CN117169657A - High-voltage cable state monitoring method and system based on artificial intelligence - Google Patents

Abstract

The present invention provides a high-voltage cable status monitoring method and system based on artificial intelligence, which includes: the first step is to draw a power grid topology map and collect the historical voltage signals, current signals, and temperatures of each detection point of each phase of the power grid and each reporting period. Data, determine the higher harmonic ratio and temperature increase value; change the thickness of the line segments in the topology map according to the temperature increase value, color the topology map according to the higher harmonic ratio, and annotate the image anomalies; use historical monitoring image sets to conduct Neural network training collects real-time data of the current cycle to obtain the current monitoring image; inputs the current monitoring image to the first model to detect whether there is an abnormality. Through the above solution, it is possible to monitor multi-branch power grids without establishing complex mathematical models.

Description

A high-voltage cable condition monitoring method and system based on artificial intelligence

Technical field

The present invention relates to the field of high-voltage cable status monitoring, and specifically to a high-voltage cable status monitoring method and system based on artificial intelligence.

Background technique

High-voltage cable condition monitoring refers to the real-time monitoring of high-voltage cables used in power systems to ensure their normal operation and safety performance. High-voltage cables are subjected to high voltages and large currents during the transmission of electrical energy, so monitoring their status is crucial to preventing failures and ensuring stable operation of the power grid.

With the accumulation of factors such as aging and damage of high-voltage cable insulation materials, excessive electric field intensity, and tip corona, partial discharge problems are prone to occur. Partial discharges can generate current within insulating materials, causing arcing and discharge phenomena. Partial discharges are small-scale electrical discharges in cable insulation materials that may be an indication of underlying faults. By monitoring partial discharge, cable insulation problems can be discovered in advance and prevented from developing into large-scale failures.

In the prior art, partial discharge sensors are usually used for partial discharge monitoring. These sensors are installed on cables and used to capture signals generated by partial discharges. Sensors usually include electromagnetic sensors, capacitive sensors, ultrasonic sensors, etc., which are used to detect electric fields, voltages, currents, sounds and other signals. After collecting the corresponding signals, mathematical methods are used to perform high-frequency current transformer analysis, spectrum analysis, etc. on the information. to locate the specific fault point.

However, the methods in the prior art usually require complex mathematical processing, so they generally only analyze a single cable. For multi-branch multi-phase power, complex mathematical models need to be established. Before using the existing technology for analysis, manual roughening is required in advance. When locating to a single cable, it is difficult to directly use methods in the existing technology to quickly solve without human intervention.

Contents of the invention

In order to solve the problems in the prior art, the present invention provides an artificial intelligence-based high-voltage cable status monitoring method and system.

In one aspect of the present invention, an artificial intelligence-based high-voltage cable status monitoring method is provided, which is characterized in that the method includes the following steps: Step 1: Draw a topology map according to the actual distribution of a multi-branch power grid; Step 2: Collect the multi-branch power grid The historical voltage signal, current signal, and temperature data of each reporting period at each detection point of each phase of the branch power grid; Step 3: For the data in the same reporting period, determine the higher harmonic ratio based on the voltage signal and current signal; Determine the minimum temperature from the temperature data, and subtract the minimum temperature from other temperatures to obtain the temperature increase value; Step 4, change the thickness of the line segment in the topology diagram according to the temperature increase value, according to the respective higher harmonic ratios of the three-phase electricity Determine the RGB value, and color the topology map according to the determined RGB value to obtain a monitoring image; Step 5, if there is an abnormal state in the current cycle, mark the abnormal position on the monitoring image; Step 6, repeat step 3 Step 5: Process all historical data to obtain a historical monitoring image set; use the historical monitoring image set for neural network training to obtain the first model; Step 7: Collect real-time data of the current period; Process real-time data of the current period according to steps 3 to 5 , obtain the current monitoring image; input the current monitoring image to the first model to detect whether there is an abnormality.

Further, the temperature increase value changing the thickness of the line segment in the topology map includes: new line thickness = original line thickness * (1 + temperature increase value/5).

Further, linear interpolation is performed on the temperature between the two detection points.

Further, determining the RGB value based on the respective higher harmonic ratios of the three-phase power includes: using Min-Max scaling to convert the higher harmonic ratio of each phase of the three-phase power into data of 0-255, and The three converted data correspond to the three numbers in the RGB value.

Further, linear interpolation is performed on the pixel colors between the two detection points.

On the other hand, the present invention also provides a high-voltage cable status monitoring system based on artificial intelligence, which is characterized in that the system includes the following modules: a drawing module for drawing a topology map according to the actual distribution of a multi-branch power grid; a first collection module, Used to collect historical voltage signals, current signals, and temperature data for each reporting cycle of each phase of each detection point of the multi-branch power grid; the calculation module is used for data in the same reporting cycle, based on the voltage signal and current signal Determine the higher harmonic ratio; determine the minimum temperature according to the temperature data, and subtract the minimum temperature from other temperatures to obtain the temperature increase value; the image processing module is used to change the thickness of the line segment in the topology map according to the temperature increase value, according to The respective higher harmonic ratios of the three-phase electricity determine the RGB values, and the topology map is colored according to the determined RGB values to obtain the monitoring image; the annotation module is used to monitor if there is an abnormal state in the current period. Annotate abnormal positions on the image; the training module is used to run the calculation module, image processing module, and annotation module to process all historical data to obtain a historical monitoring image set; conduct neural network training with the historical monitoring image set to obtain the first model; detection Module, used to collect real-time data of the current period; process the real-time data of the current period according to steps three to five to obtain the current monitoring image; input the current monitoring image to the first model to detect whether there is an abnormality.

Further, changing the thickness of the line segment in the topology map according to the temperature increase value includes: new line thickness = original line thickness * (1 + temperature increase value / 5).

Further, linear interpolation is performed on the temperature between the two detection points.

Further, determining the RGB value based on the respective higher harmonic ratios of the three-phase power includes: using Min-Max scaling to convert the higher harmonic ratio of each phase of the three-phase power into data of 0-255, and The three converted data correspond to the three numbers in the RGB value.

Further, linear interpolation is performed on the pixel colors between the two detection points.

Through the above technical solutions, the present invention can produce the following beneficial effects:

The detection data of the power grid is converted into visual image data, so that the machine learning model can be used to identify abnormal points in the image. Automatic learning and derivation through machine learning avoids the manual establishment of mathematical models, thereby enabling anomaly identification of complex multi-branch networks.

Description of drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the drawings needed to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some of the drawings of the present invention. Embodiments, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of the method of the present invention.

Detailed ways

Below, the present invention will be further described with reference to the accompanying drawings and specific embodiments.

This embodiment solves the above technical problem through the following steps:

In one embodiment, referring to Figure 1, the present invention provides an artificial intelligence-based high-voltage cable condition monitoring method, which specifically includes:

Step 1: Draw a topology diagram based on the actual distribution of the multi-branch power grid.

Multi-branch power grid refers to a power grid structure with multiple branches or branches in the power system. In power systems, electrical energy is typically transmitted and distributed through multiple lines and transformers, forming a complex network of branch connections. Since in a multi-branch power grid, the detected signals usually span multiple nodes, causing the detection signals between different line segments to interfere with each other, the mathematical model is complex and difficult to model. In the existing technology, a single cable is usually used for detection, and this method Embodiments are directly applicable to multi-branch networks. Since this embodiment is only concerned with cable issues, when drawing a topology diagram based on the actual distribution of the power grid, only the cable structure can be drawn, and the three-phase TV can be divided into one line segment.

Step 2: Collect historical voltage signals, current signals, and temperature data of each reporting period of each phase of each detection point in the multi-branch power grid.

Each phase refers to each electrical phase in a three-phase alternating current system. The three-phase power system is a method commonly used for power transmission and distribution. It has high efficiency and stability. In the three-phase power system, power is transmitted through three cable lines that are offset by 120 degrees from each other. Each cable line is called an electrical phase. These three electrical phases are phase A, phase B and phase C. Each phase carries a part of the load in the power system, and is 120 degrees out of phase with each other to balance the load and current in the power system. Therefore, when monitoring high-voltage cables or performing power system analysis, the current, voltage and other parameters of each phase are taken into account to ensure the normal operation and fault detection of the power system.

In order to monitor the power grid in real time, detection points can be set up at regular intervals in the power grid, and voltage sensors, current sensors and temperature sensors can be installed at each detection point of each phase in the multi-branch power grid to capture voltage, current and temperature data. Use appropriate data acquisition equipment to connect sensors to collect real-time voltage, current, and temperature data. Power grid sensors have been widely used in existing power grids and can collect voltage signals, current signals, and temperature data based on any existing technology. Furthermore, in order to facilitate data analysis, it is also necessary to determine the frequency of data reporting. Since power grid signals usually require timing analysis, each reporting period is a timing, and the data within the reporting period is analyzed. The reporting period may be daily, hourly, or more frequent, and may be adjusted as needed, and is not specifically limited in this embodiment.

For example, in each phase of the power grid, a set of sensors is set up every 200 meters, and the data collected within 1 hour are reported every 1 hour. The voltage and current can be time series data within 1 hour, and the temperature can be Average or maximum value within 1 hour.

Furthermore, for subsequent data analysis, the collected data are recorded in the database, including timestamps, values, detection point numbers, etc. Data storage can use a database system, such as SQL database, etc.

Step 3: For data in the same reporting period, determine the higher harmonic ratio based on the voltage signal and current signal; determine the minimum temperature based on the temperature data, and subtract the minimum temperature from other temperatures to obtain the temperature increase value.

The data in the same reporting cycle refers to the data of all detection points of the power grid to be analyzed in a collection cycle. The data in the same cycle reflects the status of the power grid in the reporting cycle.

High-order harmonic components can provide important information when diagnosing partial discharge problems. Partial discharges usually produce high-frequency components and pulse signals. These signals will be reflected in the high-order harmonic components of current and voltage. High-order harmonic ratios It can be used to quantitatively evaluate the presence of high-order harmonic components. When this ratio is high, it may indicate partial discharge. Therefore, this embodiment selects the high-order harmonic ratio as the detection index.

Determining the ratio of higher harmonics involves spectral analysis of voltage and current signals to identify the presence of different harmonic components and calculate their ratios. Preprocess the collected signals, including removing DC components, filtering, etc., to prepare for spectrum analysis. Fourier transform is performed on the voltage signal and current signal, and the time domain signal is converted into a frequency domain signal. The spectrum after Fourier transform is analyzed to identify the frequency and amplitude of each harmonic component. Higher harmonics usually refer to the 3rd harmonic and above, that is, 3 times the fundamental frequency and above. Find the corresponding harmonic frequency in the spectrum and calculate its amplitude; calculating the ratio of higher harmonics is to compare the amplitude of a specific higher harmonic with the amplitude of the fundamental frequency (50Hz in China). The higher harmonic ratio can be expressed as: higher harmonic ratio = higher harmonic amplitude/fundamental frequency amplitude, so that the amplitude ratio of higher harmonics relative to the fundamental frequency can be obtained.

When there is partial discharge, the temperature will increase due to the effect of electric heating. Temperature increase is one of the indicators that can best reflect partial discharge. However, due to the large external influence of temperature, the temperature detection error is very large when used alone. It is usually not used as a positioning indicator alone; high The sub-harmonic ratio is affected by different branches and different phases, and the abnormal location cannot be completely accurately located; therefore, this embodiment uses a combination of harmonics and temperature to locate.

Since the temperature of the cable changes with the ambient temperature at different time periods, in order to remove the influence of the ambient temperature. Determine the minimum temperature based on the temperature data, that is, the minimum temperature within the current reporting period. The minimum temperature can be regarded as the temperature without the influence of heat, which is approximately the ambient temperature. The minimum temperature is subtracted from other temperatures to obtain the temperature increase value.

For example, in a reporting cycle, the temperatures of several detection points are 23, 23, 22, 22, 22, 24..., where 22 degrees is the lowest temperature, then the temperature increase value is divided into 1, 2, 0 ,0,0,2….

Step 4: Change the thickness of the line segments in the topology map according to the temperature increase value, determine the RGB value according to the respective higher harmonic ratios of the three-phase electricity, and color the topology map according to the determined RGB values to obtain monitoring image.

Temperature increase is one of the indicators that best reflects partial discharge. In order to facilitate visual analysis of temperature, this implementation uses the temperature increase value to represent the thickness of the line segment in the topology map. The position where the temperature does not increase is determined as the original line thickness, and the position where the temperature increases is amplified by a certain proportion according to the temperature increase value. For example, new line thickness = original line thickness * (1 + temperature increase value / 5); where the new line thickness is the adjusted line segment thickness, and the original line thickness is the line segment thickness of the original topology map.

Furthermore, since there is a certain distance between the sensors, in order to adjust all line segments, the temperature between the two detection points is linearly interpolated.

Harmonics are components in a voltage or current signal whose frequency is an integer multiple of the fundamental frequency. In a three-phase electrical system, there is a phase difference between each phase, which will cause harmonics to interact and influence between different phases, which is very similar to the mixing of three primary colors. Therefore, in order to visualize three-phase harmonics, this implementation converts the higher harmonic ratio of the three-phase harmonics into RGB colors. For example, the higher harmonic ratio of phase A is mapped to R color, and the higher harmonic ratio of phase B is mapped to R color. The higher harmonic ratio of phase C is mapped to color G, and the higher harmonic ratio of phase C is mapped to color B.

When converting the high-order harmonic ratio into RGB color, it is first necessary to perform standardized conversion. Preferably, this embodiment uses Min-Max scaling to convert the high-order harmonic ratio into data of 0-255. Min-Max scaling is a conventional method in data processing, and its principle will not be described in detail in this embodiment.

After Min-Max scaling, the higher harmonic ratio is converted into pixel color, that is, three-phase electricity. Each phase of electricity corresponds to a number from 0 to 255, and these three numbers correspond to three numbers in RGB. Thus, the color value of a pixel is generated, and the pixel at the detection position is assigned a value through the pixel color.

Furthermore, since there is a certain distance between the sensors, in order to adjust all line segments, the pixel colors between the two detection points are linearly interpolated.

Step 5: If there is an abnormal state in the current cycle, mark the abnormal position on the monitoring image.

From the historical data, the time and location of partial discharge abnormalities in the cable can be determined. If there is an abnormal state in the current cycle, the abnormal location is manually marked on the monitoring image and selected using the block diagram tool provided by the annotation tool. The location where the exception occurs. Annotation can be performed using any method in the art. For example, use LabelImg, VGG Image Annotator (VIA) and other tools. Obviously, if there are no exceptions, no annotation is needed. The annotated images can be learned by the machine learning model for image recognition, and the unlabeled images can also be used by the machine learning model for verification.

Step 6: Repeat steps 3 to 5 to process all historical data and obtain a historical monitoring image set; conduct neural network training with the historical monitoring image set to obtain the first model.

In order to obtain a sufficient number of training images, all historical data are processed as in steps 3 to 5, and abnormal images are annotated to obtain a set of historical monitoring images as training samples.

Neural network training using historical monitoring image sets can use any means in the existing technology. The training process can also use any means in the existing technology, such as classifying the training data, classifying the verification set, setting the loss function, and selecting optimization. Device, verification, parameter adjustment, etc., the above means are all conventional technical means in the prior art, and are not specifically limited in this embodiment.

Furthermore, since this embodiment needs to process images and image colors, it is preferred to use a convolutional neural network (CNN) or its various variants, such as ResNet, VGG, Inception, etc.

Step seven: collect real-time data of the current period; process the real-time data of the current period according to steps three to five to obtain the current monitoring image; input the current monitoring image into the first model to detect whether there is an abnormality.

After training the model, you can apply the model to actual detection, collect real-time data of the current period through the sensor network, and process the real-time data of the current period in the aforementioned steps three to five to obtain the current monitoring image (using temperature versus topology (Adjust the line thickness and use harmonic data to adjust the color); the first model already has the ability to identify abnormal locations in the image. When the current monitoring image is input to the first model, if there is a partial discharge abnormality, the A model automatically annotates images to identify abnormal locations.

Through the above steps, this embodiment converts the detection data of the power grid into visual image data, thereby facilitating the use of machine learning models to identify abnormal points in the image. Automatic learning and derivation through machine learning avoids the manual establishment of mathematical models, thereby enabling anomaly identification of complex multi-branch networks.

On the other hand, the present invention also provides a high-voltage cable condition monitoring system based on artificial intelligence, which is characterized in that the system includes the following modules:

Drawing module, used to draw topology diagrams based on the actual distribution of multi-branch power grids;

The first collection module is used to collect historical voltage signals, current signals, and temperature data of each reporting period of each phase of each detection point of the multi-branch power grid;

The calculation module is used to determine the higher harmonic ratio according to the voltage signal and the current signal for the data of the same reporting period; determine the minimum temperature according to the temperature data, and subtract the minimum temperature from other temperatures to obtain the temperature increase value;

An image processing module, configured to change the thickness of the line segments in the topology map according to the temperature increase value, determine the RGB value according to the respective higher harmonic ratios of the three-phase electricity, and color the topology map according to the determined RGB value. , get the monitoring image;

The marking module is used to mark the abnormal position on the monitoring image if there is an abnormal state in the current cycle;

The training module is used to run the computing module, the image processing module, and the annotation module, process all historical data, and obtain a historical monitoring image set; conduct neural network training with the historical monitoring image set to obtain the first model;

The detection module is used to collect real-time data of the current period; process the real-time data of the current period according to steps three to five to obtain the current monitoring image; input the current monitoring image to the first model to detect whether there is an abnormality.

Furthermore, the specific implementation methods of the above-mentioned high-voltage cable condition monitoring system based on artificial intelligence are the same as those of a high-voltage cable condition monitoring method based on artificial intelligence. All further technical solutions are fully integrated into an artificial intelligence-based high-voltage cable condition monitoring system.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solution 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 of ordinary skill in the art should understand that the present invention can still be modified. Modifications or equivalent substitutions may be made to the specific embodiments, and any modifications or equivalent substitutions that do not depart from the spirit and scope of the invention shall be covered by the scope of the claims of the invention.

Claims (10)

1. A high-voltage cable condition monitoring method based on artificial intelligence, characterized in that the method includes the following steps: Step 1: Draw a topology diagram based on the actual distribution of the multi-branch power grid; Step 2: Collect historical voltage signals, current signals, and temperature data of each reporting period of each phase of each detection point of the multi-branch power grid; Step 3: For data in the same reporting period, determine the higher harmonic ratio based on the voltage signal and current signal; determine the minimum temperature based on the temperature data, and subtract the minimum temperature from other temperatures to obtain the temperature increase value; Step 4: Change the thickness of the line segments in the topology map according to the temperature increase value, determine the RGB value according to the respective higher harmonic ratios of the three-phase electricity, and color the topology map according to the determined RGB values to obtain monitoring image; Step 5: If there is an abnormal state in the current cycle, mark the abnormal position on the monitoring image; Step 6: Repeat steps 3 to 5 to process all historical data and obtain a historical monitoring image set; conduct neural network training with the historical monitoring image set to obtain the first model; Step seven: collect real-time data of the current period; process the real-time data of the current period according to steps three to five to obtain the current monitoring image; input the current monitoring image into the first model to detect whether there is an abnormality. 2. A high-voltage cable status monitoring method based on artificial intelligence according to claim 1, characterized in that said changing the thickness of the line segment in the topology diagram according to the temperature increase value includes: new line thickness = original line thickness * ( 1+temperature increase value/5). 3. An artificial intelligence-based high-voltage cable condition monitoring method according to claim 2, characterized in that: linear interpolation is performed on the temperature between two detection points. 4. A high-voltage cable condition monitoring method based on artificial intelligence according to claim 1, characterized in that determining the RGB value according to the respective higher harmonic ratios of the three-phase electricity includes: using Min-Max scaling, dividing the three-phase The high-order harmonic ratio of each phase in the electricity is converted into data of 0-255, and the three converted data correspond to the three numbers in the RGB value. 5. An artificial intelligence-based high-voltage cable condition monitoring method according to claim 4, characterized in that linear interpolation is performed on the pixel color between two detection points. 6. A high-voltage cable condition monitoring system based on artificial intelligence, characterized in that the system includes the following modules: Drawing module, used to draw topology diagrams based on the actual distribution of multi-branch power grids; The first collection module is used to collect historical voltage signals, current signals, and temperature data of each reporting period of each phase of each detection point of the multi-branch power grid; The calculation module is used to determine the higher harmonic ratio according to the voltage signal and the current signal for the data of the same reporting period; determine the minimum temperature according to the temperature data, and subtract the minimum temperature from other temperatures to obtain the temperature increase value; An image processing module, configured to change the thickness of the line segments in the topology map according to the temperature increase value, determine the RGB value according to the respective higher harmonic ratios of the three-phase electricity, and color the topology map according to the determined RGB value. , get the monitoring image; The marking module is used to mark the abnormal position on the monitoring image if there is an abnormal state in the current cycle; The training module is used to run the computing module, the image processing module, and the annotation module, process all historical data, and obtain a historical monitoring image set; conduct neural network training with the historical monitoring image set to obtain the first model; The detection module is used to collect real-time data of the current period; process the real-time data of the current period according to steps three to five to obtain the current monitoring image; input the current monitoring image to the first model to detect whether there is an abnormality. 7. A high-voltage cable condition monitoring system based on artificial intelligence according to claim 6, characterized in that said changing the thickness of the line segment in the topology diagram according to the temperature increase value includes: new line thickness = original line thickness * ( 1+temperature increase value/5). 8. A high-voltage cable condition monitoring system based on artificial intelligence according to claim 7, characterized in that: linear interpolation is performed on the temperature between two detection points. 9. A high-voltage cable condition monitoring system based on artificial intelligence according to claim 6, characterized in that determining the RGB value according to the respective higher harmonic ratios of the three-phase electricity includes: using Min-Max scaling, dividing the three-phase The high-order harmonic ratio of each phase in the electricity is converted into data of 0-255, and the three converted data correspond to the three numbers in the RGB value. 10. A high-voltage cable condition monitoring system based on artificial intelligence according to claim 9, characterized in that linear interpolation is performed on the pixel color between two detection points.