CN107966638A - Method and apparatus, storage medium and the processor of correction error - Google Patents

Abstract

The invention discloses a method and device for correcting errors, a storage medium and a processor. Among them, the method includes: obtaining the measured value for locating the partial discharge source, and the radial distance and azimuth angle of the above measured value; using the first model to analyze the radial distance and azimuth angle of the above measured value, and obtaining the The error value of the power supply, wherein the above-mentioned first model is obtained by using multiple sets of data through machine learning training, and each set of data in the above-mentioned multiple sets of data includes: the radial distance and azimuth angle of the above-mentioned measured value, and the above-mentioned error value; The above-mentioned measured value is corrected according to the above-mentioned error value. The invention solves the technical problem in the prior art that the error of partial discharge location cannot be corrected, resulting in low accuracy of time delay estimation.

Description

Error correction method and device, storage medium and processor

technical field

The invention relates to the field of power failure processing, in particular to a method and device for correcting errors, a storage medium and a processor.

Background technique

Partial discharge (Partial Discharge) is a manifestation of the deterioration of the insulation performance of power equipment, and it is also the cause of further deterioration of the insulation performance. Monitoring partial discharge signals is an important means to discover equipment insulation defects in time and prevent insulation breakdown failures. The UHF electromagnetic wave positioning method has become the most potential full-station partial discharge positioning method because of its characteristics of total station partial discharge positioning, high sensitivity and suitable for online monitoring.

In recent years, the partial discharge positioning method of UHF sensor array (abbreviated as UHF method) has aroused extensive research by scholars at home and abroad. The research results show that time delay estimation is the key to UHF method for partial discharge positioning. Algorithms include threshold method, correlation estimation method, and energy accumulation method. However, the UHF method produces positioning errors due to its own limitations, and the accuracy of PD positioning is low. The experiment found that the size of the positioning error has the characteristics of increasing with the increase of the distance from the PD source, and the increase of the positioning error has a nonlinear relationship with the increase of the distance from the PD source. Therefore, it is necessary to correct the UHF partial discharge location error to improve the accuracy of time delay estimation.

However, the current method to improve the accuracy of time delay estimation is to use an antenna array with better performance and an oscilloscope with a higher sampling rate in hardware, and to improve the time delay estimation algorithm in software. The improvement of hardware performance will undoubtedly greatly increase the cost of the product, and the cost performance is low. Scholars usually start with improving the software. At present, the commonly used delay estimation algorithms include threshold method, generalized correlation method and energy accumulation method. The threshold method and the energy accumulation method have higher requirements on the signal-to-noise ratio of the signal; the correlation estimation method requires that the signal and noise, noise and noise are not correlated with each other, and the time delay estimation ability for non-stationary signals is poor. The improved correlation estimation method is the generalized Weighted correlation estimation method, this algorithm needs to know the statistical prior knowledge of the noise signal in advance, and it is still greatly affected by the noise signal. However, there are many sources of electromagnetic interference on the substation site, and the signal-to-noise ratio of the collected electromagnetic wave signal is not high. It is difficult for the noise reduction algorithm to eliminate the noise interference. The existing delay estimation algorithm is more sensitive to the noise signal, which leads to the low accuracy of the delay estimation. .

Aiming at the above-mentioned problem in the prior art that errors in partial discharge positioning cannot be corrected, resulting in low accuracy of time delay estimation, no effective solution has been proposed so far.

Contents of the invention

Embodiments of the present invention provide a method and device for correcting errors, a storage medium, and a processor, so as to at least solve the technical problem in the prior art that errors in partial discharge positioning cannot be corrected, resulting in low accuracy of time delay estimation.

According to an aspect of an embodiment of the present invention, there is provided a method for correcting errors, including: acquiring measured values for locating a partial discharge source, and the radial distance and azimuth angle of the measured values; Analyze the radial distance and azimuth angle to obtain the error value for locating the above-mentioned partial discharge source, wherein the above-mentioned first model is obtained by using multiple sets of data through machine learning training, and each set of data in the above-mentioned multiple sets of data includes: the above-mentioned The radial distance and azimuth angle of the measured value, the above-mentioned error value; the above-mentioned measured value is corrected according to the above-mentioned error value.

Further, obtaining the measured value for locating the partial discharge source, and the radial distance and azimuth angle of the measured value include: determining at least one measuring point between the first sensor and the second sensor in the partial discharge simulation device; The coordinate position of a measurement point is used to obtain the measured value of the partial discharge source, wherein the measured value at least includes: radial distance and azimuth angle.

Further, the above-mentioned first model is determined at least in the following manner: Based on the BP neural network, the relationship formula between the above-mentioned error value and the radial distance and azimuth angle of the above-mentioned measured value is determined: Wherein, ΔR is the above-mentioned error value, r is the radial distance of the above-mentioned measured value, and θ is the azimuth angle of the above-mentioned measured value; the above-mentioned relational formula is used as the above-mentioned first model.

Further, correcting the above-mentioned measurement value according to the above-mentioned error value includes: determining the compensation function of the above-mentioned measurement value according to the above-mentioned error value, wherein the above-mentioned error value is determined according to the above-mentioned measurement value and a predetermined standard value, and the above-mentioned compensation function is calculated by the following formula : r' is the radial distance of the above-mentioned predetermined standard value; θ' is the azimuth angle of the above-mentioned predetermined standard value; the above-mentioned compensation function at least includes: a compensation function of the radial distance of the above-mentioned measured value, and a compensation function of the azimuth angle of the above-mentioned measured value; △r=k ₁ (r, θ) is the compensation function of the radial distance of the above measured value; △θ=k ₂ (r, θ) is the compensation function of the azimuth angle of the above measured value, k ₁ and k ₂ are constants ; Use the above compensation function to correct the above measured value.

According to another aspect of the embodiments of the present invention, there is also provided a device for correcting errors, including: an acquisition module, used to acquire the measured value for locating the partial discharge source, and the radial distance and azimuth angle of the measured value; an analysis module , for using the first model to analyze the radial distance and azimuth angle of the above-mentioned measurement value to obtain the error value for locating the above-mentioned partial discharge source, wherein the above-mentioned first model is obtained by using multiple sets of data through machine learning training, and the above-mentioned Each set of data in the multiple sets of data includes: the radial distance and azimuth angle of the above-mentioned measured value, and the above-mentioned error value; a correction module is used to correct the above-mentioned measured value according to the above-mentioned error value.

Further, the acquisition module includes: a determination submodule, used to determine at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device; a positioning submodule, used to locate the coordinates of the above at least one measurement point position, to obtain the measured value of the partial discharge source, wherein the measured value at least includes: radial distance and azimuth angle.

Further, at least the above-mentioned first model is determined by the following module: the first determination module is used to determine the relationship formula between the above-mentioned error value and the radial distance and azimuth angle of the above-mentioned measurement value based on the BP neural network: Wherein, ΔR is the above-mentioned error value, r is the radial distance of the above-mentioned measurement value, and θ is the azimuth angle of the above-mentioned measurement value; the second determination module is used to use the above-mentioned relational formula as the above-mentioned first model.

Further, the above-mentioned correction module includes: a compensation determination sub-module for determining a compensation function of the above-mentioned measurement value according to the above-mentioned error value, wherein the above-mentioned error value is determined according to the above-mentioned measurement value and a predetermined standard value, and the above-mentioned compensation function is calculated by the following formula : r' is the radial distance of the above-mentioned predetermined standard value; θ' is the azimuth angle of the above-mentioned predetermined standard value; the above-mentioned compensation function at least includes: a compensation function of the radial distance of the above-mentioned measured value, and a compensation function of the azimuth angle of the above-mentioned measured value; △r=k ₁ (r, θ) is the compensation function of the radial distance of the above measured value; △θ=k ₂ (r, θ) is the compensation function of the azimuth angle of the above measured value, k ₁ and k ₂ are constants ; A calibration sub-module, configured to correct the above-mentioned measurement value by using the above-mentioned compensation function.

According to another aspect of the embodiments of the present invention, there is also provided a storage medium, where the above storage medium includes a stored program, wherein the above program executes any one of the above methods for correcting errors.

According to still another aspect of the embodiments of the present invention, there is also provided a processor, the above-mentioned processor is used for running a program, wherein, when the above-mentioned program is running, any one of the above-mentioned methods for correcting errors is executed.

In the embodiment of the present invention, by obtaining the measured value of locating the partial discharge source, and the radial distance and azimuth angle of the above measured value; using the first model to analyze the radial distance and azimuth angle of the above measured value, the above-mentioned The error value of the partial discharge source, wherein the above-mentioned first model is obtained by using multiple sets of data through machine learning training, and each set of data in the above-mentioned multiple sets of data includes: the radial distance and azimuth angle of the above-mentioned measured value, the above-mentioned error value; the above-mentioned measurement value is corrected according to the above-mentioned error value, and the purpose of effectively correcting the error of partial discharge location is achieved, thereby achieving the technical effect of improving the accuracy of time delay estimation, and further solving the problem that the local discharge cannot be corrected in the prior art The error of discharge positioning is corrected, which leads to the technical problem of low accuracy of time delay estimation.

Description of drawings

The accompanying drawings described here are used to provide a further understanding of the present invention and constitute a part of the application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations to the present invention. In the attached picture:

Fig. 1 is a schematic diagram of a BP neural network model according to an embodiment of the present invention;

FIG. 2 is a flowchart of steps of a method for correcting errors according to an embodiment of the present invention;

FIG. 3 is a flowchart of steps of an optional error correction method according to an embodiment of the present invention;

Fig. 4a is a schematic diagram of an optional calibration point position for calibration based on actual coordinates according to an embodiment of the present invention;

Fig. 4b is a schematic diagram of an optional calibration point position after spline interpolation according to an embodiment of the present invention;

Fig. 5 is an optional fitting curve diagram based on the cubic spline interpolation method according to an embodiment of the present invention;

Fig. 6a is an optional relationship diagram of the positioning error of the radial distance with r after spline interpolation according to an embodiment of the present invention;

Fig. 6b is an optional relationship diagram of the positioning error value of the azimuth angle after spline interpolation according to an embodiment of the present invention;

Fig. 7a is a graph showing the variation law of the measurement error of an optional radial distance r with r before and after correction according to an embodiment of the present invention;

Fig. 7b is a graph showing the change law of the measurement error of an optional azimuth angle θ with r before and after correction according to an embodiment of the present invention; and

Fig. 8 is a schematic structural diagram of an error correction device according to an embodiment of the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the solutions of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only It is an embodiment of a part of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

It should be noted that the terms "first" and "second" in the description and claims of the present invention and the above drawings are used to distinguish similar objects, but not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used are interchangeable under appropriate circumstances such that the embodiments of the invention described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having", as well as any variations thereof, are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or device comprising a sequence of steps or elements is not necessarily limited to the expressly listed instead, may include other steps or elements not explicitly listed or inherent to the process, method, product or apparatus.

First of all, for the convenience of understanding the embodiments of the present invention, some terms or nouns involved in the present invention will be explained below:

BP neural network: refers to a multi-layer artificial neural network that uses the Error Back-propagation Algorithm (BP).

Example 1

According to an embodiment of the present invention, an embodiment of a method for correcting errors is provided. It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and, although A logical order is shown in the flowcharts, but in some cases the steps shown or described may be performed in an order different from that shown or described herein.

It should be noted that before introducing each optional or preferred embodiment of the present application, the basic principle of the BP neural network can be clarified first:

BP neural network is a network with multiple hidden layers, which has the ability to deal with linear inseparable problems, and realizes the learning of unknown systems through the forward propagation of signals and the reverse adjustment mechanism of errors. The linear mapping ability can approximate any continuous nonlinear function with arbitrary precision, which is the proven universal approximation theorem. At the same time, it has strong fault-tolerant ability, and has self-learning and self-adaptive ability for uncertain complex problems, which lays the foundation for the application of BP neural network in nonlinear systems.

In an optional embodiment, the network structure of the BP neural network is as shown in Figure 1, for a node with N input nodes (X1, X2, X3...XN), the hidden layer has q neurons ( The corresponding output is P1, P2, P3...Pq), the BP neural network of L output nodes (the corresponding output is y1, y2, y3...yL), and the input-output relationship of the network is:

Among them, θ is the ideal output, and ω is the corresponding weight coefficient.

The application of BP neural network needs to pay attention to the following issues:

(1) Determine the number of network layers: BP neural network is composed of input layer---hidden layer---output layer, in which the number of hidden layers can be adjusted. It has been proved in theory that the continuous function of any closed interval All can be approximated by the BP neural network of a single hidden layer. Therefore, in most application scenarios, a single hidden layer network can realize its powerful nonlinear mapping function by adjusting the number of neurons in the hidden layer.

(2) Determine the number of input nodes: the number of input nodes depends on the dimension of the input vector. The input vector of this system is composed of the radial distance error value Δr and the direction angle error value Δθ, so the number of input nodes is 2.

(3) Determine the number of hidden layer nodes: when the number of network layers has been determined, the performance of BP neural network depends largely on the number of hidden layer neurons, the choice of the number of hidden layer neurons is It is a very important and complicated problem. Currently, there is no clear analytical formula to determine the optimal number of neuron nodes in the hidden layer. Usually, the approximate number of hidden layer nodes is determined according to the following three reference formulas:

1) M=log ₂ n, n is the number of input nodes.

2)

Among them, n is the number of input layer nodes, K is the number of samples, and M is the number of hidden layer nodes. It is stipulated that when i>M, CMi=0.

3)

Where a is a constant between [0,10], n is the number of neurons in the input layer, and m is the number of neurons in the output layer.

(4) Determine the number of output layer neurons: the number of output layer neurons will be determined according to the actual problem to be solved, the present invention is to do nonlinear mapping processing, and each radial distance and direction angle are uniquely mapped to an error value , so the number of neurons in the output layer of the present invention is 1.

(5) Selection of training method: The training method of BP neural network is affected by the actual problem and the number of samples. The RPROP algorithm has better applicability to the pattern recognition problem. When , the efficiency is higher, the convergence speed is faster and the mean square error is smaller. The present invention finally selects the LM algorithm as the training method by comparing various methods.

Fig. 1 is a flow chart of the steps of a method for correcting errors according to an embodiment of the present invention. As shown in Fig. 1, the method includes the following steps:

Step S102, obtaining the measured value for locating the partial discharge source, and the radial distance and azimuth angle of the measured value;

Step S104, using the first model to analyze the radial distance and azimuth angle of the above-mentioned measured value to obtain the error value for locating the above-mentioned partial discharge source, wherein the above-mentioned first model is obtained by using multiple sets of data through machine learning training, and the above-mentioned Each set of data in the multiple sets of data includes: the radial distance and azimuth angle of the above-mentioned measured values, and the above-mentioned error value;

Step S106, correcting the measurement value according to the error value.

In an optional implementation manner, the above-mentioned first model may be a compensation surface determined based on a BP neural network.

As an optional implementation, the application can pre-store the radial distance, azimuth angle and error value of multiple measured values in the database or server, so that the above-mentioned can be obtained through machine learning training based on BP neural network using multiple sets of data first model.

Among them, the radial distance and azimuth angle of the measured value and the error value can be input into the neural network to establish corresponding neurons and determine image features and image characteristics according to a preset function (such as the Sigmoid function) between neurons Feature mapping, so as to perform mapping processing according to the determined features, and obtain an error value for locating the above-mentioned partial discharge source.

In addition, in the optional embodiment provided by the present application, after the above first model is established, deep learning or KNN algorithm can be used to filter out the measured value information with the same feature in the image information, so as to obtain the difference error Value characteristic information, and then obtain the error value of locating the above partial discharge source.

This application uses the BP neural network learning method to correct the UHF partial discharge positioning error. In an optional embodiment, this application can pre-think or set the inherent error of the positioning system, and use the BP neural network to be highly fault-tolerant And the ability to infinitely approximate the nonlinear function, identify and reproduce the inherent error of the positioning system, and then compensate the actual measured value, use the K-Means machine learning algorithm to de-jitter the measured value, through the limited calibration points (radial distance r, azimuth θ) error value training, the error compensation surface of 0<r≤66m, 0<θ≤360° is constructed, and then the measured value is corrected.

It should be noted that the BP neural network has a very strong nonlinear mapping ability, and can learn the unknown system at the same time, and because of the redundancy of the parallel mechanism, it has a strong fault tolerance.

In order to improve the accuracy of UHF substation localization positioning, this application proposes a positioning error correction method based on BP neural network, using the K-Means machine learning algorithm to de-jitter the measured value, and through a limited number of calibration points ( For the training of the error value of radial distance r, azimuth angle θ), an error compensation surface of 0<r≤66m, 0<θ≤360° is constructed, and then the measured value is corrected. Considering the engineering application factors, a method of obtaining samples based on spline interpolation is proposed, which reduces the workload and has engineering operability. The experimental results show that in the range of r∈(0,40m), the radial distance measurement error shows a decreasing trend, and in the range of r>40m, the radial distance measurement error shows an increasing trend, and the distribution of the direction angle measurement error has a similar law. After the error compensation, the error distribution tends to be stable and the error value is greatly reduced, which proves the feasibility and effectiveness of the error correction method provided by the present application.

It should be noted that the reasons for the above errors can be as follows: the UHF signal sensing antenna is a key part of the UHF detection system, and its function is to realize the acquisition of partial discharge signals, and the performance of the sensing antenna directly determines The performance of the entire detection system, whether it has the ability to suppress low-frequency interference (below 200MHZ), whether it has high detection sensitivity, how is the omnidirectional performance, and the matching characteristics within the ultra-frequency bandwidth are the key to measuring the performance of the sensing antenna; band-pass filtering The filter performance of the filter determines the purity of the collected signal. The input protection performance of the UHF amplifier has an important impact on the reliability of the system. The characteristics of the amplifier will also affect the sensitivity of the detection; The highest precision of delay estimation, the minimum error of delay estimation is one sampling time interval. Therefore, the performance of the system hardware directly affects the positioning accuracy. In addition, in addition to the above-mentioned main reasons, the following reasons are also included: electromagnetic noise interference and electromagnetic wave propagation attenuation, limitations of time delay estimation algorithms, and the like.

In an optional embodiment, FIG. 2 is a flow chart of the steps of an optional error correction method according to an embodiment of the present invention. As shown in FIG. 2, obtaining the measured value for locating the partial discharge source includes:

Step S202, determining at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device;

Step S204, by locating the coordinate position of the at least one measurement point, to obtain the measurement value of the partial discharge source, wherein the measurement value at least includes: radial distance and azimuth angle.

In an optional embodiment, there is at least one measurement point of the partial discharge source between the first sensor and the second sensor.

Based on the embodiments provided in the above step S202 to step S204, at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device can be determined, and by locating the coordinate position of the above at least one measurement point, the above local The measured value of the discharge source, wherein the above measured value at least includes: radial distance and azimuth angle.

As an optional embodiment, the application can use a partial discharge simulation device (EM TEST DITO, partial discharge simulator) to carry out a partial discharge simulation experiment, which can produce accurate Discharge pulse, optional, the direction angle of the vertical line direction of the connection between the first sensor and the second sensor is 0 degrees, and the position coordinates of the measuring points are calibrated with a cloth tape measure with a total length of 100 meters in the direction of 0 degrees, Mark a point every six meters between 0-66m and conduct positioning experiments.

In an optional embodiment, correcting the above-mentioned measurement value according to the above-mentioned error value includes: determining a compensation function of the above-mentioned measurement value according to the above-mentioned error value, wherein the above-mentioned error value is determined according to the above-mentioned measurement value and a predetermined standard value, and the above-mentioned The compensation function is calculated by the following formula: r' is the radial distance of the above-mentioned predetermined standard value; θ' is the azimuth angle of the above-mentioned predetermined standard value; the above-mentioned compensation function at least includes: a compensation function of the radial distance of the above-mentioned measured value, and a compensation function of the azimuth angle of the above-mentioned measured value; △r=k ₁ (r, θ) is the compensation function of the radial distance of the above measured value; △θ=k ₂ (r, θ) is the compensation function of the azimuth angle of the above measured value, k ₁ and k ₂ are constants ; Use the above compensation function to correct the above measured value.

In an optional embodiment, due to the large amount of environmental noise interference in the substation site, the measured value of the positioning system deviates from the predetermined standard value, and the plane position of the partial discharge source is uniquely determined by the radial distance r and the azimuth angle θ, It is necessary to correct the measured value of the partial discharge source position (radial distance r, azimuth angle θ), the core idea of the present invention is to admit the existence of the inherent error of the system, use the BP neural network to learn the distribution law of the error, and determine according to the above error value The compensation function of the above-mentioned measured value, wherein the above-mentioned error value is determined according to the above-mentioned measured value and a predetermined standard value, and the above-mentioned compensation function is calculated by the following formula: The above-mentioned measured values are corrected using the above-mentioned compensation function.

In addition, it should be noted that since the error compensation model in this application is a compensation surface whose error value varies with the radial distance r and azimuth angle θ of the above-mentioned measured values, this application uses BP neural network to construct the above-mentioned compensation surface, which cannot Find a specific analytical formula.

In the electromagnetic environment with low signal-to-noise ratio (large external interference signal), the measured position coordinates of the same discharge point (measurement point) fluctuate, which interferes with the selection of error samples and makes the obtained samples not typical, so The selection of error value samples in this application adopts the method based on K-Means clustering. Each measurement result needs to be measured multiple times, clustered by K-Means algorithm, and the cluster center is used as the final measurement value , and then compared with the predetermined standard value to obtain the error value sample.

As an optional embodiment, the above-mentioned first model is determined at least in the following manner: based on the BP neural network, the relationship formula between the above-mentioned error value and the radial distance and azimuth angle of the above-mentioned measured value is determined: Wherein, ΔR is the above-mentioned error value, r is the radial distance of the above-mentioned measured value, and θ is the azimuth angle of the above-mentioned measured value; the above-mentioned relational formula is used as the above-mentioned first model.

It should be noted that the core idea of this application is to use the BP neural network to learn the change law of the positioning error with the azimuth angle θ and the radial distance r. increase, the detection of weak signals of UHF electromagnetic waves is more difficult, and the positioning error tends to increase with the increase of radial distance, so it is necessary to conduct feasibility analysis experiments to verify the learning of BP neural network for complex uncertain random values Ability to illustrate the effectiveness of the BP neural network error correction scheme.

Based on the BP neural network, the relationship formula between the above-mentioned error value and the radial distance and azimuth angle of the above-mentioned measured value is preset: Among them, ΔR is the above error value, r=1,2,3...70; θ=15k, k=0,1,2...23.

Among them, r can increase from 1 to 70 with a step size of 3, but not limited to, θ can increase from 0 to 345° with a step size of 15°. When θ is different, that is, when UHF signal propagation paths are different, there must be delay calibration and positioning errors. Therefore, random numbers are used in the above relationship formula to represent this characteristic, so the value of Δr will be determined by θ and r. These values are now used as learning samples to train the BP neural network. By training the BP neural network with different structures, it is found that using 2-30-1 (that is, two inputs, one output, and 30 neurons in the hidden layer ) structure has the best effect. After the eighth training of the neural network, the root mean square error between the network output of the training sample and the expected output is 0.019, where 0019 is the value after normalization, and it is 0.22 after denormalization m.

Based on the above embodiments, it can be seen that in the embodiment of the present application, since the BP neural network has a high degree of fault tolerance, it can completely realize the simulation and reproduction of the randomly fluctuating curved surface, thus proving the effectiveness of the error correction method in the present application .

As an optional embodiment, due to the construction of an error compensation surface in the range of 0<r≤66m, 0<θ≤360°, a large number of samples are required to train the neural network. If the step size of r is 6m, the If the calibration point is selected as the training sample with a step length of 15°, then 264 points still need to be calibrated, as shown in Figure 4a. It is necessary to know the actual coordinates of these 264 points, and to use the analog discharger to discharge at these points, and then Obtaining the measured value of the UHF positioning system to be corrected requires a huge workload and poor operability. Considering the factors of engineering application, this application proposes a sample acquisition method based on spline interpolation. The idea of this method is More abundant training samples can be obtained by inserting several points between any two actual calibration points, thereby reducing the number of partial discharge simulation experiments to obtain error sample data, and greatly reducing the workload.

Moreover, as another optional embodiment, in substations with regular distribution of equipment, the selection of calibration points can be simplified, that is, it is considered that the azimuth angle θ has little influence on the distribution of errors. If the error value is assumed to be The distribution laws of different orientation angles are the same, so when selecting training samples, it is only necessary to measure the error value in one angle direction, as shown in Figure 4b, and then insert several points between any two calibration points by spline interpolation method, so that Get more informative training samples.

In the embodiment of this application, the simulation data can be generated by using the above relational formula, but not limited to, now take 11 data in the direction of θ = 0°, and the r step length is 6m, respectively -0.0653m, -0.6987m, 0.6556m . Insert another data point between the data points, and assume that the error value of the same r value is the same, you can get 504 (21×24) training samples, this sample and the 504 data points directly generated by the above correction formula The root mean square error is 0.48m, and the error is acceptable, which proves the feasibility of the sample acquisition method. The experiment found that the root mean square error between the network output and the expected output is 2.95×10 ^-6 for the method of obtaining samples through interpolation. The expected output value at is the training sample of the neural network. It can be seen that the method of obtaining more training samples based on spline interpolation has high feasibility, and the error is within an acceptable range.

As an optional embodiment, the present application can use EM TEST DITO (partial discharge simulator) to carry out partial discharge simulation experiment, and this simulator can produce accurate discharge pulse according to EN/IEC 61000-4-2 standard, and this experiment It is stipulated that the direction angle of the vertical line direction of the connecting line of sensors 1 and 2 is 0 degrees, and the position coordinates of the measuring points are calibrated with a cloth tape measure with a total length of 100 meters in the direction of 0 degrees, and every six meters between 0-66m Calibrate a point and conduct a positioning experiment, the measurement results are shown in Table 1.

Table 1 Measurement Results

In an optional embodiment, according to the idea of spline interpolation, two data points can be inserted between every two adjacent error data points, and the relationship between the positioning error of the interpolated radial distance and the variation of r is as follows: Figure 6a, and then obtain the error sample data in a simplified manner based on the same error value when r is the same and θ is different, and train the neural network.

In a kind of optional embodiment, can utilize simulation discharger to carry out partial discharge simulation experiment, on the direction angle of θ=90 °, between 0-66m, carry out positioning experiment every six meters, utilize the method of the present invention Compensate the measurement value given by the system to obtain the final positioning result. The positioning measurement value and the data error after compensation surface compensation are shown in Table 2. The variation law of the measurement error of the radial distance r with r before and after correction is shown in Figure 7a, and the variation law of the measurement error of the azimuth angle θ with r before and after correction is shown in Figure 7b.

Table 2 Positioning data measurement value and error after compensation

From Fig. 7a to Fig. 7b, it can be seen that the error of the positioning measurement value of the azimuth angle and the positioning measurement value of the radial distance has a tendency to increase gradually with the increase of r before correction, and the overall positioning error value of the system is relatively large. After correction The change of the error value tends to be stable, and the overall error value is small. This proves the effectiveness and stability of the UHF partial discharge error correction method based on the BP neural network proposed in this application.

Example 2

An embodiment of the present invention also provides a device for implementing the above error correction method. FIG. 8 is a schematic structural diagram of an error correction device according to an embodiment of the present invention. As shown in FIG. 8, the above error correction device , including: acquisition module 90, analysis module 92 and correction module 94, wherein,

The obtaining module 90 is used to obtain the measured value of locating the partial discharge source, as well as the radial distance and azimuth angle of the above-mentioned measured value; the analysis module 92 is used to analyze the radial distance and azimuth angle of the above-mentioned measured value by using the first model , to obtain the error value for locating the above-mentioned partial discharge source, wherein the above-mentioned first model is obtained by using multiple sets of data through machine learning training, and each set of data in the above-mentioned multiple sets of data includes: the radial distance and orientation of the above-mentioned measured values Angle, the above-mentioned error value; a correction module 94, configured to correct the above-mentioned measurement value according to the above-mentioned error value.

In an optional embodiment, the acquisition module includes: a determination submodule, configured to determine at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device; a positioning submodule, configured to locate The coordinate position of the at least one measurement point is used to obtain the measurement value of the partial discharge source, wherein the measurement value at least includes: radial distance and azimuth angle.

In an optional embodiment, the above-mentioned first model is determined by at least the following module: the first determination module is used to determine the relationship formula between the above-mentioned error value and the radial distance and azimuth angle of the above-mentioned measured value based on the BP neural network : Wherein, ΔR is the above-mentioned error value, r is the radial distance of the above-mentioned measurement value, and θ is the azimuth angle of the above-mentioned measurement value; the second determination module is used to use the above-mentioned relational formula as the above-mentioned first model.

In an optional embodiment, the correction module includes: a compensation determination submodule, configured to determine the compensation function of the measurement value according to the error value, wherein the error value is determined according to the measurement value and a predetermined standard value, and the above The compensation function is calculated by the following formula r' is the radial distance of the above-mentioned predetermined standard value; θ' is the azimuth angle of the above-mentioned predetermined standard value; the above-mentioned compensation function at least includes: a compensation function of the radial distance of the above-mentioned measured value, and a compensation function of the azimuth angle of the above-mentioned measured value; △r=k ₁ (r, θ) is the compensation function of the radial distance of the above measured value; △θ=k ₂ (r, θ) is the compensation function of the azimuth angle of the above measured value, k ₁ and k ₂ are constants ; A calibration sub-module, configured to correct the above-mentioned measurement value by using the above-mentioned compensation function.

It should be noted that each of the above-mentioned modules can be realized by software or hardware. For example, for the latter, it can be realized in the following manner: each of the above-mentioned modules can be located in the same processor; or, each of the above-mentioned modules can be implemented in any combination on a different processor.

It should be noted here that the acquisition module 90, the analysis module 92 and the correction module 94 correspond to steps S102 to S106 in Embodiment 1, and the examples and application scenarios implemented by the above modules are the same as those of the corresponding steps, but are not limited to The content disclosed in the above-mentioned embodiment 1. It should be noted that, as a part of the device, the above modules can run in the computer terminal.

It should be noted that, for optional or preferred implementation manners of this embodiment, reference may be made to relevant descriptions in Embodiment 1, and details are not repeated here.

The above-mentioned device for correcting errors may also include a processor and a memory. The above-mentioned acquisition module 90, analysis module 92, and correction module 94 are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory. corresponding function.

The processor includes a kernel, and the kernel fetches corresponding program units from the memory, and one or more kernels can be provided. Memory may include non-permanent memory in computer-readable media, random access memory (RAM) and/or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM), memory includes at least one memory chip.

The embodiment of the present application also provides a storage medium. Optionally, in this embodiment, the above-mentioned storage medium includes a stored program, wherein when the above-mentioned program is running, the device where the above-mentioned storage medium is located is controlled to execute any one of the above-mentioned methods for correcting errors.

Optionally, in this embodiment, the above-mentioned storage medium may be located in any computer terminal in the group of computer terminals in the computer network, or in any mobile terminal in the group of mobile terminals.

The embodiment of the present application also provides a processor. Optionally, in this embodiment, the above-mentioned processor is configured to run a program, wherein, when the above-mentioned program is running, any one of the above-mentioned methods for correcting errors is executed.

An embodiment of the present application provides a device, which includes a processor, a memory, and a program stored in the memory and operable on the processor. When the processor executes the program, the following steps are implemented: acquiring a measured value for locating a partial discharge source, and The radial distance and azimuth angle of the above-mentioned measured value; using the first model to analyze the radial distance and azimuth angle of the above-mentioned measured value to obtain the error value for locating the above-mentioned partial discharge source, wherein the above-mentioned first model uses multiple sets of data Obtained through machine learning training, each of the above multiple sets of data includes: the radial distance and azimuth angle of the above-mentioned measured value, and the above-mentioned error value; the above-mentioned measured value is corrected according to the above-mentioned error value.

Optionally, when the above-mentioned processor executes the program, it can also determine at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device; by locating the coordinate position of the above-mentioned at least one measurement point, the above-mentioned partial discharge source can be obtained The measured values, wherein the measured values at least include: radial distance and azimuth angle.

Optionally, when the above-mentioned processor executes the program, it can also determine the relationship formula between the above-mentioned error value and the radial distance and azimuth angle of the above-mentioned measurement value based on the BP neural network: Wherein, ΔR is the above-mentioned error value, r is the radial distance of the above-mentioned measured value, and θ is the azimuth angle of the above-mentioned measured value; the above-mentioned relational formula is used as the above-mentioned first model.

Optionally, when the above-mentioned processor executes the program, the compensation function of the above-mentioned measurement value can also be determined according to the above-mentioned error value, wherein the above-mentioned error value is determined according to the above-mentioned measurement value and a predetermined standard value, and the above-mentioned compensation function is calculated by the following formula: r' is the radial distance of the above-mentioned predetermined standard value; θ' is the azimuth angle of the above-mentioned predetermined standard value; the above-mentioned compensation function at least includes: a compensation function of the radial distance of the above-mentioned measured value, and a compensation function of the azimuth angle of the above-mentioned measured value; △r=k ₁ (r, θ) is the compensation function of the radial distance of the above measured value; △θ=k ₂ (r, θ) is the compensation function of the azimuth angle of the above measured value, k ₁ and k ₂ are constants ; Use the above compensation function to correct the above measured value.

The present application also provides a computer program product adapted, when executed on a data processing device, to execute a program initialized with the method steps of: obtaining measurements for locating a partial discharge source, and the radial distance and orientation of said measurements angle; use the first model to analyze the radial distance and azimuth angle of the above-mentioned measured value, and obtain the error value of locating the above-mentioned partial discharge source, wherein, the above-mentioned first model is obtained by using multiple sets of data through machine learning training, and the above-mentioned multiple Each set of data in the set of data includes: the radial distance and azimuth angle of the above-mentioned measured value, and the above-mentioned error value; the above-mentioned measured value is corrected according to the above-mentioned error value.

Optionally, when the above-mentioned computer program product executes the program, at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device can also be determined; by locating the coordinate position of the above-mentioned at least one measurement point, the above-mentioned partial discharge The measurement value of the power supply, wherein the above measurement value at least includes: radial distance and azimuth angle.

Optionally, when the above-mentioned computer program product executes the program, it can also determine the relationship formula between the above-mentioned error value and the radial distance and azimuth angle of the above-mentioned measurement value based on the BP neural network: Wherein, ΔR is the above-mentioned error value, r is the radial distance of the above-mentioned measured value, and θ is the azimuth angle of the above-mentioned measured value; the above-mentioned relational formula is used as the above-mentioned first model.

Optionally, when the above-mentioned computer program product executes the program, the compensation function of the above-mentioned measured value may also be determined according to the above-mentioned error value, wherein the above-mentioned error value is determined according to the above-mentioned measured value and a predetermined standard value, and the above-mentioned compensation function is calculated by the following formula: r' is the radial distance of the above-mentioned predetermined standard value; θ' is the azimuth angle of the above-mentioned predetermined standard value; the above-mentioned compensation function at least includes: a compensation function of the radial distance of the above-mentioned measured value, and a compensation function of the azimuth angle of the above-mentioned measured value; △r=k ₁ (r, θ) is the compensation function of the radial distance of the above measured value; △θ=k ₂ (r, θ) is the compensation function of the azimuth angle of the above measured value, k ₁ and k ₂ are constants ; Use the above compensation function to correct the above measured value.

The serial numbers of the above embodiments of the present invention are for description only, and do not represent the advantages and disadvantages of the embodiments.

In the above-mentioned embodiments of the present invention, the descriptions of each embodiment have their own emphases, and for parts not described in detail in a certain embodiment, reference may be made to relevant descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed technical content can be realized in other ways. Wherein, the device embodiments described above are only illustrative. For example, the division of the units may be a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or may be Integrate into another system, or some features may be ignored, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of units or modules may be in electrical or other forms.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, each unit may exist separately physically, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

If the integrated unit is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the essence of the technical solution of the present invention or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes. .

The above is only a preferred embodiment of the present invention, it should be pointed out that, for those of ordinary skill in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, and these improvements and modifications can also be made. It should be regarded as the protection scope of the present invention.

Claims (10)

1. A method for correcting errors, characterized in that, comprising: Obtaining measurements for locating partial discharge sources, and radial distances and azimuths of said measurements; Use the first model to analyze the radial distance and azimuth angle of the measured value to obtain the error value for locating the partial discharge source, wherein the first model is obtained through machine learning training using multiple sets of data, so Each set of data in the multiple sets of data includes: the radial distance and azimuth angle of the measured value, the error value; The measured value is corrected according to the error value. 2. The method according to claim 1, wherein obtaining the measured value for locating the partial discharge source, and the radial distance and azimuth angle of the measured value comprise: determining at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device; By locating the coordinate position of the at least one measurement point, the measurement value of the partial discharge source is obtained, wherein the measurement value at least includes: a radial distance and an azimuth angle. 3. The method according to claim 1, wherein the first model is determined at least in the following manner: Based on the BP neural network, determine the relational formula of the radial distance and the azimuth angle of the error value and the measured value: Wherein, ΔR is the error value, r is the radial distance of the measured value, and θ is the azimuth angle of the measured value; The relational formula is used as the first model. 4. The method according to claim 1, wherein correcting the measured value according to the error value comprises: The compensation function of the measured value is determined according to the error value, wherein the error value is determined according to the measured value and a predetermined standard value, and the compensation function is calculated by the following formula: <mrow><mfenced open = "{" close = ""><mtable><mtr><mtd><msup><mi>r</mi><mo>&amp;prime;</mo></msup><mo>=</mo><mi>r</mi><mo>-</mo><msub><mi>k</mi><mn>1</mn></msub><mo>(</mo><mi>r</mi><mo>,</mo><mi>&amp;theta;</mi><mo>)</mo></mtd></mtr><mtr><mtd><msup><mi>&amp;theta;</mi><mo>&amp;prime;</mo></msup><mo>=</mo><mi>&amp;theta;</mi>mi><mo>-</mo><msub><mi>k</mi><mn>2</mn></msub><mo>(</mo><mi>r</mi><mo>,</mo><mi>&amp;theta;</mi><mo>)</mo></mtd></mtr></mtable></mfenced><mo>;</mo></mrow> r' is the radial distance of the predetermined standard value; θ' is the azimuth angle of the predetermined standard value; the compensation function at least includes: the compensation function of the radial distance of the measured value, the azimuth of the measured value Angle compensation function; Δr=k ₁ (r, θ) is the compensation function of the radial distance of the measured value; Δθ=k ₂ (r, θ) is the compensation function of the azimuth angle of the measured value, k ₁ and k ₂ are constants; The measured values are corrected using the compensation function. 5. A device for correcting errors, characterized in that it comprises: An acquisition module, configured to acquire the measured value for locating the partial discharge source, and the radial distance and azimuth angle of the measured value; An analysis module, configured to use a first model to analyze the radial distance and azimuth angle of the measured value to obtain an error value for locating the partial discharge source, wherein the first model uses multiple sets of data through machine learning Obtained by training, each set of data in the multiple sets of data includes: the radial distance and azimuth angle of the measured value, and the error value; A correction module, configured to correct the measurement value according to the error value. 6. The device according to claim 5, wherein the acquisition module comprises: A determining submodule, configured to determine at least one measurement point between the first sensor and the second sensor in the partial discharge simulation device; The positioning sub-module is configured to obtain the measurement value of the partial discharge source by locating the coordinate position of the at least one measurement point, wherein the measurement value at least includes: a radial distance and an azimuth angle. 7. The device according to claim 5, wherein the first model is determined at least by the following modules: The first determining module is used to determine the relational formula of the radial distance and the azimuth angle of the error value and the measured value based on the BP neural network: Wherein, ΔR is the error value, r is the radial distance of the measured value, and θ is the azimuth angle of the measured value; The second determining module is configured to use the relational formula as the first model. 8. The device according to claim 5, wherein the correction module comprises: The compensation determining submodule is used to determine the compensation function of the measured value according to the error value, wherein the error value is determined according to the measured value and a predetermined standard value, and the compensation function is calculated by the following formula: <mrow><mfenced open = "{" close = ""><mtable><mtr><mtd><msup><mi>r</mi><mo>&amp;prime;</mo></msup><mo>=</mo><mi>r</mi><mo>-</mo><msub><mi>k</mi><mn>1</mn></msub><mo>(</mo><mi>r</mi><mo>,</mo><mi>&amp;theta;</mi><mo>)</mo></mtd></mtr><mtr><mtd><msup><mi>&amp;theta;</mi><mo>&amp;prime;</mo></msup><mo>=</mo><mi>&amp;theta;</mi>mi><mo>-</mo><msub><mi>k</mi><mn>2</mn></msub><mo>(</mo><mi>r</mi><mo>,</mo><mi>&amp;theta;</mi><mo>)</mo></mtd></mtr></mtable></mfenced><mo>;</mo></mrow> r' is the radial distance of the predetermined standard value; θ' is the azimuth angle of the predetermined standard value; the compensation function at least includes: the compensation function of the radial distance of the measured value, the azimuth of the measured value Angle compensation function; Δr=k ₁ (r, θ) is the compensation function of the radial distance of the measured value; Δθ=k ₂ (r, θ) is the compensation function of the azimuth angle of the measured value, k ₁ and k ₂ are constants; A corrector module, configured to correct the measured value by using the compensation function. 9. A storage medium, characterized in that the storage medium includes a stored program, wherein the program executes the error correction method according to any one of claims 1 to 4. 10. A processor, characterized in that the processor is used to run a program, wherein the error correction method according to any one of claims 1 to 4 is executed when the program is running.