Digital employee-based power supply station business automation test method and system

Abstract

The invention relates to the technical field of power supply station power business, in particular to a power supply station business automatic test method and system based on digital employees, and the method comprises the steps: carrying out the feature extraction of single-phase meter measurement data of a to-be-tested business region in a power supply station, and obtaining measurement misalignment feature data; determining single-phase meter metering misalignment grade distribution data according to the single-phase meter metering data deviation amplitude and the metering misalignment characteristic data; establishing a correlation model by taking the single-phase meter measurement misalignment grade distribution data as an independent variable and the key assessment index value of each transformer area as a dependent variable; in combination with the correlation model, predicting the change trend of the key assessment indexes of the test points under different misalignment levels by using a long-short-term memory neural network to obtain expected index data; and generating a service test result according to the expected index data and the actual assessment index data. According to the invention, the digital employee technology and the machine learning algorithm are combined, the accuracy and reliability of the power supply station business automation test are improved, and manual intervention and errors are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power business of power supply stations, and in particular to a power supply station business automation testing method and system based on digital employees. Background Art

[0002] In the modern power system, power supply stations play an important role in delivering electricity from power plants to end users. With the development of artificial intelligence technology, in the wave of digital transformation, digital employees based on artificial intelligence technology are gradually penetrating into power supply stations. With their powerful data processing capabilities and intelligent decision-making support, digital employees have brought unprecedented opportunities for the automation of power supply station business. However, in practical applications, especially for the problems of single-phase meter inaccuracy and deviation of substation assessment indicators, the testing methods of digital employees still face many technical bottlenecks and challenges. As the core equipment of power supply station business, the accuracy and integrity of single-phase meter measurement data are directly related to the evaluation of substation assessment indicators. However, due to various reasons such as equipment aging and environmental factors, single-phase meters often have inaccurate measurement problems, which brings great challenges to the accuracy and fairness of substation assessment indicators. How to accurately test and identify the inaccurate measurement of single-phase meters is a major problem that digital employees need to solve in the current power supply station business testing.

[0003] Secondly, the analysis of the changing rules of assessment indicators under different inaccuracy levels needs to consider multiple factors, such as load type, power consumption time distribution, line parameters, etc. There are complex coupling relationships between these factors, which brings great difficulties to the testing process. For example, different types of loads have different degrees of influence on meter inaccuracy, and the uneven distribution of power consumption time may aggravate this influence. In addition, the impact of single-phase meter measurement inaccuracy on the assessment indicators of the substation has a time accumulation effect, which may be difficult to detect in the short term, but after long-term accumulation, it may cause the assessment indicators to deviate seriously from reality. How to quantify this time accumulation effect is a major problem. In actual operation, due to the lack of effective quantitative methods, it is often difficult to accurately evaluate the long-term impact of meter inaccuracy on assessment indicators, which brings great uncertainty to the business management of power supply stations. Therefore, how to solve the above problems and improve the accuracy and reliability of digital employees in power supply station business testing is an issue that urgently needs to be studied and solved in the current power industry. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for automatic testing of power supply station services based on digital employees.

[0005] In a first aspect, the present invention provides a method for automated testing of power supply services based on digital employees, the method comprising the following steps: The digital employee technology is used to trigger the collection of single-phase meter measurement data of the business area to be tested in the power supply station, and feature extraction is performed on the single-phase meter measurement data to obtain measurement inaccuracy feature data; the single-phase meter measurement data includes power data, voltage data and current data; The single-phase meter measurement data is analyzed by using a support vector machine to calculate the key assessment index values ​​of each substation; the key assessment index values ​​include the line loss rate and the power difference rate of each substation; Determine the single-phase meter measurement inaccuracy grade distribution data based on the single-phase meter measurement data deviation amplitude and measurement inaccuracy characteristic data; The distribution data of the inaccuracy level of single-phase meter is used as the independent variable, and the key assessment index value of each substation is used as the dependent variable, and the correlation model between the inaccuracy level and the key assessment index of the substation is established; Combined with the association model, a pre-built long short-term memory neural network is used to predict the change trend of key assessment indicators of test points under different misalignment levels to obtain expected indicator data; Generate business test results based on the expected indicator data and actual assessment indicator data.

[0006] In a further embodiment, the step of extracting features from the single-phase meter measurement data to obtain measurement misalignment feature data comprises: A five-point linear interpolation method is used to fill in the breakpoints and missing points of the single-phase meter measurement data to obtain the single-phase meter measurement completion data; The single-phase meter metering completion data is filtered through a seven-point sliding median filter window to obtain single-phase meter metering filtered data; Using Daubechies fourth-order wavelet to perform three-layer decomposition on the single-phase meter measurement filter data to obtain high-frequency component data, and extracting amplitude mutation points from the high-frequency component data as feature points; Constructing a statistical period data window, and using the statistical period data window to extract a voltage feature group, a current feature group, and a power feature group at the feature point under a preset reference value; Performing maximum and minimum value normalization processing on the voltage feature group, the current feature group, and the power feature group respectively to obtain corresponding standardized feature data; According to all the standardized characteristic data under the voltage characteristic dimension, the current characteristic dimension and the power characteristic dimension, a characteristic weight coefficient of each standardized characteristic data is calculated by using a weighted average method; The standardized feature data whose feature weight coefficient exceeds a preset misalignment threshold is used as measurement misalignment feature data.

[0007] In a further embodiment, the voltage characteristic group includes voltage root mean square value, voltage peak-to-average ratio and voltage fluctuation rate; The current characteristic group includes a current peak factor and a current distortion degree; The power characteristic group includes power fluctuation and phase offset.

[0008] In a further implementation scheme, the step of analyzing the single-phase meter measurement data using a support vector machine to calculate the key assessment index values ​​of each substation includes: Extracting key characteristic points from the single-phase meter measurement data using a principal component analysis algorithm, and collecting historical load data of the key characteristic points; Using a support vector regression algorithm to analyze the historical load data and identify the load distribution pattern of key characteristic points; According to the load distribution pattern of the key characteristic points, the mean load value of the same period is calculated by using a statistical analysis method; Calculate the difference between the current power value of the key feature point and the average load value of the same period, and calculate the power difference rate of the substation; Determine a line loss rate reference value, and calculate a transformer loss compensation value, a conductor loss compensation value, and a contact loss compensation value based on the single-phase meter measurement data; The line loss rate of the substation is determined according to the difference between the line loss rate baseline value and the transformer loss compensation value, the conductor loss compensation value, and the contact loss compensation value, and the line loss rate of the substation and the power difference rate of the substation are integrated to form the key assessment indicator value of the substation.

[0009] In a further embodiment, the step of determining the single-phase meter measurement mismatch grade distribution data according to the single-phase meter measurement data deviation amplitude and measurement mismatch characteristic data comprises: According to the measurement inaccuracy characteristic data and the single-phase meter measurement standard value, the single-phase meter measurement data deviation amplitude is obtained, and the single-phase meter measurement data deviation amplitude includes the voltage deviation amplitude, the current deviation amplitude and the power factor deviation amplitude of the single-phase meter measurement data; Determine the moving data window, and perform statistical analysis on the measurement data deviation amplitude of the single-phase meter within the moving data window to obtain the measurement deviation amplitude characteristics; The measurement deviation amplitude feature is modeled by using a Gaussian naive Bayes classifier, and the characteristic distribution probability value of each measurement deviation amplitude feature at different misalignment levels is calculated; The metering inaccuracy characteristic data is divided into inaccuracy levels according to the characteristic distribution probability value and a preset characteristic inaccuracy level probability distribution interval, and the single-phase meter metering inaccuracy level distribution data is determined.

[0010] In a further embodiment, the metering deviation amplitude characteristics include the deviation mean, deviation standard deviation and maximum deviation value of the deviation amplitude of the single-phase meter metering data.

[0011] In a further implementation scheme, the step of establishing a correlation model between the degree of inaccuracy and the key assessment indicators of the substations with the single-phase meter measurement inaccuracy level distribution data as the independent variable and the key assessment indicator values ​​of each substation as the dependent variable includes: Normalizing the single-phase meter measurement inaccuracy level distribution data to obtain normalized inaccuracy level distribution data; A random forest regressor is used to model the relationship between the normalized misalignment level distribution data and the key assessment indicator values ​​of each substation, so as to obtain a correlation model between the misalignment degree and the key assessment indicators of the substation.

[0012] In a further embodiment, the step of combining the association model and using a pre-built long short-term memory neural network to predict the change trend of key assessment indicators of test points under different misalignment levels to obtain expected indicator data includes: Performing linear interpolation on the probability distribution interval of the feature misalignment level to obtain a test data point; Calculating the confidence interval value of each of the test data points, and detecting the validity of the test data points according to the confidence interval value, and screening out valid test data points; Acquire the single-phase meter measurement test data of the valid test data point, and obtain the corresponding key assessment indicators of the test point through the association model; The key assessment indicators of the test points are input into the pre-built long short-term memory neural network to predict the change trends of the key assessment indicators of the test points under different inaccuracy levels to obtain the expected indicator data.

[0013] In a further implementation scheme, the step of generating a business test result according to the expected indicator data and the actual assessment indicator data includes: The expected indicator data is compared with the actual assessment indicator data. If the comparison results of the expected indicator data and the actual assessment indicator data are consistent, the business test result is determined to be a test pass; if the comparison results of the expected indicator data and the actual assessment indicator data are inconsistent, the business test result is determined to be a test failure.

[0014] In a second aspect, the present invention provides a power supply station service automation testing system based on digital employees, the system comprising: A data acquisition module is used to trigger the acquisition of single-phase meter measurement data of the business area to be tested in the power supply station by using digital employee technology, and to extract features of the single-phase meter measurement data to obtain measurement inaccuracy feature data; the single-phase meter measurement data includes power data, voltage data and current data; A data analysis module is used to analyze the single-phase meter measurement data using a support vector machine to calculate the key assessment index values ​​of each substation; the key assessment index values ​​include the line loss rate and the power difference rate of each substation; The inaccuracy analysis module is used to determine the single-phase meter inaccuracy grade distribution data according to the single-phase meter metering data deviation amplitude and metering inaccuracy characteristic data; A model building module is used to establish a correlation model between the degree of inaccuracy and the key assessment indicators of the substation using the distribution data of the inaccuracy level of the single-phase meter as the independent variable and the key assessment indicator values ​​of each substation as the dependent variable; An expected test module is used to combine the association model and use a pre-built long short-term memory neural network to predict the change trend of key assessment indicators of test points under different misalignment levels to obtain expected indicator data; The test result generation module is used to generate business test results based on the expected indicator data and the actual assessment indicator data.

[0015] The present invention provides a method and system for automated testing of power supply station services based on digital employees. The method uses digital employee technology to trigger the collection of single-phase meter metering data of a service area to be tested in a power supply station, and extracts features from the single-phase meter metering data to obtain metering inaccuracy feature data; uses a support vector machine to analyze the single-phase meter metering data to calculate key assessment index values ​​for each substation; the key assessment index values ​​include line loss rate and electricity quantity difference rate for each substation; determines single-phase meter metering inaccuracy grade distribution data according to the single-phase meter metering data deviation amplitude and metering inaccuracy feature data; uses the single-phase meter metering inaccuracy grade distribution data as an independent variable and the key assessment index value of each substation as a dependent variable, and establishes a correlation model between the degree of inaccuracy and the key assessment index of the substation; combines the correlation model, uses a pre-constructed long short-term memory neural network to predict the change trend of the key assessment index of the test point under different inaccuracy levels, and obtains expected index data; generates a business test result according to the expected index data and the actual assessment index data. Compared with the existing technology, this method can effectively identify the problem of single-phase meter inaccuracy and predict the impact of inaccuracy on the assessment indicators of the substation, thereby realizing the automated management of the power supply station business and improving the work efficiency and management level of the power supply station. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of a flow chart of a power supply station service automation testing method based on digital employees provided by an embodiment of the present invention; Figure 2 The invention discloses a block diagram of a power supply station service automation test system based on digital employees provided in an embodiment of the invention. DETAILED DESCRIPTION

[0017] The following specifically illustrates the implementation mode of the present invention in conjunction with the accompanying drawings. The embodiments are provided for illustrative purposes only and cannot be understood as limiting the present invention. The accompanying drawings are provided for reference and illustration only and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from the spirit and scope of the present invention.

[0018] refer to Figure 1 The embodiment of the present invention provides a method for automated testing of power supply services based on digital employees, such as Figure 1 As shown, the method comprises the following steps: S1. Use digital employee technology to trigger the collection of single-phase meter measurement data in the business area to be tested in the power supply station, and perform feature extraction on the single-phase meter measurement data to obtain measurement inaccuracy feature data; the single-phase meter measurement data includes power data, voltage data and current data.

[0019] In this embodiment, the step of extracting features from the single-phase meter measurement data to obtain measurement inaccuracy feature data includes: A five-point linear interpolation method is used to fill in the breakpoints and missing points of the single-phase meter measurement data to obtain the single-phase meter measurement completion data; The single-phase meter metering completion data is filtered through a seven-point sliding median filter window to obtain single-phase meter metering filtered data; Using Daubechies fourth-order wavelet to perform three-layer decomposition on the single-phase meter measurement filter data to obtain high-frequency component data, and extracting amplitude mutation points from the high-frequency component data as feature points; Constructing a statistical period data window, and using the statistical period data window to extract a voltage feature group, a current feature group, and a power feature group at the feature point under a preset reference value; Performing maximum and minimum value normalization processing on the voltage feature group, the current feature group and the power feature group respectively to obtain corresponding standardized feature data; the standardized feature data includes standardized voltage feature data, standardized current feature data and standardized power feature data; According to all the standardized characteristic data under the voltage characteristic dimension, the current characteristic dimension and the power characteristic dimension, a characteristic weight coefficient of each standardized characteristic data is calculated by using a weighted average method; The standardized feature data whose feature weight coefficient exceeds a preset misalignment threshold is used as measurement misalignment feature data.

[0020] Specifically, the business data of the power supply station covers the metering data of various types of electric energy meters, including but not limited to single-phase meters, three-phase meters, etc. These data are important components of the power management of the power supply station, involving the management, verification, rotation, and prevention and handling of fault errors of electric energy metering devices. The metering data of single-phase meters can be used to test and evaluate multiple businesses of the power supply station. Digital employees, as executors or tools of automated testing, instruct or trigger the data collection process of data acquisition equipment through their program logic. This embodiment uses digital employee technology to automatically trigger the data collection system to collect single-phase meter metering data involved in the business area to be tested in the power supply station. The data contains key information such as voltage data, current data, and power data. For the breakpoint and missing point data in the single-phase meter metering time series data, this embodiment uses a five-point linear interpolation method to fill the breakpoint and missing point data in the single-phase meter metering data. For example, two valid data points before and after the breakpoint are selected, for a total of five points, and the completion value at the breakpoint is calculated by a linear interpolation formula. Thus, the single-phase meter metering completion data is obtained. For the completed single-phase meter metering data, this embodiment filters the single-phase meter metering completion data through a seven-point sliding median filtering window, that is, a window containing seven data points is slid on the data sequence, the data in each window is sorted by amplitude, and the value in the middle position is taken as the filtering result of the window. At the same time, in the filtering result, the mean and standard deviation of the voltage, current and power factor are calculated, and the values ​​exceeding the range of three times the standard deviation are regarded as abnormal values ​​and eliminated to ensure the accuracy of the data, so as to remove noise and abnormal values, thereby obtaining the single-phase meter metering filtered data. Then, this embodiment uses Daubechies fourth-order wavelet (db4 wavelet for short) to perform three-layer decomposition on the single-phase meter metering filtered data to obtain high-frequency component data, and extracts amplitude mutation points from the high-frequency component data. These mutation points are feature points, and based on the feature points, a statistical period data window is constructed. For example, with a period of 10 minutes, the data in each period is counted to provide a basis for subsequent feature calculation.

[0021] According to the characteristic points and the statistical period data window, under the preset reference values ​​(such as rated voltage, rated current, power factor, etc.), the voltage characteristic group, current characteristic group and power characteristic group are extracted respectively. These characteristic groups contain characteristic values ​​of their respective fields, such as voltage fluctuation rate, phase offset, peak factor, etc. Specifically, in this embodiment, the voltage root mean square value, voltage peak-to-average ratio and voltage fluctuation rate are used as the voltage characteristic group; the current peak factor and current distortion are used as the current characteristic group; the power fluctuation and phase offset are used as the power characteristic group; then the voltage characteristic group, the current characteristic group and the power characteristic group are normalized to the maximum and minimum values ​​respectively, and the characteristic values ​​are mapped to the interval from 0 to 1 to obtain the corresponding The standardized voltage characteristic data, the standardized current characteristic data and the standardized power characteristic data are obtained. According to all the standardized characteristic data (i.e., the standardized voltage characteristic data, the standardized current characteristic data and the standardized power characteristic data), the characteristic weight coefficient of each standardized characteristic data is calculated by adopting the weighted average method according to the voltage characteristic dimension, the current characteristic dimension and the power characteristic dimension. The characteristic weight coefficient is calculated to evaluate the importance of each characteristic point in the overall dimensional characteristics. The characteristic weight coefficient exceeding the preset threshold value indicates that the characteristic point has a significant influence in the misalignment feature identification, so that the standardized characteristic data whose characteristic weight coefficient exceeds the preset misalignment threshold value is marked as measurement misalignment characteristic data.

[0022] S2. Use support vector machine to analyze the single-phase meter measurement data and calculate the key assessment indicator values ​​of each substation.

[0023] In this embodiment, the step of using a support vector machine to analyze the single-phase meter measurement data and calculate the key assessment index value of each substation includes: Extracting key characteristic points from the single-phase meter measurement data using a principal component analysis algorithm, and collecting historical load data of the key characteristic points; Using a support vector regression algorithm to analyze the historical load data and identify the load distribution pattern of key characteristic points; According to the load distribution pattern of the key characteristic points, the mean load value of the same period is calculated by using a statistical analysis method; Calculate the difference between the current power value of the key feature point and the average load value of the same period, and calculate the power difference rate of the substation; Determine a line loss rate reference value, and calculate a transformer loss compensation value, a conductor loss compensation value, and a contact loss compensation value based on the single-phase meter measurement data; The line loss rate of the substation is determined according to the difference between the line loss rate baseline value and the transformer loss compensation value, the conductor loss compensation value, and the contact loss compensation value, and the line loss rate of the substation and the power difference rate of the substation are integrated to form the key assessment indicator value of the substation.

[0024] Specifically, this embodiment uses a principal component analysis algorithm to extract key feature points from the single-phase meter measurement data, collects historical load data of the key feature points, and uses a support vector regression (SVR) algorithm to analyze these data. The support vector regression algorithm, as a supervised learning algorithm, can find the optimal hyperplane so that the data points are within the boundary of the hyperplane, thereby mining and identifying the law of load changes and accurately identifying the load distribution pattern. According to the load distribution pattern of the identified key feature points, a statistical analysis method (such as mean calculation) is used to predict the current load level and obtain the mean load of the same period. This mean is used as a benchmark for subsequent comparisons and can be used to identify load fluctuations.

[0025] In this embodiment, the key evaluation index values ​​include the line loss rate and the electricity difference rate of each substation. These indicators are the key to evaluating the business performance of the power supply station. This embodiment calculates the degree of deviation between the current electricity value of the key feature point and the average load value of the same period, and obtains the substation electricity difference rate based on this to reflect the change in the substation electricity. For the substation line loss rate, this embodiment determines the line loss rate baseline value according to the power loss level of the substation under ideal conditions, and calculates the actual line loss rate of the substation based on the line loss rate baseline value, combined with various loss factors such as transformer loss compensation value, conductor loss compensation value and contact loss compensation value, so as to more accurately evaluate the power loss of the substation. The calculation of these compensation values ​​is based on the physical characteristics of the transformer and the power system, such as no-load loss, load loss, etc. Among them, the calculation of the transformer loss compensation value In the calculation process, this embodiment obtains the input power and output power of the transformer according to the single-phase meter measurement data, calculates the load loss according to the ratio of the difference between the input power and output power of the transformer and the transformer efficiency, and obtains the transformer loss compensation value according to the no-load loss value and load loss of the transformer; for the calculation process of the wire loss compensation value, this embodiment obtains the current value flowing through the wire according to the single-phase meter measurement data, and calculates the wire loss according to the current value flowing through the wire. When calculating the line loss rate of the substation, the average value of the wire loss is used as the wire loss compensation value; for the contact loss compensation value, this embodiment obtains the current value flowing through the contact point according to the single-phase meter measurement data, and calculates the contact loss according to the current value flowing through the contact point and the contact resistance. When calculating the line loss rate of the substation, the average value of the contact loss is used as the contact loss compensation value.

[0026] S3. Determine the single-phase meter measurement inaccuracy grade distribution data based on the single-phase meter measurement data deviation amplitude and measurement inaccuracy characteristic data.

[0027] In this embodiment, the step of determining the single-phase meter measurement inaccuracy level distribution data according to the single-phase meter measurement data deviation amplitude and measurement inaccuracy characteristic data includes: According to the measurement inaccuracy characteristic data and the single-phase meter measurement standard value, the single-phase meter measurement data deviation amplitude is obtained, and the single-phase meter measurement data deviation amplitude includes the voltage deviation amplitude, the current deviation amplitude and the power factor deviation amplitude of the single-phase meter measurement data; Determine a mobile data window, and perform statistical analysis on the deviation amplitude of the single-phase meter measurement data within the mobile data window to obtain a measurement deviation amplitude feature; the measurement deviation amplitude feature includes a deviation mean, a deviation standard deviation, and a maximum deviation value of the deviation amplitude of the single-phase meter measurement data; The measurement deviation amplitude feature is modeled by using a Gaussian naive Bayes classifier, and the characteristic distribution probability value of each measurement deviation amplitude feature at different misalignment levels is calculated; The metering inaccuracy characteristic data is divided into inaccuracy levels according to the characteristic distribution probability value and a preset characteristic inaccuracy level probability distribution interval, and the single-phase meter metering inaccuracy level distribution data is determined.

[0028] Specifically, this embodiment calculates the voltage deviation amplitude, current deviation amplitude and power factor deviation amplitude of the single-phase meter metering data based on the metering inaccuracy characteristic data and the single-phase meter metering standard value. For example, for each metering parameter (voltage, current, power factor), the deviation amplitude can be represented by the difference between the actual measured value and the standard value. Then, this embodiment determines an appropriate mobile data window size. The mobile data window should contain enough data points to reflect the dynamic changes of the single-phase meter metering data. The deviation amplitude of the single-phase meter metering data in the mobile data window is statistically analyzed to calculate the metering deviation amplitude characteristics such as the deviation mean, deviation standard deviation and maximum deviation value. These metering deviation amplitude characteristics can reflect the overall situation and discreteness of the deviation amplitude in the data window.

[0029] This embodiment uses the metering deviation amplitude characteristics and the corresponding misalignment level information in the historical data to train a Gaussian naive Bayes classifier. After the training is completed, the metering deviation amplitude characteristics (deviation mean, deviation standard deviation, maximum deviation value) calculated above are used as input features of the Gaussian naive Bayes classifier to calculate the feature distribution probability value of each metering deviation amplitude characteristic at different misalignment levels. It should be ensured that the feature distribution probability value is within a preset error range. These feature distribution probability values ​​can reflect the distribution of features at different levels and provide a basis for subsequent level division. According to the feature distribution probability value output by the classifier, different feature misalignment level probability distribution intervals are set to divide the single-phase meter metering misalignment level. For example, the misalignment level includes normal, mild misalignment, moderate misalignment, and severe misalignment. If the probability of the feature at the normal level exceeds 95%, it is determined to be normal, thereby obtaining the single-phase meter metering misalignment level distribution data.

[0030] S4. Taking the distribution data of the inaccuracy level of single-phase meters as the independent variable and the key assessment index values ​​of each substation as the dependent variable, a correlation model between the inaccuracy level and the key assessment index of the substation is established.

[0031] In this embodiment, the single-phase meter measurement inaccuracy level distribution data is used as the independent variable, and the key assessment index value of each substation is used as the dependent variable. The steps of establishing a correlation model between the inaccuracy level and the key assessment index of the substation include: Normalizing the single-phase meter measurement inaccuracy level distribution data to obtain normalized inaccuracy level distribution data; A random forest regressor is used to model the relationship between the normalized misalignment level distribution data and the key assessment indicator values ​​of each substation, so as to obtain a correlation model between the misalignment degree and the key assessment indicators of the substation.

[0032] Specifically, this embodiment normalizes the metering inaccuracy level distribution data of single-phase meters, that is, maps the numerical values ​​of each level to the interval [0, 1] to eliminate the impact of dimensions between different levels, which is convenient for subsequent modeling and analysis. At the same time, the key assessment indicator values ​​of the substation (substation line loss rate and substation electricity difference rate) are preprocessed to ensure their consistency with the normalized inaccuracy level distribution data in time and space dimensions. Then, the normalized inaccuracy level distribution data is used as input features, and the preprocessed substation assessment indicator values ​​are used as target variables. The random forest regressor is used to learn and model the potential relationship between the normalized inaccuracy level distribution data and the key assessment indicator values ​​of the substation, capture the impact of different metering error levels on the substation assessment indicators, and finally obtain the association model of the relationship between the inaccuracy degree and the assessment indicator, providing a scientific basis for the intelligent management and decision-making of the power supply station.

[0033] S5. Combined with the association model, the pre-built long short-term memory neural network is used to predict the changing trend of key assessment indicators of test points under different inaccuracy levels to obtain expected indicator data.

[0034] In this embodiment, the step of combining the association model and using the pre-built long short-term memory neural network to predict the change trend of the key assessment indicators of the test points under different misalignment levels to obtain the expected indicator data includes: Performing linear interpolation on the probability distribution interval of the feature misalignment level to obtain a test data point; Calculating the confidence interval value of each of the test data points, and detecting the validity of the test data points according to the confidence interval value, and screening out valid test data points; Acquire the single-phase meter measurement test data of the valid test data point, and obtain the corresponding key assessment indicators of the test point through the association model; The key assessment indicators of the test points are input into the pre-built long short-term memory neural network to predict the change trends of the key assessment indicators of the test points under different inaccuracy levels to obtain the expected indicator data.

[0035] Specifically, this embodiment performs linear interpolation on the probability distribution interval of the feature misalignment level (i.e., the misalignment interval) to supplement the data, so as to generate more test data points within the misalignment interval, provide a more complete data set for the model, and thus more accurately evaluate the performance of the association model. This embodiment uses linear interpolation technology to generate a series of test data points within the probability distribution interval of the feature misalignment level, and calculates the confidence interval value of each test data point, and judges the validity of the test data point by the confidence interval value, so as to screen out valid test data points and ensure the accuracy of subsequent analysis. For example, according to a preset confidence level (e.g., a 95% confidence level), it is determined whether the test data point falls within an acceptable confidence interval, and the data points falling within the confidence interval are deemed valid, otherwise they are deemed invalid. In order to be effective, and to convert the test data into specific assessment index values ​​for the convenience of subsequent analysis and prediction, this embodiment calculates the corresponding key assessment indicators of the test points through the association model based on the single-phase meter measurement test data of the screened effective test data points, and inputs the calculated key assessment indicators of the test points into the long short-term memory neural network (LSTM). The network includes input layer nodes, hidden layer nodes, and output layer nodes. The LSTM network is used to predict the change trend of the key assessment indicators of the test points under different inaccuracy levels to obtain the expected indicator data, so as to evaluate the accuracy and effectiveness of the test by comparing the expected indicator data and the actual assessment indicator data output by the LSTM network, generate business test results, and provide strong support for the business automation test of the power supply station.

[0036] S6. Generate business test results based on the expected indicator data and actual assessment indicator data.

[0037] In this embodiment, the step of generating a business test result according to the expected indicator data and the actual assessment indicator data includes: The expected indicator data is compared with the actual assessment indicator data. If the comparison results of the expected indicator data and the actual assessment indicator data are consistent, the business test result is determined to be a test pass; if the comparison results of the expected indicator data and the actual assessment indicator data are inconsistent, the business test result is determined to be a test failure.

[0038] The embodiment of the present invention provides a method for automated testing of power supply station services based on digital employees. The method uses digital employee technology to trigger the collection of single-phase meter metering data of the business area to be tested in the power supply station, and extracts features of the single-phase meter metering data to obtain metering inaccuracy feature data; uses a support vector machine to analyze the single-phase meter metering data to calculate the key assessment index values ​​of each substation; determines the single-phase meter metering inaccuracy level distribution data according to the deviation amplitude of the single-phase meter metering data and the metering inaccuracy feature data; takes the single-phase meter metering inaccuracy level distribution data as the independent variable and the key assessment index values ​​of each substation as the dependent variable, and establishes a correlation model between the degree of inaccuracy and the key assessment index of the substation; combines the correlation model, uses a pre-constructed long short-term memory neural network to predict the change trend of the key assessment index of the test point under different inaccuracy levels, and obtains the expected index data; generates business test results according to the expected index data and the actual assessment index data. Compared with the prior art, the automated testing method proposed in this embodiment uses digital employee technology and machine learning algorithms to accurately identify metering inaccuracy problems and quantitatively analyze the deviation of substation assessment indicators, thereby improving the accuracy and reliability of power supply station business testing and reducing the time and cost of manual testing.

[0039] It should be noted that the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0040] In one embodiment, Figure 2 As shown, an embodiment of the present invention provides a power supply station service automation testing system based on digital employees, and the system includes: The data acquisition module 101 is used to trigger the acquisition of single-phase meter measurement data of the business area to be tested in the power supply station by using digital employee technology, and perform feature extraction on the single-phase meter measurement data to obtain measurement inaccuracy feature data; the single-phase meter measurement data includes power data, voltage data and current data; The data analysis module 102 is used to analyze the single-phase meter measurement data using a support vector machine to calculate the key assessment index values ​​of each substation; the key assessment index values ​​include the line loss rate and the power difference rate of each substation; The misalignment analysis module 103 is used to determine the misalignment level distribution data of the single-phase meter according to the measurement data deviation amplitude and measurement misalignment characteristic data of the single-phase meter; Model building module 104, used to establish a correlation model between the inaccuracy level and the key assessment index of the substation using the single-phase meter measurement inaccuracy level distribution data as an independent variable and the key assessment index value of each substation as a dependent variable; The expected test module 105 is used to combine the association model and use the pre-built long short-term memory neural network to predict the change trend of the key assessment indicators of the test points under different misalignment levels to obtain the expected indicator data; The test result generating module 106 is used to generate a business test result according to the expected indicator data and the actual assessment indicator data.

[0041] For the specific definition of a power supply station business automation test system based on digital employees, please refer to the above-mentioned definition of a power supply station business automation test method based on digital employees, which will not be repeated here. A person of ordinary skill in the art will appreciate that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0042] The embodiment of the present invention provides a power supply station service automation test system based on digital employees. The system uses the digital employee technology to trigger the collection of single-phase meter metering data of the service area to be tested in the power supply station through a data acquisition module, and extracts features from the single-phase meter metering data to obtain metering inaccuracy feature data; the data analysis module uses a support vector machine to analyze the single-phase meter metering data to calculate the key assessment index values ​​of each substation; the inaccuracy analysis module determines the single-phase meter metering inaccuracy level distribution data according to the single-phase meter metering data deviation amplitude and metering inaccuracy feature data; the model construction module uses the single-phase meter metering inaccuracy level distribution data as an independent variable and the key assessment index value of each substation as a dependent variable to establish a correlation model between the inaccuracy degree and the key assessment index of the substation; the expected test module combines the association model and uses a pre-constructed long short-term memory neural network to predict the change trend of the key assessment index of the test point under different inaccuracy levels to obtain the expected index data; the test result generation module generates a business test result according to the expected index data and the actual assessment index data. Compared with the existing technology, the automated testing system proposed in this embodiment uses digital employee technology and machine learning algorithms to accurately identify metering inaccuracies and quantitatively analyze deviations in substation assessment indicators, thereby improving the accuracy and reliability of power supply station business testing and reducing the time and cost of manual testing.

[0043] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. A power supply station service automation testing method based on digital employees, characterized in that: The following steps are involved: The digital employee technology is used to trigger the collection of single-phase meter measurement data of the business area to be tested in the power supply station, and feature extraction is performed on the single-phase meter measurement data to obtain measurement inaccuracy feature data; The single-phase meter measurement data includes power data, voltage data and current data; The single-phase meter measurement data is analyzed by using a support vector machine to calculate the key assessment index values ​​of each substation; the key assessment index values ​​include the line loss rate and the power difference rate of each substation; Determine the single-phase meter measurement inaccuracy grade distribution data based on the single-phase meter measurement data deviation amplitude and measurement inaccuracy characteristic data; The distribution data of the inaccuracy level of single-phase meter is used as the independent variable, and the key assessment index value of each substation is used as the dependent variable, and the correlation model between the inaccuracy level and the key assessment index of the substation is established; Combined with the association model, a pre-built long short-term memory neural network is used to predict the change trend of key assessment indicators of test points under different misalignment levels to obtain expected indicator data; Generate business test results based on the expected indicator data and actual assessment indicator data.

2. A method for automated testing of power supply services based on digital employees as claimed in claim 1, characterized in that: The step of extracting features from the single-phase meter measurement data to obtain measurement inaccuracy feature data comprises: A five-point linear interpolation method is used to fill in the breakpoints and missing points of the single-phase meter measurement data to obtain the single-phase meter measurement completion data; The single-phase meter metering completion data is filtered through a seven-point sliding median filter window to obtain single-phase meter metering filtered data; Using Daubechies fourth-order wavelet to perform three-layer decomposition on the single-phase meter measurement filter data to obtain high-frequency component data, and extracting amplitude mutation points from the high-frequency component data as feature points; Constructing a statistical period data window, and using the statistical period data window to extract a voltage feature group, a current feature group, and a power feature group at the feature point under a preset reference value; Performing maximum and minimum value normalization processing on the voltage feature group, the current feature group, and the power feature group respectively to obtain corresponding standardized feature data; According to all the standardized characteristic data under the voltage characteristic dimension, the current characteristic dimension and the power characteristic dimension, a characteristic weight coefficient of each standardized characteristic data is calculated by using a weighted average method; The standardized feature data whose feature weight coefficient exceeds a preset misalignment threshold is used as measurement misalignment feature data.

3. The method for automated testing of power supply station services based on digital employees as claimed in claim 2, characterized in that: The voltage characteristic group includes voltage root mean square value, voltage peak-to-average ratio and voltage fluctuation rate; The current characteristic group includes a current peak factor and a current distortion degree; The power characteristic group includes power fluctuation and phase offset.

4. The method for automated testing of power supply station services based on digital employees as claimed in claim 1, characterized in that: The step of using a support vector machine to analyze the single-phase meter measurement data and calculate the key assessment index values ​​of each substation includes: Extracting key characteristic points from the single-phase meter measurement data using a principal component analysis algorithm, and collecting historical load data of the key characteristic points; Using a support vector regression algorithm to analyze the historical load data and identify the load distribution pattern of key characteristic points; According to the load distribution pattern of the key characteristic points, the mean load value of the same period is calculated by using a statistical analysis method; Calculate the difference between the current power value of the key feature point and the average load value of the same period, and calculate the power difference rate of the substation; Determine a line loss rate reference value, and calculate a transformer loss compensation value, a conductor loss compensation value, and a contact loss compensation value based on the single-phase meter measurement data; The line loss rate of the substation is determined based on the difference between the line loss rate baseline value and the transformer loss compensation value, the conductor loss compensation value, and the contact loss compensation value, and the line loss rate of the substation and the power difference rate of the substation are integrated to form the key assessment indicator value of the substation.

5. The method for automated testing of power supply station services based on digital employees as claimed in claim 1, characterized in that: The step of determining the single-phase meter measurement inaccuracy grade distribution data according to the single-phase meter measurement data deviation amplitude and measurement inaccuracy characteristic data comprises: According to the measurement inaccuracy characteristic data and the single-phase meter measurement standard value, the single-phase meter measurement data deviation amplitude is obtained, and the single-phase meter measurement data deviation amplitude includes the voltage deviation amplitude, the current deviation amplitude and the power factor deviation amplitude of the single-phase meter measurement data; Determine the moving data window, and perform statistical analysis on the measurement data deviation amplitude of the single-phase meter within the moving data window to obtain the measurement deviation amplitude characteristics; The measurement deviation amplitude feature is modeled by using a Gaussian naive Bayes classifier, and the characteristic distribution probability value of each measurement deviation amplitude feature at different misalignment levels is calculated; The metering inaccuracy characteristic data is divided into inaccuracy levels according to the characteristic distribution probability value and a preset characteristic inaccuracy level probability distribution interval, and the single-phase meter metering inaccuracy level distribution data is determined.

6. The method for automated testing of power supply station services based on digital employees as claimed in claim 5, characterized in that: The metering deviation amplitude characteristics include the deviation mean, deviation standard deviation and maximum deviation value of the deviation amplitude of the single-phase meter metering data.

7. The method for automated testing of power supply station services based on digital employees as claimed in claim 1, characterized in that: The steps of establishing a correlation model between the inaccuracy level distribution data of the single-phase meter as the independent variable and the key assessment index value of each substation as the dependent variable include: Normalizing the single-phase meter measurement inaccuracy level distribution data to obtain normalized inaccuracy level distribution data; A random forest regressor is used to model the relationship between the normalized misalignment level distribution data and the key assessment indicator values ​​of each substation, so as to obtain a correlation model between the misalignment degree and the key assessment indicators of the substation.

8. The method for automated testing of power supply station services based on digital employees as claimed in claim 5, characterized in that: The step of combining the association model and using the pre-built long short-term memory neural network to predict the change trend of the key assessment indicators of the test points under different misalignment levels to obtain the expected indicator data includes: Performing linear interpolation on the probability distribution interval of the feature misalignment level to obtain a test data point; Calculating the confidence interval value of each of the test data points, and detecting the validity of the test data points according to the confidence interval value, and screening out valid test data points; Acquire the single-phase meter measurement test data of the valid test data point, and obtain the corresponding key assessment indicators of the test point through the association model; The key assessment indicators of the test points are input into the pre-built long short-term memory neural network to predict the change trends of the key assessment indicators of the test points under different inaccuracy levels to obtain the expected indicator data.

9. The method for automated testing of power supply station services based on digital employees as claimed in claim 1, characterized in that: The step of generating a business test result according to the expected indicator data and the actual assessment indicator data includes: The expected indicator data is compared with the actual assessment indicator data. If the comparison results of the expected indicator data and the actual assessment indicator data are consistent, the business test result is determined to be a test pass; if the comparison results of the expected indicator data and the actual assessment indicator data are inconsistent, the business test result is determined to be a test failure.

10. A power supply station business automation test system based on digital employees, characterized in that: The system comprises: A data acquisition module is used to trigger the acquisition of single-phase meter measurement data of the business area to be tested in the power supply station by using digital employee technology, and to extract features of the single-phase meter measurement data to obtain measurement inaccuracy feature data; the single-phase meter measurement data includes power data, voltage data and current data; A data analysis module is used to analyze the single-phase meter measurement data using a support vector machine to calculate the key assessment index values ​​of each substation; the key assessment index values ​​include the line loss rate and the power difference rate of each substation; The inaccuracy analysis module is used to determine the single-phase meter inaccuracy grade distribution data according to the single-phase meter metering data deviation amplitude and metering inaccuracy characteristic data; A model building module is used to establish a correlation model between the degree of inaccuracy and the key assessment indicators of the substation using the distribution data of the inaccuracy level of the single-phase meter as the independent variable and the key assessment indicator values ​​of each substation as the dependent variable; An expected test module is used to combine the association model and use a pre-built long short-term memory neural network to predict the change trend of key assessment indicators of test points under different misalignment levels to obtain expected indicator data; The test result generation module is used to generate business test results based on the expected indicator data and the actual assessment indicator data.

Images