A low-voltage area line loss rate reasonable range detection method and system

Abstract

The application discloses a low-voltage area line loss rate reasonable range detection method and system. The method comprises the following steps: user electric energy meter historical data mining and standardization processing; calculating the reasonable range of line loss rate; and calculating the user electric energy meter error prediction result. The application improves the accuracy of historical data, which is beneficial to improving the reasonable interval detection result precision; the CHMM model is selected for user electric energy meter error prediction, which can not only improve the prediction efficiency, but also make the prediction result more accurate; the application standardizes the line loss historical data mining result by the regression analysis method, eliminates the differences between the data, such as nature, dimension and order of magnitude, and applies the processed data sample to model training, which is beneficial to improving the training result precision. No matter how much the load rate is, the detection result accuracy of the method is higher than that of the traditional method.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of low-voltage transformer area line loss rate detection, and in particular to a low-voltage transformer area line loss rate reasonable range detection method and system. BACKGROUND

[0002] At present, the line loss rate detection methods for low-voltage transformer areas mainly include equivalent power method and pressure drop method. The above methods are difficult to meet the lean management needs of transformer area line loss because they do not consider the actual situation of low-voltage transformer areas. Therefore, it is crucial to improve the existing line loss rate detection methods.

[0003] Document [1] is Zhao Qingming. Transformer area line loss rate calculation method based on nearest neighbor propagation algorithm and random forest regression model [J]. Electric Power Systems and Automation, 2020, 32 (09): 94-98, which proposes a transformer area line loss rate calculation method based on nearest neighbor propagation clustering algorithm and random forest regression model, constructs a line loss model, and uses the model to predict electrical characteristic indexes; the PCA method is used to cluster the characteristic indexes obtained by the nearest neighbor propagation clustering algorithm; finally, the clustering data is trained by the random forest algorithm to obtain the line loss calculation result, and the calculation result is compared with the line loss theoretical value to determine whether the transformer area line loss rate is in a reasonable range.

[0004] Document [2] is Wang Peng, Bai Yuling, Wang Linmei, etc. Transformer area line loss rate calculation based on clustering division and bidirectional LSTM network [J]. Electronic Devices, 2022, 45 (04): 964-969, which proposes a transformer area line loss rate calculation method based on clustering division and bidirectional LSTM network, analyzes the clustering division of transformer types according to line loss rate influencing factors, establishes a line loss rate calculation model based on bidirectional LSTM network, obtains line loss rate theoretical value, and compares the theoretical value with the calculation result to analyze the rationality of line loss rate.

[0005] The detection results of document [1] and document [2] have low accuracy. SUMMARY

[0006] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art, and to provide a low-voltage transformer area line loss rate reasonable range detection method and system to improve the accuracy of the detection result.

[0007] To achieve the above purpose, a technical solution adopted by the present application is as follows: a low-voltage transformer area line loss rate reasonable range detection method, comprising:

[0008] Step S1, user electric energy meter historical data mining and standardization processing;

[0009] Step S2, calculating the reasonable range of line loss rate;

[0010] Step S3, calculating the user electric energy meter error prediction result.

[0011] Further, the user electric energy meter historical data mining and standardization processing includes the following steps:

[0012] Step a1, using K-means algorithm to mine and process the user electric energy meter historical data;

[0013] Step a2, standardizing the user electric energy meter historical data obtained by mining and processing by regression analysis method.

[0014] Further, the process of using K-means algorithm to mine and process the user electric energy meter historical data is as follows:

[0015] Step b1: selecting multiple initial clustering centers, and dividing different data sample objects to the nearest clustering center;

[0016] Step b2: obtaining new clustering centers according to the data sample object division result;

[0017] Step b3: setting the total number of independent variables equal to k, and the corresponding sample number is N, comparing the newly obtained clustering centers with the initial clustering centers, if they are consistent, output the clustering result, that is, the user electric energy meter historical data mining result D ij , otherwise return to step b1,

[0018] The The x ij represents the standardized independent variable, the represents the mean of the original data, and the s 2 represents the standard deviation,

[0019] Further, the user electric energy meter historical data obtained by mining and processing is standardized by the following regression analysis method: X ij represents the user electric energy meter historical data sample; the x max and x min are the maximum and minimum values of the original data.

[0020] Further, using Bayesian multivariate linear inference theory, the probability density function of the power point spacing resistance is obtained, and on this basis, the least square estimation method is used to estimate the fuzzy solution interval range of the power point load proportion, thereby obtaining the reasonable range of line loss rate.

[0021] Further, the process of calculating the reasonable range of line loss rate is as follows:

[0022] Step c1: calculating the service line loss value ΔL, the The R il represents the power supply point spacing resistance; the R L represents the service line resistance; the P d represents the power supply amount; Y ij represents the user electric energy meter historical data;

[0023] Step c2, convert ΔL into ΔL=r l0 +F1r l1 +F2r l2 +,...,+F n r ln , the r l0 , r l1 , r l2 , … r ln represents the expected value of the service line loss setting, the F1, F2…F n represents the service line length, and the calculation formula is: The α1, β1, x1 represent the coordinates of the distribution transformer substation, and the α2, β2, x2 represent the coordinates of the user electrical equipment;

[0024] Step c3, according to Y ij , the probability density function of the power supply point spacing resistance R il is calculated by the Bayesian multivariate linear inference theory: f(R il ) = 1 / (u-a), where a represents the resistance measurement value, that is, the cable resistance value from the distribution transformer substation to the user electrical equipment, and u is 2.5;

[0025] Step c4, calculate the m sample area Y-R il relationship: Y = Sr×φ, where Sr is an m×m r matrix, where m r represents the number of power supply points, m represents the sample area, and the relationship between them is m>m r , and φ represents the service line unit length resistance. The specific data of the m×m r matrix is determined according to the actual number of power supply points and the sample area;

[0026] Step c5, calculate the reasonable range of line loss rate [ΔL min , ΔL max ] = [R ij,min , R ij,max ] × [α, β], where [R ij,min , R ij,max ] represents the value range of R il at the 95% confidence level; and [α, β] represents the fuzzy solution interval range of the power supply point load proportion.

[0027] Further, the process of calculating the user electric energy meter error prediction result is as follows:

[0028] Step d1, data acquisition and preprocessing: set that there are Z groups of observation value sequences G={G 1 ,G 2 ,.....,G Z}, wherein the values of G 1 ,G 2 ,...,G Z are determined according to the real electric energy meter reading data, and then a mixed one-dimensional Gaussian probability density function is used to convert the CHMM observation probability density function:

[0029] The μ jn , η jn and respectively represent the mixed weight, mean value and variance of the mixed Gaussian density;

[0030] Step d2, determination of the historical line loss database, that is, determining the line loss database according to the mined user electric energy meter historical data, denoted as D={d1, d2,...,d N};

[0031] Step d3, auxiliary variable selection, that is, selecting measurable variables that affect the user electric energy meter error, and according to a certain sequence, the observation vectors of the CHMM are composed, which are used for CHMM model training;

[0032] Step d4, inputting G={G 1 ,G 2 ,.....,G Z} into the trained CHMM model, calculating the output probability P(S|σ i ) of the sample in the CHMM, and thus obtaining the user electric energy meter error prediction result H, so as to realize the line loss rate reasonable range detection.

[0033] Further, the measurable variables include load change, power supply line voltage size and line voltage asymmetry.

[0034] Further, the nearest cluster center is C γ , the The E represents the sum of square errors of the user electric energy meter historical data sample, h i represents the average value of the cluster center C k , u i represents the initial cluster center, and γ represents the data cluster.

[0035] Another technical solution of the present application is: a low-voltage transformer area line loss rate reasonable range detection system, which is used to realize the low-voltage transformer area line loss rate reasonable range detection method.

[0036] The present application has the following advantages: the present application mines user electric energy meter historical data by K-means algorithm, adopts regression analysis method to perform standardization and inverse normalization processing on the mined data, improves the accuracy of historical data, and is beneficial to improve the reasonable interval detection result precision; selecting CHMM model for user electric energy meter error prediction can not only improve the prediction efficiency, but also make the prediction result more accurate; the present application performs standardization processing on the line loss historical data mining result by regression analysis method, eliminates the differences between data in nature, dimension, order of magnitude, etc., and applies the processed data samples to model training, which is beneficial to improve the training result precision, so that the detection result accuracy of the method is higher than that of the traditional method regardless of the load rate. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The present application is a flowchart of a low-voltage transformer area line loss rate reasonable range detection method;

[0038] Figure 2 The present application is a ROC curve diagram of three methods in the specific embodiment;

[0039] Figure 3 The present application is a detection result diagram of three methods in the specific embodiment under the condition of 10% load rate;

[0040] Figure 4 The present application is a detection result diagram of three methods in the specific embodiment under the condition of 30% load rate;

[0041] Figure 5 The present application is a detection result diagram of three methods in the specific embodiment under the condition of 40% load rate. DETAILED DESCRIPTION

[0042] The technical solutions of the present application will be further described below in combination with the drawings and specific embodiments of the present application:

[0043] The present embodiment provides a low-voltage transformer area line loss rate reasonable range detection method, as shown in Figure 1 The present application includes the following steps:

[0044] Step S1: user electric energy meter historical data mining and standardization processing;

[0045] Step S2: line loss rate reasonable range calculation;

[0046] Step S3: user electric energy meter error prediction result calculation.

[0047] Specifically, the user electric energy meter historical data mining and standardization processing includes the following steps:

[0048] Step a1: using K-means algorithm to mine and process the user electric energy meter historical data;

[0049] Step a2: using regression analysis method to standardize the user electric energy meter historical data obtained by mining and processing.

[0050] More specifically, the process of using K-means algorithm to mine and process the user electric energy meter historical data is as follows:

[0051] Step b1: selecting multiple initial cluster centers, and dividing different data sample objects to the nearest cluster center;

[0052] Step b2: obtaining new cluster centers according to the data sample object division result;

[0053] Step b3: setting the total number of independent variables equal to k, and the corresponding sample number is N, comparing the newly obtained cluster centers with the initial cluster centers, if they are consistent, the clustering result, i.e. the user electric energy meter historical data mining result D ij , can be output, otherwise return to step b1, the The x ij represents the standardized independent variable, the represents the mean of the original data, and the s 2 represents the standard deviation, and the

[0054] Specifically, the standardization processing of the user electric energy meter historical data obtained by mining and processing by using regression analysis method is X ij represents the user electric energy meter historical data sample; the x max and x min are the maximum and minimum values of the original data.

[0055] There are two important influencing factors for user electric energy meter loss: (1) power point spacing resistance; (2) power point load proportion. Therefore, using Bayesian multivariate linear inference theory, the probability density function of the power point spacing resistance is obtained; on this basis, using the least square estimation method to estimate the fuzzy solution interval range of the power point load proportion, thereby obtaining the reasonable range of line loss rate.

[0056] Specifically, the process of calculating the reasonable range of line loss rate is as follows:

[0057] Step c1: calculating the service line loss value ΔL, the The R il represents the power point spacing resistance; the RL represents the service line resistance; the P d represents the power supply amount; Y ij represents the user electric energy meter historical data;

[0058] Step c2: convert ΔL into ΔL=r l0 +F1r l1 +F2r l2 +,...,+F n r ln , the r l0 , r l1 , r l2 , … r ln represents the set expected value of the service line loss, the F1, F2…F n represents the service line length, and the calculation formula is The α1, β1, x1 represent the coordinates of the distribution transformer substation, and the α2, β2, x2 represent the coordinates of the user electrical equipment;

[0059] Step c3: according to Y ij , the probability density function of the power receiving point spacing resistance R il is calculated by the Bayesian multivariate linear inference theory as f(R il ) = 1 / (u-a), where a represents the resistance measurement value, that is, the cable resistance value from the distribution transformer substation to the user electrical equipment, and u is 2.5;

[0060] Step c4: calculate the m sample areas Y=Sr×φ relationship as Y=Sr×φ, where S is an m×m r matrix, where m r represents the number of power receiving points, m represents the sample area, and the relationship between them is m>m r , and φ represents the unit length resistance of the service line; the specific data of the m×m r matrix is determined according to the actual number of power receiving points and the sample area;

[0061] Step c5: calculate the reasonable range of line loss rate [ΔL min , ΔL max ] = [R ij,min , R ij,max ]×[α, β], where [R ij,min , R ij,max ] represents the value range of R il at the 95% confidence level; and [α, β] represents the fuzzy solution interval range of the power receiving point load proportion.

[0062] Specifically, the process of calculating the user electric energy meter error prediction result is as follows:

[0063] Step d1, data collection and pre-processing: set Z groups of observation value sequences G = {G 1 , G 2 , ..., G Z}, wherein G 1 , G 2 , ..., G Z values are determined according to real electric energy meter readings, and then a mixed one-dimensional Gaussian probability density function is used to convert the CHMM observation value probability density function:

[0064] The μ jn , η jn and respectively represent user electric energy meter error prediction results, mixed weights of mixed Gaussian density, mean values and variances; in order to obtain more accurate user electric energy meter error prediction results, the CHMM model is trained through multiple groups of observation sample sequences;

[0065] Step d2, determination of a historical line loss database, that is, the line loss database is determined according to the mined user electric energy meter historical data, denoted as D = {d1, d2,..., d N};

[0066] Step d3, auxiliary variable selection, that is, measurable variables affecting user electric energy meter errors are selected, and observation vectors of the CHMM are formed in a certain sequence, which are used for CHMM model training;

[0067] Step d4, input G = {G 1 , G 2 , ..., G Z} into the trained CHMM model, according to the correlation between electric energy meter errors and influencing factors, select factors with higher correlation coefficients to form observation vectors, calculate the output probability P(S|σ i ) of the sample in the CHMM, and thus obtain user electric energy meter error prediction results H, so as to determine whether the user electric energy meter error result is in a reasonable range, thereby realizing line loss rate reasonable range detection.

[0068] Specifically, the measurable variables include load changes, power supply line voltage sizes and line voltage asymmetry. The nearest cluster center is C γ , the The E represents the sum of square errors of user electric energy meter historical data samples, h i represents the average value of the cluster center C k , and u i represents the initial cluster center. Under this criterion, each cluster itself is as compact as possible, and the clusters are as independent as possible, and γ represents the data cluster.

[0069] The embodiment also provides a low-voltage transformer area line loss rate reasonable range detection system for realizing the low-voltage transformer area line loss rate reasonable range detection method.

[0070] Experimental results and analysis:

[0071] In order to verify the feasibility of the low-voltage transformer area line loss rate reasonable range detection method based on historical data mining and improved hidden Markov model, experimental verification and analysis are performed.

[0072] According to the user electric energy meter error prediction result, different load rates (10%, 20%, 30%, 40%) are set, and the reasonable range of line loss rate under different conditions is calculated through the formula f(R il ) = 1 / (u-a). Table 1 is the specific calculation result.

[0073] Table 1 Line loss rate reasonable range

[0074]

[0075]

[0076] Based on the collected user electric energy meter historical data and line loss rate reasonable range calculation result, the method in document [1] and the method in document [2] are selected as comparison methods, and the line loss rate reasonable range detection results of the methods are compared with the line loss rate reasonable range detection result of the method, and the detection results of different methods are obtained. In order to ensure the accuracy of the experimental results, the MATLAB simulation software is used to process the experimental results, and at the same time, the consistency of the experimental conditions is ensured.

[0077] Results and analysis:

[0078] The detection performance of the above three methods is verified by introducing the concept of ROC curve, and the detection performance can be reflected by the area under the ROC curve. The larger the area is, the better the performance is. The detection results of the three methods are shown in Figure 2 .

[0079] As shown in Figure 2 , the uppermost curve is the detection result of the method, the middle curve is the detection result of document [1], and the lowermost curve is the detection result of document [2].

[0080] Figure 2 Among them, FPR represents the false positive rate. From Figure 2It can be seen that the area under the ROC curve of the method described in this invention is higher than that of the methods in references [1] and [2], indicating that the detection results of the method described in this invention within the reasonable range of line loss rate are more reasonable and the detection effect is better than the two traditional methods. The reason is that this method trains the CHMM model through multiple sets of observed sample sequences, making the model training results more reasonable, thereby making the detection results more reliable.

[0081] To further verify the reliability of the detection results of the method, three methods were used to detect the reasonable range of line loss rate at load rates of 10%, 30%, and 40%. Figures 2 to 4 The results compare the detection accuracy of different methods.

[0082] Depend on Figures 3 to 5 As can be seen, because the method described in this invention standardizes the results of historical line loss data mining through regression analysis, it eliminates differences in properties, dimensions, and orders of magnitude among the data. Applying the processed data samples to model training helps improve the accuracy of the training results. Therefore, under different load rates, the detection accuracy of the method described in this invention is higher than that of traditional methods. When the load rate is 10%, the highest detection accuracy of the method reaches 99.5%; when the load rate is 30%, the highest detection accuracy reaches 98.6%; and when the load rate is 40%, the highest detection accuracy reaches 97.9%. This indicates that the detection results of the method described in this invention are more accurate and can provide a reliable data foundation for controlling the line loss rate of power equipment such as meter boxes and energy meters.

[0083] Finally, the detection results of the three methods were verified by the deviation index. Here, the deviation specifically refers to the difference between the error of the user's electricity meter and the reasonable range of the line loss rate. The deviation of the measurement results of different methods was calculated, and the specific comparison results are given in Table 2.

[0084] Table 2 Comparison of Deviation Results of Calculation Results from Different Methods

[0085]

[0086] As can be seen from the data in Table 2, the minimum value of the calculation result of the method described in this invention is 0.18, the minimum value of the calculation result of the method in reference [1] is 0.52, and the minimum value of the calculation result of the method in reference [2] is 0.73. The deviation result of the method described in this invention is reduced by 0.22 and 0.43 compared with the methods in reference [1] and reference [2], respectively. It can be seen that the line loss rate calculation result of the method described in this invention is more accurate, which also reflects the rationality of its line loss rate reasonable range detection result.

[0087] Comprehensive analysis of the above experimental results can be known that the method can not only realize accurate measurement of line loss rate, and line loss rate reasonable interval detection result is more reliable.

[0088] Conclusion:

[0089] With the research goal of improving the reliability of the detection result of the reasonable range of line loss rate, a low-voltage transformer area line loss rate reasonable range detection method based on historical data mining and improved hidden Markov model is proposed. The historical data is mined and standardized to improve the accuracy of the detection result of the reasonable range of line loss rate. The reasonable range of line loss rate is calculated, the CHMM model is applied to the user electric energy meter error prediction, the prediction result is measured with the calculation result of the reasonable range, and the low-voltage transformer area line loss rate reasonable range detection is completed. The experimental results show that the measurement deviation of the method is reduced by 0.22 and 0.43 respectively compared with the method in document [1] and the method in document [2], and the area under the ROC curve of the method is significantly higher than that of the traditional method. The above results show that the detection effect of the method is good, and can provide help for the lean management of transformer line loss.

[0090] It should be noted that the above enumeration is only one specific embodiment of the present application. Obviously, the present application is not limited to the above embodiments, but can have many variations. In short, all variations that can be directly derived or inferred from the disclosed content by those skilled in the art should be considered as the protection scope of the present application.

Claims

1. A method for detecting the reasonable range of line loss rate in low-voltage distribution areas, characterized in that, include: Step S1: Mining and standardizing historical data from user electricity meters; Step S2: Calculate the reasonable range of line loss rate; Step S3: Calculate the error prediction result of the user's electricity meter; In step S2, the probability density function of the resistance between the receiving points is obtained using Bayesian multivariate linear inference theory. Based on this, the range of the fuzzy solution of the load proportion of the receiving point is estimated using the least squares estimation method, thereby obtaining a reasonable range of the line loss rate. The process for calculating the reasonable range of line loss rate is as follows: Step c1, calculate the service connection loss value ΔL, the The R fl Represents the resistance between the power receiving points; the R L Indicates the service line resistance; the P d Indicates power supply; X ij This represents a sample of historical data from a user's electricity meter. Step c2, convert ΔL to ΔL = r l0 +F1r l1 +F2r l2 +,...,+F n r ln The r l0 r l1 r l2 ...r ln This represents the expected value of the service drop setting, where F1, F2…F… represent the service drop setting. n The length of the service connection line is indicated by the following formula: i represents the serial number of the service line, α1, β1, and x1 represent the coordinates of the substation, and α2, β2, and x2 represent the coordinates of the user's electrical equipment. Step c3, according to X ij The resistance R between the receiving points is determined using Bayesian multivariate linear inference theory. fl The probability density function is used to calculate: f(R) fl ) = 1 / (ua), where a represents the resistance measurement value, that is, the cable resistance value from the substation to the user's electrical equipment, and u is 2.5; Step c4, calculate YR for m sample areas fl Relationship: Y = Sr × φ, where Sr is an m × m r The matrix, where m r This represents the number of power receiving points, and m represents the sample station area. The relationship between the two is m > m r φ represents the resistance per unit length of the service line, and the m×m r The specific data for the matrix are determined based on the actual number of power receiving points and the sample station area; Step c5, calculate the reasonable range of line loss rate [ΔL] min ,ΔL max ] = [R fl,min ,R fl,max ]×[α,β], the [R fl,min ,R fl,max ] indicates R fl The range of values ​​at a 95% confidence level; [α,β] represents the fuzzy solution range of the load proportion at the power receiving point.

2. The method for detecting the reasonable range of line loss rate in low-voltage distribution areas according to claim 1, characterized in that, The historical data mining and standardization process for user electricity meters includes the following steps: Step a1: The K-means algorithm is used to mine and process the historical data of user electricity meters; Step a2 involves standardizing the historical user electricity meter data obtained through the mining process using regression analysis.

3. The method for detecting the reasonable range of line loss rate in low-voltage distribution areas according to claim 2, characterized in that, The process of mining and processing historical data of user electricity meters using the K-means algorithm is as follows: Step b1: Select multiple initial cluster centers and assign different data sample objects to the nearest cluster center; Step b2: Obtain new cluster centers based on the data sample object partitioning results; Step b3: Set the total number of independent variables to k, and the corresponding number of samples to N. Compare the newly acquired cluster centers with the initial cluster centers. If they match, output the clustering result, i.e., the user's electricity meter historical data mining result D. ij Otherwise, return to step b1, where... The x ij The standardized independent variable is described below. The s represents the mean of the original data. 2 The standard deviation is represented by the following.

4. The method for detecting the reasonable range of line loss rate in low-voltage distribution areas according to claim 3, characterized in that, The historical user electricity meter data obtained through data mining was standardized using regression analysis as follows: X ij This represents a sample of historical data from a user's electricity meter; the x max and x min These are the maximum and minimum values ​​of the original data.

5. The method for detecting the reasonable range of line loss rate in low-voltage distribution areas according to claim 1, characterized in that, The process for calculating the error prediction result of the user's electricity meter is as follows: Step d1, Data Acquisition and Preprocessing: Assume there are Z sets of observation sequences G = {G 1 G 2 ,.....,G Z }, where G represents the observed value, G 1 G 2 ,...,G Z The values ​​are determined based on actual electricity meter readings, and then the probability density function of the CHMM observations is transformed and represented using a mixture of one-dimensional Gaussian probability density functions: The μ jn η jn and These represent the error prediction results of the user's electricity meter, the mixing weights of the Gaussian mixture density, the mean, and the variance, respectively. Step d2: Determining the historical line loss database, i.e., determining the line loss database based on the mined historical data of user electricity meters, denoted as D={d1,d2,...,d N }; Step d3, auxiliary variable selection, that is, selecting measurable variables that affect the error of user electricity meters, forming the observation vector of CHMM according to a certain sequence, and using it for CHMM model training; Step d4: Set G = {G} 1 G 2 ,.....,G Z The input is fed into the trained CHMM model to calculate the output probability P(G|σ) of the sample in the CHMM. i This allows us to obtain the user's electricity meter error prediction result H. This allows us to determine whether the error result of the user's electricity meter is within a reasonable range, thereby enabling the detection of a reasonable range for line loss rate.

6. The method for detecting the reasonable range of line loss rate in low-voltage distribution areas according to claim 5, characterized in that, The measurable variables include load changes, power supply line voltage magnitude, and line voltage asymmetry.

7. The method for detecting the reasonable range of line loss rate in low-voltage distribution areas according to claim 3, characterized in that, The nearest cluster center is C. γ The E represents the sum of squared errors of historical data samples from user electricity meters, h i Represents the cluster center C γ The average value, u i γ represents the initial cluster center, and γ represents the data cluster.

8. A system for detecting the reasonable range of line loss rate in low-voltage distribution areas, characterized in that, This method is used to implement the method for detecting the reasonable range of line loss rate in low-voltage distribution areas as described in any one of claims 1-7.

Images