Dual-source checked wideband transaction identification system and detection method

Abstract

The present application belongs to the technical field of wideband detection of power system, and particularly relates to a wideband abnormality identification system and detection method with double-source checking. The present application comprises: supplementing SCADA data with Lagrange interpolation method for data correspondence, evenly interpolating between adjacent two measurement values until matching of two-source measurement data is achieved, and each wideband measurement value has corresponding SCADA measurement value; difference calculation is performed on the same node of wideband measurement value and SCADA measurement value interpolation data; the improved genetic algorithm is used to optimize the correct data of BP neural network for difference training; the collected measurement difference which has been subjected to time processing is input into the trained model, and whether there is bad data can be judged through the output value; the feature smoothing line checking method is used to identify the existence of bad data of wideband device. The present application can effectively identify the front-end bad data of two data sources, has small calculation amount, short time consumption, and the checking result is more reliable.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of wideband detection of power systems, and particularly relates to a wideband abnormality identification system and detection method based on double-source checking, more particularly to a wideband abnormality identification system and detection method based on mutual checking of two sources composed of wideband and SCADA. BACKGROUND

[0002] In recent years, with the wide application of power electronic technology, various energies coexist in the power grid, and the scale and structure of the power grid are developing continuously. The data type is increasing, and the data volume is also expanding rapidly. When the dispatch personnel observes and controls the power grid, they need to master comprehensive and accurate data, and collect, sort, mine and analyze different forms of data based on the measurement information from multiple front-end data sources. In the process of collecting data, due to communication transmission and other reasons, there may be contradictions between the front-end measurements of various sources, indicating that there are measurement values deviating from the true value in the data, i.e. bad data. In order to avoid the bad data from causing errors in subsequent analysis and making the dispatch center make wrong decisions to affect the subsequent operation of the power grid, the original measurement data needs to be cleaned to find out the bad data.

[0003] The data acquisition and monitoring control system is widely used in the power grid and has absolute advantages in the number of distribution points and the amount of collected data information, which has improved the observability of the power system. However, due to measurement errors or transmission errors, there may be bad data.

[0004] Traditional bad data identification methods mainly include residual search method, zero residual method and estimation identification method, and there are new methods combining the characteristics obtained after state estimation with machine learning and data fusion. The large amount of data generated by intensive measurement also brings challenges to checking. Some scholars use state estimation and abnormal point detection algorithms such as pattern recognition and local outlier factor algorithm to identify bad data of wideband device measurement values, which are all focused on checking single source data. Considering that SCADA and wideband devices are distributed in the power grid to collect measurement values, SCADA and wideband devices may measure the same node at the same time and can complement each other. For this situation, some documents propose to combine the SCADA state estimation results and wideband device measurement values to perform secondary linear state estimation, and simultaneously update the residual covariance matrix of the secondary linear state estimation to detect whether there are bad data in the key measurements of SCADA measurements. However, it does not check the wideband device data, and needs to be iterated multiple times, which is relatively complex. With the continuous enrichment of multi-source information in the power system, reasonable use of the identification results of bad data detection will be more reliable than single source judgment. SUMMARY

[0005] To address the shortcomings of the existing technology, this invention provides a broadband anomaly identification system and detection method with dual-source verification. Its purpose is to achieve the objectives of the invention.

[0006] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0007] The dual-source verified broadband anomaly identification system uses time-interpolation and subtraction of measurements from two sources at the same node, and the resulting difference is used as the detection feature. Based on the difference between the two sources under normal conditions, an improved BP neural network algorithm is used to train the model, obtaining the optimal training parameters. The trained model is then used for rapid initial screening of the data to be detected. A gap statistics algorithm is used to verify the boundary data to determine if there are contradictions between the two source data. After identifying contradictions, the broadband device data at that moment is subjected to abnormal data identification, ultimately determining the source of the abnormal data.

[0008] The broadband anomaly detection method with dual-source verification includes the following steps:

[0009] Step 1. Use the Lagrange interpolation method to supplement SCADA data and make data correspondence. Interpolate evenly between two adjacent measurement values ​​until the matching of the two source measurement data is achieved, and each broadband measurement value has a corresponding SCADA measurement value.

[0010] Step 2. Calculate the difference between the broadband measurement value and the SCADA measurement value interpolation data at the same node;

[0011] Step 3. Use the difference between the correct data and the optimized BP neural network using the improved genetic algorithm to train and obtain the output model;

[0012] Step 4. Input the collected, time-synchronized measurement differences into the trained model, and the output value can be used to determine whether bad data exists;

[0013] Step 5. In the bad data sequence, the feature smoothing line verification method is used to determine that there is bad data in the broadband device.

[0014] Furthermore, the method of using Lagrange interpolation to supplement SCADA data for data correspondence involves uniform interpolation between adjacent measurement values ​​until matching of the two source measurement data is achieved. Each broadband measurement value has a corresponding SCADA measurement value, including:

[0015] Lagrange interpolation is performed on n sampling points in the SCADA measurement to fill in the missing values ​​compared with the broadband.

[0016]

[0017] Where n represents the number of known data points, i represents the i-th data point, j represents the j-th data point, x represents the x-axis component corresponding to the data point, y represents the y-axis component corresponding to the data point, and L(x) represents the missing data.

[0018] Furthermore, the difference between the broadband measurement value and the SCADA measurement value interpolated data at the same node is calculated, and the expression is:

[0019] D(t i )=y(t i )-L(t i (2)

[0020] Where D(t) i Let y(t) be the difference at point i. i Let L(i) be the broadband measurement value at the i-th point. t ) is the SCADA interpolation supplementary value for the i-th point.

[0021] Furthermore, the training of the difference between correct data and the optimized BP neural network using the improved genetic algorithm includes the following steps:

[0022] Step 31. Assume the input layer has M nodes, the hidden layer has L nodes, and the output layer has J nodes, with the following relationship:

[0023]

[0024] In the above formula, n is the number of iterations, and y kj For the actual output in the network, w ml Let θ be the weights of the input layer and the hidden layer. l v is the threshold of the hidden layer. ij Let r be the weights of the hidden layer and the output layer. j The threshold of the output layer, D(t) i () represents the difference at point i;

[0025] The error formula is obtained through training:

[0026]

[0027] In the above formula, d kj (n) represents the budget output, y kj Here, E is the actual output of the network, n is the number of iterations, ε is a number close to 0, J is the number of output layer nodes, and L is the number of hidden layer nodes.

[0028] The fitness function is the reciprocal of the network error.

[0029]

[0030] In the above formula: fitness(w,v,θ,r) is the fitness value, x i For the experimental results; y i ζ is the predicted value; N is the total number of samples; ζ is a small value that ensures the denominator is not zero at any time, preventing the genetic algorithm from being limited to a small local range for optimization.

[0031] Step 32. Perform selection, crossover, and mutation operations on the improved genetic algorithm; the crossover and mutation probabilities are adaptively adjusted as follows:

[0032]

[0033]

[0034] In the formula, f max It is the maximum fitness value, f ave f is the average fitness value, f' is the larger fitness value between the two individuals in the crossover probability, f is the fitness value of the individual to be mutated, and c1, c2, c3, c4 are constant values ​​from 0 to 1.

[0035] Step 33. Obtain the output result after optimization by the improved genetic algorithm;

[0036] The input and output layers of the neural network after optimization using the adaptive genetic algorithm are as follows:

[0037] The input variable net of the i-th node in the hidden layer of the neural network i :

[0038]

[0039] In the formula: w ml To initialize the weights of each layer of the neural network, θ l The threshold value for the hidden layer is M, and the number of nodes in the input layer is M.

[0040] The output variable y of the i-th node in the hidden layer of the neural network j k :

[0041]

[0042] The output variable of the k-th node in the output layer of the neural network

[0043]

[0044] In the formula: v ij Let be the weights from the i-th point to the j-th point in the output layer. Let q be the output variable of the k-th node, and a be the number of nodes. kTo initialize the threshold, i is the i-th node, M is the number of learning pairs, and φ is the activation function of the hidden layer of the network.

[0045] Furthermore, the step of inputting the collected, time-synchronized measurement differences into the trained model, and then determining whether bad data exists based on the output value, involves inputting the collected, time-synchronized measurement differences into the trained improved BP neural network, and then determining whether bad data may exist based on the output value.

[0046] Furthermore, the step of identifying bad data in the broadband device using a feature smoothing line verification method for bad data sequences includes:

[0047] Step 51. For the wideband data sample set U, calculate the maximum and minimum values ​​of the wideband signal data at the same time:

[0048]

[0049] In the above formula, X max X represents the maximum value of the broadband signal data at the same time. min The minimum value of broadband signal data at the same time. The maximum value of the collected data within the same period. This represents the minimum value of the collected data within the same period.

[0050] Step 52. Extract the feature curves through comparison:

[0051] X d =[x d1 ,x d2 ,...,x d200 (13)

[0052] In the above formula, X d For the characteristic curve, X d1 ...x d200 These are the 200 values ​​that constitute the characteristic curve;

[0053] Step 53. Calculate the upper and lower limits of the historical data domain using the maximum and minimum values ​​of current, voltage, or harmonics from historical data:

[0054]

[0055] In the above formula, All are dimensionless values ​​of the allowable variation threshold, x d_i For x d The i-th value, where i is the i-th point;

[0056] Step 54. Calculate the upper and lower limits of the feasible region for current, voltage, or harmonic data:

[0057]

[0058] In the above formula, This represents the upper limit of the feasible region. K is the lower bound of the feasible region. t The threshold coefficient that allows for variation, and All are dimensionless values ​​of the allowable change threshold, where i is the i-th point;

[0059] Using the calculated upper and lower limits of the feasible region, a feasible region matrix is ​​formed for current, voltage, or harmonic data. The feasible region matrix for broadband signals is:

[0060]

[0061] Each column element of the feasible region matrix Represents the and All are [-K1,+K i ] Dimensionless values ​​between; in anomaly data identification, for the test data X new =[x new,1 ,x new,2 ,...,x new,96 When X new The i-th component satisfies x new,i ≤x d_i At that time, calculate the upper limit of the data to be inspected:

[0062]

[0063] In the above formula, x represents the upper limit of the data to be inspected. new，i For the i-th component;

[0064] When the first component satisfies the condition, calculate the lower limit of the data to be inspected:

[0065]

[0066] The dual-source verified broadband anomaly detection device includes:

[0067] The acquisition module is used to supplement SCADA data and perform data correspondence using the Lagrange interpolation method;

[0068] The first calculation module is used to interpolate between two adjacent measurements in SCADA data until the matching of the two source measurement data is achieved. Each broadband measurement value has a corresponding SCADA measurement value. The module also performs difference calculation on the same node of the interpolated data of the broadband measurement value and the SCADA measurement value.

[0069] The broadband anomaly discrimination model for dual-source verification of substations is a model trained by the difference of correct data using an improved BP neural network, which is used to verify the collected data.

[0070] The second computation module is used to input the collected, time-synchronized measurement difference values ​​into the trained support vector machine. The output value can be used to determine whether bad data exists. The feature smoothing line verification method is used to identify bad data in the broadband device, verify whether the broadband signal jumps, and determine whether there are errors in the broadband device data.

[0071] A computer device includes a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the dual-source verification broadband anomaly identification and detection methods described above.

[0072] A computer storage medium storing a computer program, wherein when the computer program is executed by a processor, the steps of any of the dual-source verification broadband anomaly identification and detection methods are implemented.

[0073] The present invention has the following beneficial effects and advantages:

[0074] The method of this invention employs an optimized BP neural network-based identification approach to identify anomalies based on the difference between two source measurements. It then examines the broadband device measurements at the corresponding time point to ultimately determine the location of the defective data. This method effectively identifies defective data from both data sources, requires less computation, and is less time-consuming, resulting in more reliable verification than data from only one source. Attached Figure Description

[0075] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0076] Figure 1 This is a simulation flowchart of an embodiment of the present invention;

[0077] Figure 2 This is a flowchart of the algorithm proposed in this invention;

[0078] Figure 3 This invention provides an improved BP neural network and a traditional BP neural network for regression analysis. Detailed Implementation

[0079] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0080] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0081] The following reference Figures 1-3 The technical solutions of some embodiments of the present invention are described below.

[0082] Example 1

[0083] This invention provides an embodiment of a broadband anomaly identification system with dual-source verification. The invention uses the time-interpolated and subtracted differences of measurements from two sources at the same node as the detection feature. To overcome the problem of insufficient defective data samples in the power grid, the system is trained solely on the differences between the two sources under normal conditions. An improved BP neural network algorithm is used to obtain the optimal training parameters of the model, which is then used to quickly screen the data to be detected. Next, a gap statistics algorithm is used to verify the boundary data to determine whether there are contradictions between the two source data. After identifying contradictions, defective data identification is performed on the broadband device data at that moment, ultimately determining the defective data source.

[0084] Example 2

[0085] This invention provides another embodiment of a broadband anomaly identification and detection method with dual-source verification, comprising the following steps:

[0086] Step 1. Use the Lagrange interpolation method to supplement SCADA data and make data correspondence. Interpolate evenly between two adjacent measurement values ​​until the matching of the two source measurement data is achieved, and each broadband measurement value has a corresponding SCADA measurement value.

[0087] Step 2. Calculate the difference between the broadband measurement value and the SCADA measurement value interpolation data at the same node;

[0088] Step 3. Use the difference between the correct data and the optimized BP neural network using the improved genetic algorithm to train and obtain the output model;

[0089] Step 4. Input the collected, time-synchronized measurement differences into the trained model, and the output value can be used to determine whether bad data exists;

[0090] Step 5. In the bad data sequence, the feature smoothing line verification method is used to determine that there is bad data in the broadband device.

[0091] SCADA refers to a computer-based automated production process control and scheduling system. It can monitor and control the operating equipment on site.

[0092] Furthermore, in step 1, according to formula (1), Lagrange interpolation is performed on the n sampling points in the SCADA measurement values ​​to fill in the missing amount compared with the broadband.

[0093]

[0094] Where n represents the number of known data points, i represents the i-th data point, j represents the j-th data point, x represents the x-axis component corresponding to the data point, y represents the y-axis component corresponding to the data point, and L(x) represents the missing data.

[0095] Furthermore, in step 2, the difference between the broadband measurement value and the SCADA measurement value interpolation data at the same node is calculated, and the difference expression is obtained as follows:

[0096] D(t i )=y(t i )-L(t i (2)

[0097] Where D(t) i Let y(t) be the difference at point i. i Let L(i) be the broadband measurement value at the i-th point. t ) is the SCADA interpolation supplementary value for the i-th point.

[0098] Furthermore, in step 3, the difference training using the improved BP neural network with correct data is employed, specifically including the following steps:

[0099] Step 31. Assume the input layer has M nodes, the hidden layer has L nodes, and the output layer has J nodes, with the following relationship:

[0100]

[0101] In the above formula, n is the number of iterations, and y kj For the actual output in the network, w ml Let θ be the weights of the input layer and the hidden layer. l v is the threshold of the hidden layer. ij Let r be the weights of the hidden layer and the output layer. j The threshold of the output layer, D(t) i ) represents the difference at point i.

[0102] The error formula can be obtained through training:

[0103]

[0104] In the above formula, d kj (n) represents the budget output, y kjLet E be the actual output of the network, n be the number of iterations, and ε be a number close to 0. J is the number of output layer nodes, and L is the number of hidden layer nodes.

[0105] The fitness function is the reciprocal of the network error.

[0106]

[0107] In the above formula: fitness(w,v,θ,r) is the fitness value. i For the experimental results; y i ζ is the predicted value; N is the total number of samples; ζ is a small value that ensures the denominator is never zero, preventing the genetic algorithm from being limited to a small local range for optimization.

[0108] Step 32. Perform selection, crossover, and mutation operations on the improved genetic algorithm; wherein the crossover probability and mutation probability are adaptively adjusted, as shown in equations (7) and (8).

[0109]

[0110]

[0111] In the formula, f max It is the maximum fitness value, f ave f is the average fitness value, f' is the larger fitness value between the two individuals in the crossover probability, f is the fitness value of the individual to be mutated, and c1, c2, c3, c4 are constant values ​​from 0 to 1.

[0112] Step 33. Finally, the output result optimized by the improved genetic algorithm is obtained.

[0113] The formulas for the input and output layers of the neural network after optimization by the adaptive genetic algorithm are shown in equations (9) and (11).

[0114] The input variable net of the i-th node in the hidden layer of the neural network i :

[0115]

[0116] In the formula: w ml To initialize the weights of each layer of the neural network, θ l is the threshold of the hidden layer, and M is the number of nodes in the input layer.

[0117] The output variable of the i-th node in the hidden layer of a neural network

[0118]

[0119] The output variable of the k-th node in the output layer of the neural network

[0120]

[0121] In the formula: v ij Let be the weights from the i-th point to the j-th point in the output layer. Let q be the output variable of the k-th node, and a be the number of nodes. k To initialize the threshold, i is the i-th node, M is the number of learning pairs, and φ is the activation function of the hidden layer of the network.

[0122] Furthermore, in step 4, the collected measurement differences that have undergone time synchronization are then input into the trained improved BP neural network, and the output value can be used to determine whether bad data may exist.

[0123] Furthermore, in step 5, the characteristic curve verification method is used to determine that there is bad data in the broadband device.

[0124] Step 51. For the wideband data sample set U, calculate the maximum and minimum values ​​of the wideband signal data at the same time:

[0125]

[0126] In the above formula, X max X represents the maximum value of the broadband signal data at the same time. min The minimum value of broadband signal data at the same time. The maximum value of the collected data within the same period. This represents the minimum value of the collected data within the same period.

[0127] Step 52. Extract the feature curves through comparison:

[0128] X d =[x d1 ,x d2 ,...,x d200 (13)

[0129] In the above formula, X d For the characteristic curve, X d1 ...x d200 These are the 200 values ​​that constitute the characteristic curve.

[0130] Step 53. Using the maximum and minimum values ​​of current, voltage, or harmonics from historical data, the upper and lower limits of the historical data domain can be calculated using equation (14):

[0131]

[0132] In the above formula, All are dimensionless values ​​of the allowable variation threshold, x d_i For x d The i-th value, where i is the i-th point.

[0133] Step 54. Calculate the upper and lower limits of the feasible region for current, voltage, or harmonic data:

[0134]

[0135] In the above formula, This represents the upper limit of the feasible region. K is the lower bound of the feasible region. t The threshold coefficient that allows for variation, and All are dimensionless values ​​of the allowable change threshold, where i is the i-th point.

[0136] Using the calculated upper and lower limits of the feasible region, a feasible region matrix is ​​formed for the current, voltage, or harmonic data. The feasible region matrix for this broadband signal is:

[0137]

[0138] Each column element of the feasible region matrix Represents the and All are [-K1,+K i ] Dimensionless values ​​between []. In anomaly data identification, for the test data X... new =[x new,1 ,x new,2 ,...,x new,96 When X new The i-th component satisfies x new,i ≤x d_i Then, use equation (17) to calculate the upper limit of the data to be inspected:

[0139]

[0140] In the above formula, x represents the upper limit of the data to be inspected. new，i Let i be the i-th component.

[0141] When the first component satisfies the condition, the lower limit of the data to be inspected is calculated using equation (18):

[0142]

[0143] Example 3

[0144] The present invention provides another embodiment of a broadband anomaly identification device with dual-source verification, comprising:

[0145] The acquisition module is used to supplement SCADA data and perform data correspondence using the Lagrange interpolation method;

[0146] The first calculation module is used to interpolate between two adjacent measurements in SCADA data until the matching of the two source measurement data is achieved. Each broadband measurement value has a corresponding SCADA measurement value. The module also performs difference calculation on the same node of the interpolated data of the broadband measurement value and the SCADA measurement value.

[0147] The broadband anomaly discrimination model for dual-source verification of substations is a model trained by the difference of correct data using an improved BP neural network, which is used to verify the collected data.

[0148] The second computation module is used to input the collected, time-synchronized measurement difference values ​​into the trained support vector machine. The output value can be used to determine whether bad data exists. The feature smoothing line verification method is used to identify bad data in the broadband device, verify whether the broadband signal jumps, and determine whether there are errors in the broadband device data.

[0149] Example 4

[0150] Based on the same inventive concept, embodiments of the present invention also provide a computer device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor. When the processor executes the computer program, it implements the steps of any of the dual-source verification broadband anomaly identification and detection methods described in embodiments 1-3.

[0151] Example 5

[0152] Based on the same inventive concept, this embodiment of the invention also provides a computer storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the dual-source verification broadband anomaly identification and detection methods described in embodiments 1-3.

[0153] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0154] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0155] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0156] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A broadband anomaly detection method with dual-source verification, characterized by: Includes the following steps: Step 1. Use the Lagrange interpolation method to supplement SCADA data and make data correspondence. Interpolate evenly between two adjacent measurement values ​​until the matching of the two source measurement data is achieved, and each broadband measurement value has a corresponding SCADA measurement value. Step 2. Calculate the difference between the broadband measurement value and the SCADA measurement value interpolation data at the same node; Step 3. Use the improved genetic algorithm to optimize the BP neural network and train the difference from Step 2 to obtain the output model; Step 4. Input the collected, time-synchronized measurement differences into the trained model, and the output value can be used to determine whether bad data exists; Step 5. Next, use the feature smoothing line verification method to identify bad data in the broadband device; check whether the broadband signal jumps to determine whether there are errors in the broadband device data; The method of supplementing SCADA data using Lagrange interpolation is used to achieve data correspondence. Interpolation is performed uniformly between adjacent measurements until matching of the two source measurements is achieved. Each broadband measurement value has a corresponding SCADA measurement value, including: Lagrange interpolation is performed on n sampling points in the SCADA measurement to fill in the missing values ​​compared with the broadband. （1）； Where n represents the number of known collection points, i represents the i-th data point, j represents the j-th data point, x represents the x-axis component corresponding to the collection point, y represents the y-axis component corresponding to the collection point, and L(x) represents the missing data. The difference between the broadband measurement value and the SCADA measurement value interpolated data at the same node is calculated, and the expression is: （2）； Where: D(t) i Let be the difference at point i. For the broadband measurement value at the i-th point, Provides the SCADA interpolation supplementary value for the i-th point; The step of further identifying bad data sequences using the feature smoothing line verification method includes: Step 51. For the broadband data sample set U, calculate the maximum and minimum values ​​of the broadband signal data at the same time: （12）； In the above formula, X max X represents the maximum value of the broadband signal data at the same time. min The minimum value of broadband signal data at the same time. ... The maximum value of the collected data within the same period. ... This represents the minimum value of the collected data within the same period. Step 52. Extract the feature curves through comparison: （13）； In the above formula, X d For the characteristic curve, X d1…… X d200 These are the 200 values ​​that constitute the characteristic curve; Step 53. Calculate the upper and lower limits of the historical data domain using the maximum and minimum values ​​of current, voltage, or harmonics from historical data: （14）； In the above formula, , All are dimensionless values ​​representing allowable variation thresholds. For x d The i-th value, where i is the i-th point; Step 54. Calculate the upper and lower limits of the feasible region for current, voltage, or harmonic data: （15）； In the above formula, This represents the upper limit of the feasible region. K is the lower bound of the feasible region. t The threshold coefficient that allows for variation, and All are dimensionless values ​​of the allowable change threshold, where i is the i-th point; Using the calculated upper and lower limits of the feasible region, a feasible region matrix is ​​formed for current, voltage, or harmonic data. The feasible region matrix for broadband signals is: （16）； Each column element of the feasible region matrix [ , ] represents the and All are [-K1,+K i ] Dimensionless values ​​between; in anomaly data identification, for the data to be tested ,when The i-th component satisfies ≤ At that time, calculate the upper limit of the data to be inspected: （17）； In the above formula, x represents the upper limit of the data to be inspected. new，i For the i-th component; when The i-th component satisfies ≥ At that time, calculate the lower limit of the data to be inspected: （18）。 2. The broadband anomaly identification and detection method with dual-source verification according to claim 1, characterized in that: The training of the difference between correct data and the optimized BP neural network using the improved genetic algorithm includes the following steps: Step 31. Assume the input layer has M nodes, the hidden layer has L nodes, and the output layer has J nodes, with the following relationship: (4) ； In the above formula, n is the number of iterations. This is the actual output in the network. These are the weights in the input layer and the hidden layer. The threshold of the hidden layer, These are the weights of the hidden layer and the output layer. The threshold of the output layer, D(t) i () represents the difference at point i; The error formula is obtained through training: (5) ； In the above formula, d kj (n) represents the budget output, y kj Here, E represents the actual output of the network, E is the network error, and n is the number of iterations. Let J be a number close to 0, where J is the number of output layer nodes and L is the number of hidden layer nodes. The fitness function is the reciprocal of the network error. (6) ； In the above formula: x is the fitness value. i For the experimental results; y i The predicted value is N; N is the total sample size. It is a small value, and its purpose is to ensure that the denominator is never zero, so as to prevent the genetic algorithm from being limited to a small local range for optimization; Step 32. Perform selection, crossover, and mutation operations on the improved genetic algorithm; the crossover and mutation probabilities are adaptively adjusted as follows: (7) ； (8) ； In the formula, It is the maximum fitness value. It is the average fitness value. It is the larger fitness value among the two individuals in the crossover probability. It is the fitness value of the individual to be mutated. It is a constant value from 0 to 1; Step 33. Obtain the output result after optimization by the improved genetic algorithm; The input and output layers of the neural network after optimization using the adaptive genetic algorithm are as follows: The input variable net of the i-th node in the hidden layer of the neural network i : (9) ； In the formula: To initialize the weights of each layer of the neural network, The threshold value for the hidden layer is M, and the number of nodes in the input layer is M. The output variable of the i-th node in the hidden layer of a neural network : (10)； The output variable of the k-th node in the output layer of the neural network : ； (11)； In the formula: Let be the weights from point i to point j in the output layer. Let q be the output variable of the k-th node, and a be the number of nodes. k To initialize the threshold, i is the i-th node, M is the number of learning pairs, and φ is the activation function of the hidden layer of the network.

3. The broadband anomaly identification and detection method with dual-source verification according to claim 1, characterized in that: The step of inputting the collected, time-synchronized measurement differences into the trained model and determining whether bad data exists based on the output value involves inputting the collected, time-synchronized measurement differences into the trained improved BP neural network and determining whether bad data may exist based on the output value.

4. A broadband anomaly identification device with dual-source verification, characterized in that: include: The data acquisition module is used to supplement SCADA data and perform data correspondence using the Lagrange interpolation method; The first calculation module is used to interpolate between two adjacent measurement values ​​in SCADA data until the matching of the two source measurement data is achieved. Each broadband measurement value has a corresponding SCADA measurement value. The difference between the broadband measurement value and the SCADA measurement value interpolation data is calculated at the same node. The wideband anomaly discrimination model for dual-source verification of substations is obtained by training the difference values ​​using an improved genetic algorithm-optimized BP neural network to obtain the output model. The second calculation module is used to input the collected measurement difference values ​​that have been time-synchronized into the trained model. The output value can be used to determine whether there is bad data. Then, the bad data sequence is used to identify bad data in the broadband device using the feature smoothing line verification method, and to verify whether the broadband signal jumps, and to determine whether there are errors in the broadband device data. The method of supplementing SCADA data using Lagrange interpolation is used to achieve data correspondence. Interpolation is performed uniformly between adjacent measurements until matching of the two source measurements is achieved. Each broadband measurement value has a corresponding SCADA measurement value, including: Lagrange interpolation is performed on n sampling points in the SCADA measurement to fill in the missing values ​​compared with the broadband. （1）； Where n represents the number of known collection points, i represents the i-th data point, j represents the j-th data point, x represents the x-axis component corresponding to the collection point, y represents the y-axis component corresponding to the collection point, and L(x) represents the missing data. The difference between the broadband measurement value and the SCADA measurement value interpolated data at the same node is calculated, and the expression is: （2）； Where: D(t) i Let be the difference at point i. For the broadband measurement value at the i-th point, Provides the SCADA interpolation supplementary value for the i-th point; The step of further identifying bad data sequences using the feature smoothing line verification method includes: Step 51. For the broadband data sample set U, calculate the maximum and minimum values ​​of the broadband signal data at the same time: （12）； In the above formula, X max X represents the maximum value of the broadband signal data at the same time. min The minimum value of broadband signal data at the same time. ... The maximum value of the collected data within the same period. ... This represents the minimum value of the collected data within the same period. Step 52. Extract the feature curves through comparison: （13）； In the above formula, X d For the characteristic curve, X d1…… X d200 These are the 200 values ​​that constitute the characteristic curve; Step 53. Calculate the upper and lower limits of the historical data domain using the maximum and minimum values ​​of current, voltage, or harmonics from historical data: （14）； In the above formula, , All are dimensionless values ​​representing allowable variation thresholds. For x d The i-th value, where i is the i-th point; Step 54. Calculate the upper and lower limits of the feasible region for current, voltage, or harmonic data: （15）； In the above formula, This represents the upper limit of the feasible region. K is the lower bound of the feasible region. t For the threshold coefficient that allows for variation, and All are dimensionless values ​​of the allowable change threshold, where i is the i-th point; Using the calculated upper and lower limits of the feasible region, a feasible region matrix is ​​formed for current, voltage, or harmonic data. The feasible region matrix for broadband signals is: （16）； Each column element of the feasible region matrix [ , ] represents the and All are [-K1,+K i ] dimensionless values ​​between; in anomaly data identification, for the data to be tested ,when The i-th component satisfies ≤ At that time, calculate the upper limit of the data to be inspected: （17）； In the above formula, x represents the upper limit of the data to be inspected. new，i For the i-th component; when The i-th component satisfies ≥ At that time, calculate the lower limit of the data to be inspected: （18）。 5. A computer device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the broadband anomaly identification and detection method with dual-source verification as described in any one of claims 1-3.

6. A computer storage medium, characterized in that: The computer storage medium contains a computer program, which, when executed by a processor, implements the steps of the broadband anomaly identification and detection method with dual-source verification as described in any one of claims 1-3.

Images