value of the first-order difference of the current data is used to pre-screen the users of charging piles in the ledger, and the users of charging piles with abnormal power current are screened out, including: First, remove irrelevant and duplicate data in the measurement curve data set, and use Python The isnull function in the measurement curve dataset identifies special identifiers NaN , NULL , NA , thereby identifying missing values; Then, the identified missing values ​​are interpolated using the Lagrange polynomial, which is calculated as follows: ) in, Represents the polynomial function obtained by Lagrange interpolation, which is used to estimate the given position of the measured data point x The interpolation result at the value; Indicates the number of existing measurement data points; Indicates The measured data value of each data point; ) Indicates i Lagrangian basis functions of the measured data values; Finally, the average value of the absolute value of the first-order difference of the current data is used to pre-screen the users whose accounts are charging piles. The calculation formula is as follows: ; By limiting AVG Less than 1, preliminarily screen out charging pile users with abnormal power consumption current.

[0008] The statistical dimension data features include the standard deviation, coefficient of variation, kurtosis, skewness, maximum value and median value of the user's time-divided current measurement data.

[0009] The standard deviation of the user's time-sharing point current measurement data , which is used to describe the discrete degree and fluctuation range of data. Its calculation formula is as follows: ; in, Indicates the specific value in the original data; represents the mean of the original data; Indicates the amount of data; The coefficient of variation of the user's time-sharing point current measurement data , which is used to describe the degree of dispersion of data relative to its average and reflects the relative fluctuation characteristics of data. The calculation formula is as follows:

[0010] Furthermore, the kurtosis of the user's time-divided current measurement data , which is used to describe the sharpness or flatness of the data, is calculated as follows: ; The skewness of the user's time-sharing current measurement data , which is used to describe the direction and degree of data offset, and its expression is as follows:

[0011] Furthermore, the business dimension data features include the number of points where the user's time-sharing point current is less than 1.2 times the valley current, the number of points where the user's time-sharing point current is greater than 0.8 times the peak current, the number of points where the user's time-sharing point current is greater than 1.5 times the median or less than 0.5 times the median, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing point current is less than 0.1, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing point current is greater than 0.8 times the peak current, and the Pearson correlation coefficient between the user's time-sharing point current on the same day and the time-sharing point current on similar days.

[0012] Furthermore, the list of charging pile users with abnormal electricity consumption will be fed back to the prefecture-level cities, and the on-site inspection results of the marketing team will be obtained to calculate the XGBoost The accuracy of the algorithm's low-voltage charging pile abnormal power consumption recognition model includes: Based on XGBoost The recognition results of the algorithm's low-voltage charging pile abnormal power consumption identification model will systematically feed back the list of charging pile users determined to be abnormal power users to the corresponding district and county marketing teams; subsequently, the district and county marketing teams will conduct detailed on-site verification of the list of charging pile users with abnormal power usage; finally, by summarizing and analyzing the verification results of the on-site teams, the precision of the model in identifying abnormal power consumption of charging piles is scientifically calculated, which serves as the basis for evaluating model performance and subsequent optimization directions.

[0013] Based on the same inventive concept, a method of the present invention is based on XGBoost The algorithm-based low-voltage charging pile abnormal power consumption identification system includes: The measurement data acquisition module is used to obtain the ledgers of charging pile users and non-charging pile users, time-divided current and time-divided power data for multiple consecutive days from the electricity consumption information collection system; The measurement data preprocessing module is used to preprocess the acquired data, identify missing values ​​using the isnull function, and fill the identified missing values ​​using the Lagrange polynomial interpolation method; the average value of the absolute value of the first-order difference of the current data is used to pre-screen the users of charging piles in the ledger and screen out users of charging piles with abnormal current consumption; The feature extraction module is used to use the current curve of the charging pile user as the current data of the normal power consumption of the charging pile, and mark it with a normal power consumption label; use the current curve of the non-charging pile user as the current data of the abnormal power consumption of the charging pile, and mark it with an abnormal power consumption label; extract multiple statistical dimension data features and business dimension data features from the pre-processed current data of the charging pile user and the non-charging pile user, and combine the normal power consumption and abnormal power consumption labels to form a training sample set; Model building and training module, used to select XGBoost Algorithm model, set according to sample information XGBoost The initial parameter range of the algorithm model builds the basic XGBoost Algorithm model; import the feature data of the training sample set into the basic XGBoost The algorithm model is used for model training. XGBoost In the process of algorithm modeling, the grid search method is used to adjust the parameters within the range, and the model is selected to achieve the best prediction accuracy on the training sample set, and the prediction result based on XGBoost An algorithm-based model for identifying abnormal power consumption in low-voltage charging piles; The model prediction module is used to obtain the current data of all charging piles in the district and county, using XGBoost The algorithm's low-voltage charging pile abnormal power consumption recognition model is used to identify and obtain charging pile users with abnormal power consumption; The non-charging load identification module is used to perform multiple smoothing filters on the power data of the charging pile users with abnormal power consumption for multiple consecutive days by using a filtering algorithm based on the power first-order difference threshold. At the same time, the non-charging load is separated based on the historical average power peak value of the charging electric vehicle, and the non-electric vehicle charging load is identified and filtered. The proportion of non-charging load is calculated, and finally a list of charging pile users with abnormal power consumption is obtained; The algorithm testing module is used to feed back the list of charging pile users with abnormal electricity consumption to the prefecture-level cities, and calculate the XGBoost The precision rate of the algorithm's low-voltage charging pile abnormal power consumption identification model.

[0014] Based on the same inventive concept, a method of the present invention is based on XGBoost The device for identifying abnormal power consumption of a low-voltage charging pile based on an algorithm is characterized in that it includes a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device realizes the above-mentioned abnormal power consumption recognition based on XGBoost The steps of the method for identifying abnormal power consumption of a low-voltage charging pile based on an algorithm.

[0015] Based on the same inventive concept, a computer-readable storage medium of the present invention stores a computer program, which, when executed by a processor, implements the above-mentioned XGBoost The steps of the method for identifying abnormal power consumption of a low-voltage charging pile based on an algorithm.

[0016] Beneficial effects: Compared with the prior art, the significant technical effects of the present invention are: This solution obtains user records, current at different time points, and power data for multiple consecutive days from the electricity consumption information collection system. Deeply mine data characteristics, obtain statistical dimension and business dimension characteristics, and establish a XGBoost The algorithm uses a low-voltage charging pile abnormal power consumption recognition model to identify charging pile users with abnormal power consumption.

[0017] Through experimental verification, the present invention is based on XGBoost The algorithm's low-voltage charging pile abnormal power consumption recognition model has demonstrated excellent performance and significant improvement in recognition tasks. The model fully utilizes the powerful ability of machine learning in processing sequence data, significantly improving the accuracy and efficiency of identifying abnormal power consumption by charging pile users, effectively ensuring power grid security, eliminating potential safety hazards, and reducing operating costs.

[0018] The identification method adopted by the present invention is highly automated, greatly reducing the reliance on manual operation, thereby effectively improving the operating efficiency and safety performance of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The embodiment of the present invention discloses a method based on XGBoost A flowchart of a method for identifying abnormal power consumption of a low-voltage charging pile based on an algorithm; Figure 2 It is a schematic diagram of a current curve of a charging pile user and a current curve of a residential user disclosed in an embodiment of the present invention; Figure 3 It is a schematic diagram of a process of smoothing and filtering a power curve using a filtering algorithm of a first-order difference threshold disclosed in an embodiment of the present invention; Figure 4 The embodiment of the present invention discloses a method based on XGBoost Structural diagram of the abnormal power consumption identification system of low-voltage charging piles based on the algorithm; Figure 5 The embodiment of the present invention discloses a method based on XGBoost Schematic diagram of the structure of the low-voltage charging pile abnormal power consumption identification device based on the algorithm. DETAILED DESCRIPTION

[0020] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It will be understood by those skilled in the art that the purposes and advantages that can be achieved with the present invention are not limited to the specific description of the above beneficial effects, and the above and other purposes that can be achieved by the present invention will be more clearly understood based on the following detailed description.

[0021] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed in the present invention can be implemented in hardware, software or a combination of the two. Whether it is implemented in hardware or software depends on the specific application and design of the technical solution. Professional and technical personnel can use different methods to implement the functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0022] Reference to "embodiments" in the present invention means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0023] Example 1: Please refer to Figure 1 , Figure 1 The embodiment of the present invention discloses a method based on XGBoost Flow chart of the method for identifying abnormal power consumption of low-voltage charging piles based on the algorithm. Figure 1 The described method for identifying abnormal power consumption of a low-voltage charging pile is applied to a power system, such as identifying abnormal power consumption of a low-voltage charging pile, etc., which is not limited in the embodiments of the present invention. Figure 1 As shown, based on XGBoost The algorithm-based method for identifying abnormal power consumption of low-voltage charging piles may include the following operations: S1. Acquisition of measurement data: From the electricity consumption information collection system, the ledgers of charging pile users and non-charging pile users, time-divided current and time-divided power data for multiple consecutive days are obtained.

[0024] Since the abnormal electricity consumption among charging pile users is the same as that of residential users, in order to identify the abnormal electricity consumption behavior of charging pile users, the data of charging pile users and non-charging pile users are selected for training to obtain their time-sharing current and time-sharing power data for multiple consecutive days.

[0025] S2. Preprocessing of measurement data: Preprocess the acquired data, use the isnull function to identify missing values, and use the Lagrange polynomial interpolation method to fill in the identified missing values; use the average value of the absolute value of the first-order difference of the current data to pre-screen the charging pile users in the ledger, screen out charging pile users with abnormal power consumption, and ensure the purity of the training data.

[0026] During the data collection process, due to the influence of various factors, the raw data often contains noise. In order to ensure the accuracy and reliability of the data, it is very important to preprocess the raw data. Preprocessing mainly includes data cleaning and eliminating missing values, thereby reducing the negative impact on model training. Step S2 specifically includes the following: S2.1. Delete irrelevant and duplicate data in the measurement curve data set, and use Python The isnull function in the measurement curve dataset identifies special identifiers NaN , NULL , NA , thereby identifying missing values; S2.2. Use Lagrange polynomials to interpolate the identified missing values. The calculation formula is as follows: ); in, Represents the polynomial function obtained by Lagrange interpolation, which is used to estimate the given position of the measured data point x The interpolation result at the value; Indicates the number of existing measurement data points; Indicates The measured data value of each data point; ) Indicates i Lagrangian basis functions of the measured data values; S2.3. Use the average value of the absolute value of the first-order difference of the current data to pre-screen the users of charging piles in the ledger, and screen out the users of charging piles with abnormal current consumption to ensure the purity of the training data. The calculation formula is as follows: ; After in-depth analysis of the daily current curve characteristics of a large number of charging piles, it was found that the average absolute value of the first-order difference of the current of these charging piles is AVG Less than 1. By limiting AVG Less than 1, preliminarily screen out charging pile users with abnormal power consumption to ensure the purity of training data.

[0027] S3. Feature extraction: The current curve of the charging pile user's electricity consumption is used as the current data of the charging pile when the electricity consumption is normal, and a normal electricity consumption label is added; the current curve of the non-charging pile user's electricity consumption is used as the current data of the charging pile when the electricity consumption is abnormal, and an abnormal electricity consumption label is added; from the pre-processed current data of the charging pile user and the non-charging pile user, a number of statistical dimension data features and business dimension data features are extracted, and combined with the normal electricity consumption and abnormal electricity consumption labels to form a training sample set.

[0028] 6-dimensional statistical features and 6-dimensional business features are extracted from the 96-dimensional time-sharing current data of the user on the same day after preprocessing, including: 6-dimensional statistical dimension features of the standard deviation, coefficient of variation, kurtosis, skewness, maximum value and median of the user's time-sharing current measurement data; the number of points where the user's time-sharing current is less than 1.2 times the valley current, the number of points where the user's time-sharing current is greater than 0.8 times the peak current, the number of points where the user's time-sharing current is greater than 1.5 times the median or less than 0.5 times the median, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing current is less than 0.1, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing current is greater than 0.8 times the peak current, and the 6-dimensional business dimension features of the Pearson correlation coefficient between the user's time-sharing current on the same day and the time-sharing current on similar days. Finally, the above 12-dimensional feature data are selected, combined with the normal power consumption and abnormal power consumption labels to form a training sample set.

[0029] Step S3 specifically includes the following steps: S3.1. The current curve of the charging pile user is used as the current data of the normal power usage of the charging pile, and a normal power usage label is added; the current curve of the non-charging pile user is used as the current data of the abnormal power usage of the charging pile, and an abnormal power usage label is added; Since it is impossible to obtain the label of whether the charging pile is using electricity abnormally, the current curve of the charging pile user's electricity consumption is used as the current data when the charging pile is using electricity normally, and the current curve of the non-charging pile user's electricity consumption is used as the current data when the charging pile is using electricity abnormally.

[0030] S3.2. Obtain statistical dimension features through user active power data, including standard deviation, coefficient of variation, kurtosis, skewness, maximum value and median of user time-divided current data. The calculation process of skewness and kurtosis is as follows: Standard deviation of user's time-sharing current measurement data , which is used to describe the discrete degree and fluctuation range of data. Its calculation formula is as follows: ; in, Indicates the specific value in the original data; represents the mean of the original data; Indicates the amount of data.

[0031] The coefficient of variation of the user's time-sharing current measurement data , which is used to describe the degree of dispersion of data relative to its average and reflects the relative fluctuation characteristics of data. The calculation formula is as follows: .

[0032] The kurtosis of the user's time-sharing current measurement data , which is used to describe the sharpness or flatness of the data, is calculated as follows: .

[0033] The skewness of the user's time-sharing current measurement data , which is used to describe the direction and degree of data offset, and its expression is as follows: .

[0034] S3.3. Obtaining business dimension features through user current data is conceived as follows: S3.3.1. Considering that the current of the charging pile user should be 0A or only a small current during the non-electric vehicle charging stage, the number of points where the user's time-sharing point current is less than 1.2 times the valley current is calculated; S3.3.2. Considering that the current of charging pile users tends to be stable or fluctuates within a small range during the electricity consumption stage, and the charging time of a large number of low-voltage users is within a certain range, the number of points where the current of the user's time-sharing point is greater than 0.8 times the peak current is calculated; S3.3.3. Considering that the current of the charging pile users in the power consumption and non-power consumption stages should hardly have points near the median, therefore, calculate the number of points where the current of the user's time-sharing points is greater than 1.5 times the median or less than 0.5 times the median; S3.3.4. Considering that the current of charging pile users hardly fluctuates or fluctuates within a very small range during the power consumption and non-power consumption stages, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing current is less than 0.1 is calculated; S3.3.5. Considering that the current of the charging pile user changes suddenly during the power consumption and non-power consumption phases, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing current is greater than 0.8 times the peak current is calculated; S3.3.6. Considering that street lamps and charging piles have very similar curves, and the street lamps are turned on and off at fixed times every day, the Pearson correlation coefficient between the user's current at different time points on the same day and the current at different time points on similar days is calculated.

[0035] S3.4. Based on the multiple statistical dimension data features and business dimension data features extracted in step S3.2 and step S3.3, combined with normal power consumption and abnormal power consumption labels, a training sample set is formed.

[0036] S4. Model construction and training: selection XGBoost Algorithm model, set according to sample information XGBoost The initial parameter range of the algorithm model builds the basic XGBoost Algorithm model; import the feature data of the training sample set into the basic XGBoost The algorithm model is used for model training. XGBoost In the process of algorithm modeling, the grid search method is used to adjust the parameters within the range, and the model is selected to achieve the best prediction accuracy on the training sample set, and the prediction result based on XGBoost An abnormal power consumption identification model for low-voltage charging piles based on an algorithm.

[0037] Loading basic XGBoost Algorithm model, set the parameter range, use grid search to adjust the parameters within the range, and finally use the test set to evaluate the performance of the model.

[0038] In this embodiment, the specific model construction, training and testing process is as follows: S4.1. Loading the basic XGBoost Algorithm model, set the range of main parameters according to the data, including learning rate, number of trees, maximum depth, etc., and use the default values ​​for other parameters, as follows: S4.1.1, eta : Learning rate. This controls the contribution rate of each tree to the final result, which is used to prevent overfitting. The range is limited to [0.01, 0.3]; S4.1.2, num_boost_round : The number of trees. This specifies the number of trees to be generated during the training process. There is no fixed upper limit, but too many trees will increase the computational cost and may cause overfitting. The range is limited to [100,500]; S4.1.3, max_depth : Maximum depth. This controls the maximum depth of the tree to limit the complexity of the model. Deeper trees can capture more complex patterns, but may also lead to overfitting. The range is limited to [3,10]; S4.1.4, subsample : Subsample ratio. This is the ratio of samples used to train each tree to the total samples. A smaller value can increase the diversity of the model and help prevent overfitting. The range is limited to [0.5, 1]; S4.1.5, colsample_bytree : Feature sampling ratio. This is the ratio used to sample features when building a tree. Smaller values ​​can increase the diversity of the model. The range is limited to [0.5, 1]; S4.1.6, min_child_weight : Minimum child node weight sum. This is the minimum sample weight sum required for leaf node splitting, and the range is limited to [0,10].

[0039] S4.2. Import the feature data of the training sample set into XGBoost In the algorithm model. XGBoost In the process of algorithm modeling, the grid search method is used to adjust the parameters within the range and select the model to achieve the best prediction accuracy on the training sample set. XGBoost An abnormal power consumption identification model for low-voltage charging piles based on an algorithm.

[0040] S5. Model prediction: Obtain the current data of all charging piles in the district and county, and use XGBoost The algorithm's low-voltage charging pile abnormal power consumption recognition model is used to identify and obtain charging pile users with abnormal power consumption.

[0041] A complete data set of charging piles is collected from within a certain district or county, and then the preset model is used to accurately identify and classify the data set to obtain charging pile users with abnormal electricity consumption.

[0042] S6. Identification of non-charging loads: For the above-mentioned charging pile users with abnormal electricity consumption, a filtering algorithm based on the power first-order difference threshold is used to perform multiple smoothing filters on their time-point power data for multiple consecutive days. At the same time, combined with the historical average power peak value of charging electric vehicles, the non-charging load (spike wave) is separated, the non-electric vehicle charging load is identified and filtered, and the proportion of non-charging load is calculated. Finally, a list of charging pile users with abnormal electricity consumption is obtained.

[0043] S7, algorithm testing: Feedback the list of charging pile users with abnormal electricity consumption to the prefecture-level cities, obtain the on-site verification results of the marketing team, and calculate the algorithm based on XGBoost The precision rate of the algorithm's low-voltage charging pile abnormal power consumption identification model.

[0044] Based on XGBoost The recognition results of the algorithm's low-voltage charging pile abnormal power consumption identification model will systematically feed back the list of charging pile users who are judged to be abnormal power users to the corresponding district and county marketing teams; subsequently, the district and county marketing teams will conduct detailed on-site verification of the list of charging pile users with abnormal power consumption, with the main purpose of verifying whether these users do have abnormal power consumption behavior or misjudgment; finally, by summarizing and analyzing the verification results of the on-site teams, the precision of the model in identifying abnormal power consumption of charging piles is scientifically calculated, which serves as an important basis for evaluating model performance and subsequent optimization directions.

[0045] The present invention is described below through specific embodiments. XGBoost The algorithm-based method for identifying abnormal power consumption of low-voltage charging piles is further introduced.

[0046] like Figure 1 As shown, the embodiment of the present invention provides a method based on XGBoost The method for identifying abnormal power consumption of a low-voltage charging pile based on an algorithm specifically includes the following steps: S1. Data acquisition: Obtain the records of all non-charging pile users and charging pile users, obtain their time-sharing current and continuous time-sharing power data for multiple days, and preliminarily remove some users with questionable labels in the records based on the electricity consumption characteristics of charging pile users and non-charging pile users, and randomly obtain records and current data with labels of 100,000 groups of charging pile users and 100,000 groups of non-charging pile users. The electricity consumption pattern of charging pile users is different from that of residential users. The current curve of charging pile users is mostly rectangular waves, while there is no pattern for residential users. For example, Figure 2 shown.

[0047] S2. Data preprocessing: Preprocess the acquired data using isnull The function identifies the missing values ​​and fills them with Lagrange polynomial interpolation. Finally, the average of the absolute value of the first-order difference of the current data is used to pre-screen the users of charging piles in the ledger, and to screen out users of charging piles with abnormal current consumption to ensure the purity of the training data.

[0048] S3. Feature extraction: Extract 6-dimensional statistical features and 6-dimensional business features from the pre-processed 96-dimensional time-sharing current data of the user on the same day, including: 6-dimensional statistical dimension features of the standard deviation, coefficient of variation, kurtosis, skewness, maximum value and median of the user's time-sharing current data; the number of points where the user's time-sharing current is less than 1.2 times the valley current, the number of points where the user's time-sharing current is greater than 0.8 times the peak current, the number of points where the user's time-sharing current is greater than 1.5 times the median or less than 0.5 times the median, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing current is less than 0.1, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing current is greater than 0.8 times the peak current, and the 6-dimensional business dimension features of the Pearson correlation coefficient between the user's time-sharing current on the same day and the time-sharing current on similar days. By classifying charging pile users and non-charging pile users, labeling normal electricity consumption and abnormal electricity consumption, and combining the above features, a training sample set is formed.

[0049] S4, model building, training: loading basic XGBoost Algorithm model, set parameter range, use grid search to adjust parameters within the range, and finally use the test set to evaluate the performance of the model. The specific process is as follows: S4.1. First load the basic XGBoost Algorithm model, according to data characteristics, learning rate eta Set to a value in the list [0.01, 0.05, 0.1, 0.2, 0.3], the number of trees num_boost_round Set to a value in the list [100, 200, 300, 400, 500], the maximum depth max_depth Set to the value in the list [3, 4, 5, 6, 7, 8, 9, 10], subsample ratio subsample Set to a value in the list [0.5, 0.7, 0.8, 0.9, 1.0], feature sampling ratio colsample_bytree Set to a value in the list [0.5, 0.6, 0.7, 0.8, 0.9, 1.0], minimum child node weight and min_child_weight Set to a value in the list [1, 3, 5, 7, 10].

[0050] S4.2. Import the feature data of the training sample set into XGBoost In the algorithm model, during training XGBoost In the algorithm model process, the grid search method is used to adjust the parameters within the range described above. By limiting the maximum number of iterations to 500, the final model selects the learning rate eta is 0.1, the number of trees num_boost_round 400, maximum depth max_ depth 7, sub-sample ratio colsample_bytreeis 0.8, feature sampling ratio colsample_bytree is 0.9, the minimum child node weight and min_child_weight The value is 7, and a low-voltage charging pile abnormal power consumption recognition model is trained based on this parameter.

[0051] S5. Model prediction: A complete data set of charging piles was collected from a certain district or county, totaling 23,241 sets of data, and then the preset model was used to accurately identify and classify the data set. 98 charging pile users with abnormal electricity consumption were obtained.

[0052] S6. Identification of non-charging loads: For the above-mentioned charging pile users with abnormal power consumption, a filtering algorithm based on the power first-order difference threshold is used to perform multiple smoothing filters on their power data at different time points for multiple consecutive days. At the same time, combined with the historical average power peak value of charging electric vehicles, the non-charging load (spike wave) is separated, the non-electric vehicle charging load is identified and filtered, the non-charging load ratio is calculated, and finally a list of charging pile users with abnormal power consumption is obtained. The specific filtering process is as follows: Figure 3 shown.

[0053] S7, Algorithm testing: Based on the recognition results of the model, some users who are judged to be abnormal electricity users are systematically fed back to the corresponding district and county marketing teams. Subsequently, the district and county marketing teams conducted detailed on-site inspections on these suspected abnormal electricity users of charging piles, with the main purpose of verifying whether these users do have abnormal electricity usage behaviors or misjudgments. Finally, by summarizing and analyzing the verification results of the on-site teams, 96 of them had abnormal electricity usage behaviors, and the precision rate of the model reached 97.96%.

[0054] In summary, the technical solution of the present invention obtains the time-sharing current and time-sharing power data of charging pile users and non-charging pile users by obtaining the ledger labels of charging pile users and non-charging pile users; pre-processes the current and power data, and pre-screens out charging pile users with abnormal current data by combining the average value of the absolute value of the first-order difference; extracts 6-dimensional statistical dimension features and 6-dimensional business dimension features from the pre-processed current data of charging pile users and non-charging pile users, classifies charging pile users and non-charging pile users, labels normal electricity consumption and abnormal electricity consumption, and combines the statistical dimension and business dimension features to form a training sample set construction. XGBoost The algorithm identifies the model, trains the model, predicts the data of a certain district and county, and obtains the charging pile users with abnormal power current; then, through the power data of multiple consecutive days, the filtering algorithm based on the power first-order difference threshold is used to identify and filter the non-electric vehicle charging load, calculate the proportion of non-charging load, and obtain the details of the charging pile user list with abnormal power consumption; the above-mentioned charging pile user list with abnormal power consumption is fed back to the prefecture-level city, and the marketing team verifies the situation on site to calculate the accuracy of the algorithm. The present invention is based on XGBoostThe algorithm model significantly improves the efficiency and accuracy of identifying abnormal power consumption of low-voltage charging piles.

[0055] Example 2: Please refer to Figure 4 , Figure 4 The embodiment of the present invention discloses a method based on XGBoost The structural diagram of the low-voltage charging pile abnormal power consumption recognition system based on the algorithm, which can realize the abnormal power consumption recognition of low-voltage charging piles, specifically includes: The measurement data acquisition module is used to obtain the ledgers of charging pile users and non-charging pile users, time-divided current and time-divided power data for multiple consecutive days from the electricity consumption information collection system; The measurement data preprocessing module is used to preprocess the acquired data, identify missing values ​​using the isnull function, and fill the identified missing values ​​using the Lagrange polynomial interpolation method; the average value of the absolute value of the first-order difference of the current data is used to pre-screen the users of charging piles in the ledger and screen out users of charging piles with abnormal current consumption; The feature extraction module is used to use the current curve of the charging pile user as the current data of the normal power consumption of the charging pile, and mark it with a normal power consumption label; use the current curve of the non-charging pile user as the current data of the abnormal power consumption of the charging pile, and mark it with an abnormal power consumption label; extract multiple statistical dimension data features and business dimension data features from the pre-processed current data of the charging pile user and the non-charging pile user, and combine the normal power consumption and abnormal power consumption labels to form a training sample set; Model building and training module, used to select XGBoost Algorithm model, set according to sample information XGBoost The initial parameter range of the algorithm model builds the basic XGBoost Algorithm model; import the feature data of the training sample set into the basic XGBoost The algorithm model is used for model training. XGBoost In the process of algorithm modeling, the grid search method is used to adjust the parameters within the range, and the model is selected to achieve the best prediction accuracy on the training sample set, and the prediction result based on XGBoost An algorithm-based model for identifying abnormal power consumption in low-voltage charging piles; The model prediction module is used to obtain the current data of all charging piles in the district and county, using XGBoost The algorithm's low-voltage charging pile abnormal power consumption recognition model is used to identify and obtain charging pile users with abnormal power consumption; The non-charging load identification module is used to perform multiple smoothing filters on the power data of the charging pile users with abnormal power consumption for multiple consecutive days using a filtering algorithm based on the power first-order difference threshold. At the same time, the non-charging load is separated based on the historical average power peak value of the charging electric vehicle, and the non-electric vehicle charging load is identified and filtered. The proportion of non-charging load is calculated, and finally a list of charging pile users with abnormal power consumption is obtained; The algorithm testing module is used to feed back the list of charging pile users with abnormal electricity consumption to the prefecture-level cities, and calculate the XGBoost The precision rate of the algorithm's low-voltage charging pile abnormal power consumption identification model.

[0056] In an optional embodiment, based on XGBoost The algorithm-based method for identifying abnormal power consumption in low-voltage charging piles includes: a) obtaining the ledgers of charging pile users and non-charging pile users, current at different time points, and power data for multiple consecutive days; b) preprocessing the data to eliminate noise; c) extracting 6-dimensional statistical dimensions and 6-dimensional business dimension features from the preprocessed data, marking normal power consumption and abnormal power consumption labels for the data set to form a training sample set; d) constructing XGBoost The algorithm identifies the model and trains the model; e) For all charging piles in a certain district or county, the model is used to identify the users of charging piles with abnormal electricity usage through current data; f) Through multiple consecutive days of time-sharing power data, the filtering algorithm based on the power first-order difference threshold is used to identify and filter the charging load of non-electric vehicles, calculate the proportion of non-charging load, and obtain the details of the list of users of charging piles with abnormal electricity usage; g) The above-mentioned list of users of charging piles with abnormal electricity usage is fed back to the prefecture-level city, and the marketing team verifies the situation on site and calculates the accuracy of the algorithm.

[0057] Example 3: Please refer to Figure 5 , Figure 5 The embodiment of the present invention discloses a method based on XGBoost The structural diagram of the low-voltage charging pile abnormal power consumption identification device based on the algorithm. Figure 5 The described device can be applied to a power system, such as for identifying abnormal power consumption of a low-voltage charging pile, etc., and the embodiments of the present invention are not limited thereto.

[0058] like Figure 5 As shown, the device may include a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device implements the steps of the method described in the above embodiment and can achieve the technical effect consistent with the above method.

[0059] The memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the memory may be used to read and write non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). A program / utility having a set (at least one) of program modules may be stored in, for example, the memory, such program modules including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include the implementation of a network environment. The program modules typically perform the functions and / or methods in the embodiments described herein.

[0060] The processor executes various functional applications and data processing by running the programs stored in the memory, such as implementing the method provided in Embodiment 1 of the present invention.

[0061] Embodiment 4: Embodiment 4 of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in the above embodiment are implemented and the technical effect consistent with the above method can be achieved.

[0062] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0063] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0064] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0065] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0066] Of course, the storage medium containing computer executable instructions provided by an embodiment of the present invention, whose computer executable instructions are not limited to the above method operations, can also execute related operations in the method provided by any embodiment of the present invention.

[0067] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method based on XGBoost The method for identifying abnormal power consumption of low-voltage charging piles based on an algorithm is characterized in that: include: Measurement data acquisition: Obtain the records of charging pile users and non-charging pile users, time-divided current, and time-divided power data for multiple consecutive days from the electricity consumption information collection system; Measurement data preprocessing: preprocess the acquired data, use the isnull function to identify missing values, and use the Lagrange polynomial interpolation method to fill in the identified missing values; use the average value of the absolute value of the first-order difference of the current data to pre-screen the users of charging piles in the ledger and screen out users of charging piles with abnormal current consumption; Feature extraction: The current curve of the charging pile user's electricity consumption is used as the current data of the charging pile when it is in normal electricity consumption, and a normal electricity consumption label is added; The current curve of non-charging pile users is used as the current data of abnormal charging pile power consumption, and an abnormal power consumption label is marked; Extract multiple statistical dimension data features and business dimension data features from the pre-processed current data of charging pile users and non-charging pile users, and combine them with normal power consumption and abnormal power consumption labels to form a training sample set; Model building and training: selection XGBoost Algorithm model, set according to sample information XGBoost The initial parameter range of the algorithm model builds the basic XGBoost Algorithm model; import the feature data of the training sample set into the basic XGBoost The algorithm model is used for model training. XGBoost In the process of algorithm modeling, the grid search method is used to adjust the parameters within the range, and the model is selected to achieve the best prediction accuracy on the training sample set, and the prediction result based on XGBoost An algorithm-based model for identifying abnormal power consumption in low-voltage charging piles; Model prediction: Obtain the current data of all charging piles in the district and county, and use XGBoost The algorithm's low-voltage charging pile abnormal power consumption recognition model is used to identify and obtain charging pile users with abnormal power consumption; Identification of non-charging loads: For the charging pile users with abnormal power consumption, a filtering algorithm based on the power first-order difference threshold is used to perform multiple smoothing filters on their time-sharing power data for multiple consecutive days. At the same time, combined with the historical average power peak value of charging electric vehicles, the non-charging load is separated, the non-electric vehicle charging load is identified and filtered, the non-charging load ratio is calculated, and finally a list of charging pile users with abnormal power consumption is obtained; Algorithm testing: Feedback the list of charging pile users with abnormal electricity consumption to the prefecture-level cities, obtain the on-site verification results of the marketing team, and calculate the XGBoost The precision rate of the algorithm's low-voltage charging pile abnormal power consumption identification model.

2. According to claim 1, based on XGBoost The method for identifying abnormal power consumption of low-voltage charging piles based on an algorithm is characterized in that: Preprocess the acquired data, use the isnull function to identify missing values, and use the Lagrange polynomial interpolation method to fill the missing values; The average value of the absolute value of the first-order difference of the current data is used to pre-screen the users of charging piles in the ledger, and to screen out users of charging piles with abnormal current consumption, including: First, remove irrelevant and duplicate data in the measurement curve data set, and use Python The isnull function in the measurement curve dataset identifies special identifiers NaN , NULL , NA , thereby identifying missing values; Then, the identified missing values ​​are interpolated using the Lagrange polynomial, which is calculated as follows: ) in, Represents the polynomial function obtained by Lagrange interpolation, which is used to estimate the given position of the measured data point x The interpolation result at the value; Indicates the number of existing measurement data points; Indicates The measured data value of each data point; ) Indicates i Lagrangian basis functions of the measured data values; Finally, the average value of the absolute value of the first-order difference of the current data is used to pre-screen the users of charging piles in the ledger. The calculation formula is as follows: ； By limiting AVG Less than 1, preliminarily screen out charging pile users with abnormal power consumption current.

3. According to claim 1 based on XGBoost The method for identifying abnormal power consumption of low-voltage charging piles based on an algorithm is characterized by: The statistical dimension data features include the standard deviation, coefficient of variation, kurtosis, skewness, maximum value and median value of the user's time-divided current measurement data.

4. The method according to claim 3 XGBoost The method for identifying abnormal power consumption of low-voltage charging piles based on an algorithm is characterized in that: The standard deviation of the user's time-sharing point current measurement data , which is used to describe the discrete degree and fluctuation range of data. Its calculation formula is as follows: ； in, Indicates the specific value in the original data; represents the mean of the original data; Indicates the amount of data; The coefficient of variation of the user's time-sharing point current measurement data , which is used to describe the degree of dispersion of data relative to its average and reflects the relative fluctuation characteristics of data. The calculation formula is as follows:

5. The method according to claim 3 XGBoost The method for identifying abnormal power consumption of low-voltage charging piles based on an algorithm is characterized in that: The kurtosis of the user's time-sharing current measurement data , which is used to describe the sharpness or flatness of the data, is calculated as follows: ； The skewness of the user's time-sharing current measurement data , which is used to describe the direction and degree of data offset, and its expression is as follows:

6. The method according to claim 1 XGBoost The method for identifying abnormal power consumption of low-voltage charging piles based on an algorithm is characterized by: The business dimension data features include the number of points where the user's time-sharing point current is less than 1.2 times the valley current, the number of points where the user's time-sharing point current is greater than 0.8 times the peak current, the number of points where the user's time-sharing point current is greater than 1.5 times the median or less than 0.5 times the median, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing point current is less than 0.1, the number of points where the absolute value of the increase or decrease between adjacent data points of the user's time-sharing point current is greater than 0.8 times the peak current, and the Pearson correlation coefficient between the user's time-sharing point current on the same day and the time-sharing point current on similar days.

7. The method according to claim 1 XGBoost The method for identifying abnormal power consumption of low-voltage charging piles based on an algorithm is characterized by: Feedback the list of charging pile users with abnormal electricity consumption to the prefecture-level cities, and obtain the on-site inspection results of the marketing team to calculate the XGBoost The accuracy of the algorithm's low-voltage charging pile abnormal power consumption recognition model includes: Based on XGBoost The recognition results of the algorithm's low-voltage charging pile abnormal power consumption identification model will systematically feed back the list of charging pile users determined to be abnormal power users to the corresponding district and county marketing teams; subsequently, the district and county marketing teams will conduct detailed on-site verification of the list of charging pile users with abnormal power usage; finally, by summarizing and analyzing the verification results of the on-site teams, the precision of the model in identifying abnormal power consumption of charging piles is scientifically calculated, which serves as the basis for evaluating model performance and subsequent optimization directions.

8. A method based on XGBoost The low-voltage charging pile abnormal power consumption identification system based on the algorithm is characterized by: include: The measurement data acquisition module is used to obtain the ledgers of charging pile users and non-charging pile users, time-divided current and time-divided power data for multiple consecutive days from the electricity consumption information collection system; The measurement data preprocessing module is used to preprocess the acquired data, identify missing values ​​using the isnull function, and fill the identified missing values ​​using the Lagrange polynomial interpolation method; the average value of the absolute value of the first-order difference of the current data is used to pre-screen the users of charging piles in the ledger and screen out users of charging piles with abnormal current consumption; A feature extraction module is used to use the current curve of the charging pile user as the current data of the normal power consumption of the charging pile and label it as normal power consumption; The current curve of non-charging pile users is used as the current data of abnormal charging pile power consumption, and an abnormal power consumption label is marked; Extract multiple statistical dimension data features and business dimension data features from the pre-processed current data of charging pile users and non-charging pile users, and combine them with normal power consumption and abnormal power consumption labels to form a training sample set; Model building and training module, used to select XGBoost Algorithm model, set according to sample information XGBoost The initial parameter range of the algorithm model builds the basic XGBoost Algorithm model; import the feature data of the training sample set into the basic XGBoost The algorithm model is used for model training. XGBoost In the process of algorithm modeling, the grid search method is used to adjust the parameters within the range, and the model is selected to achieve the best prediction accuracy on the training sample set, and the prediction result based on XGBoost An algorithm-based model for identifying abnormal power consumption in low-voltage charging piles; The model prediction module is used to obtain the current data of all charging piles in the district and county, using XGBoost The algorithm's low-voltage charging pile abnormal power consumption recognition model is used to identify and obtain charging pile users with abnormal power consumption; The non-charging load identification module is used to perform multiple smoothing filters on the power data of the charging pile users with abnormal power consumption for multiple consecutive days by using a filtering algorithm based on the power first-order difference threshold. At the same time, the non-charging load is separated based on the historical average power peak value of the charging electric vehicle, and the non-electric vehicle charging load is identified and filtered. The proportion of non-charging load is calculated, and finally a list of charging pile users with abnormal power consumption is obtained; The algorithm testing module is used to feed back the list of charging pile users with abnormal electricity consumption to the prefecture-level cities, and calculate the XGBoost The precision rate of the algorithm's low-voltage charging pile abnormal power consumption identification model.

9. A method based on XGBoost The low-voltage charging pile abnormal power consumption identification device based on the algorithm is characterized by: The electronic device comprises a processor and a memory, wherein the memory stores computer instructions, and the processor is used to execute the computer instructions stored in the memory. When the computer instructions are executed by the processor, the electronic device implements the method according to any one of claims 1 to 7. XGBoost The steps of the method for identifying abnormal power consumption of a low-voltage charging pile based on an algorithm.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 7. XGBoost The steps of the method for identifying abnormal power consumption of a low-voltage charging pile based on an algorithm.

Images