Distribution network fault diagnosis method and system based on AI technology

Abstract

The embodiment of the invention relates to the technical field of information, in particular to a distribution network fault diagnosis method and system based on the AI technology. The method comprises the following steps: acquiring current monitoring data, voltage monitoring data, oil sample monitoring data, working noise of a transformer, ultrasonic data, a cable inspection image and electromagnetic radiation data; current features and voltage features are extracted and obtained, a first fault prediction model is input, and a first fault prediction result is obtained; comparing the oil sample monitoring data with reference oil sample data to obtain a second fault prediction result; extracting noise frequency domain features, and inputting the noise frequency domain features into a second fault prediction model to obtain a third fault prediction result; extracting ultrasonic frequency domain features, and comparing the ultrasonic frequency domain features with a preset reference baseline to obtain a first comparison result; extracting radiation characteristics, and comparing the radiation characteristics with reference radiation characteristics to obtain a second comparison result; obtaining a partial discharge fault prediction result; surface defects are identified, and a cable fault prediction result is obtained; and generating a distribution network fault diagnosis result.

Description

Technical Field

[0001] Multiple embodiments of this specification relate to the field of information technology, and particularly to a distribution network fault diagnosis method and system based on AI technology. Background Art

[0002] Common fault types in distribution networks mainly include short - circuit faults, open - circuit faults, grounding faults, partial discharges, overloads, and over - voltages. If these faults are not dealt with in time, they will affect the quality of life of residents and the production efficiency of enterprises. In addition, they will also cause damage to electrical equipment and pose a threat to personal safety. With the continuous expansion of the power grid scale and the continuous growth of power load, the safe and stable operation of the distribution network faces more and more challenges. Traditional distribution network fault diagnosis methods mainly rely on manual inspections and simple automated monitoring systems, and these methods are unable to cope when faced with complex and changeable power faults. Especially in the early detection of faults and accurate prediction of potential risks, traditional means are difficult to meet the requirements of modern power systems for high reliability and efficient operation and maintenance. Therefore, it is necessary to continue researching the technology of distribution network fault diagnosis. Summary of the Invention

[0003] Multiple embodiments of this specification describe a distribution network fault diagnosis method and system based on AI technology.

[0004] In a first aspect, an embodiment of this specification provides a distribution network fault diagnosis method based on AI technology, including the steps of: Obtain periodically collected current monitoring data, voltage monitoring data, oil sample monitoring data, the working noise of the transformer, ultrasonic data collected from the transformer shell, cable inspection images, and electromagnetic radiation data; Extract the features of the current monitoring data and voltage monitoring data to obtain current features and voltage features, and input the current features and voltage features into a pre - configured first fault prediction model to obtain a first fault prediction result; Compare the oil sample monitoring data with reference oil sample data to obtain a second fault prediction result; Extract the frequency - domain features of the working noise, denoted as noise frequency - domain features, and input the noise frequency - domain features into a preset second fault prediction model to obtain a third fault prediction result; Extract the frequency - domain features of the ultrasonic data, denoted as ultrasonic frequency - domain features, and compare the ultrasonic frequency - domain features with a preset reference baseline to obtain a first comparison result; Extract the features of the electromagnetic radiation data, denoted as radiation features, and compare the radiation features with a preset reference radiation feature to obtain a second comparison result; Obtain a partial discharge fault prediction result according to the first comparison result and the second comparison result; Identify surface defects of the cable inspection image using a preset image recognition model, and obtain a cable fault prediction result based on the surface defects; Generate a distribution network fault diagnosis result based on the first fault prediction result, the second fault prediction result, the third fault prediction result, the partial discharge fault prediction result, and the cable fault prediction result.

[0005] In a second aspect, an embodiment of this specification provides a distribution network fault diagnosis system based on AI technology, including: An acquisition module that obtains periodically acquired current monitoring data, voltage monitoring data, oil sample monitoring data, the operating noise of a transformer, ultrasonic data collected from the transformer housing, cable inspection images, and electromagnetic radiation data; A first prediction module that extracts features of the current monitoring data and the voltage monitoring data to obtain current features and voltage features, and inputs the current features and the voltage features into a pre-configured first fault prediction model to obtain a first fault prediction result; A first comparison module that compares the oil sample monitoring data with reference oil sample data to obtain a second fault prediction result; A second prediction module that extracts the frequency domain features of the operating noise, denoted as noise frequency domain features, and inputs the noise frequency domain features into a preset second fault prediction model to obtain a third fault prediction result; A second comparison module that extracts the frequency domain features of the ultrasonic data, denoted as ultrasonic frequency domain features, and compares the ultrasonic frequency domain features with a preset reference baseline to obtain a first comparison result; A third comparison module that extracts the features of the electromagnetic radiation data, denoted as radiation features, and compares the radiation features with preset reference radiation features to obtain a second comparison result; A third prediction module that obtains a partial discharge fault prediction result based on the first comparison result and the second comparison result; An identification module that uses a preset image recognition model to identify surface defects of the cable inspection image, and obtains a cable fault prediction result based on the surface defects; A diagnosis module that generates a distribution network fault diagnosis result based on the first fault prediction result, the second fault prediction result, the third fault prediction result, the partial discharge fault prediction result, and the cable fault prediction result.

[0006] In a third aspect, an embodiment of this specification provides an electronic device, including a processor and a memory; The processor is connected to the memory; The memory is used to store executable program code; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the method described in any of the above aspects.

[0007] In a fourth aspect, embodiments of the present specification provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method described in any of the above aspects is implemented.

[0008] In a fifth aspect, embodiments of the present specification provide a computer program product, including a computer program. When the computer program is executed by a processor, the method described in any of the above aspects is implemented.

[0009] The beneficial effects brought by the technical solutions provided in some embodiments of the present specification at least include: In multiple embodiments of the present specification, the provided power distribution network fault diagnosis method and system based on AI technology combine multiple information sources such as current monitoring data, voltage monitoring data, oil sample monitoring data, working noise frequency domain characteristics, ultrasonic data, and electromagnetic radiation data, and can more comprehensively cover potential fault points, improving the accuracy of fault detection. Partial discharge is a common hidden fault inside high-voltage equipment. By analyzing the ultrasonic data on the transformer shell and the electromagnetic radiation data in the surrounding environment, partial discharge problems can be detected early, which helps to prevent power distribution network faults.

[0010] Other features and advantages of multiple embodiments of the present specification will be further revealed in the following specific implementation manners and drawings. Description of the Drawings

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present specification, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present specification. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0012] Figure 1 It is a schematic diagram of the power distribution network composition provided by the embodiments of the present specification.

[0013] Figure 2 It is a schematic diagram of the power distribution network fault diagnosis method flow provided by the embodiments of the present specification.

[0014] Figure 3 It is a schematic diagram of the reference oil sample data provided by the embodiments of the present specification.

[0015] Figure 4 It is a schematic diagram of the second sample data provided by the embodiments of the present specification.

[0016] Figure 5Schematic diagram of the distribution network fault diagnosis system provided by the embodiments of this specification.

[0017] Figure 6 Schematic diagram of the electronic device provided by the embodiments of this specification. Specific embodiments

[0018] The technical solutions of the embodiments of this specification will be explained and described below with reference to the accompanying drawings of the embodiments of this specification. However, the following embodiments are only the preferred embodiments of this specification, not all of them. Based on the embodiments in the implementation manners, other embodiments obtained by those skilled in the art without creative efforts all fall within the protection scope of this specification.

[0019] The terms "first", "second", "third", etc. in the specification, claims and the above-mentioned drawings of this specification are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0020] In the following description, terms such as "inner", "outer", "upper", "lower", "left", "right", etc. indicating orientation or positional relationship are only for the convenience of describing the embodiments and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation of this specification.

[0021] The data involved in this application are all information and data authorized by users or fully authorized by all parties, and the collection, use and processing of the relevant data all comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0022] Before introducing the technical solutions described in this specification, the application scenarios of the technical solutions and related technologies will be introduced.

[0023] The distribution network is a network composed of all power transmission and distribution facilities starting from the low-voltage side of the substation 11 until before the user's household meter 14. Please refer to the appendix Figure 1, The distribution network mainly includes transmission lines (such as overhead lines or cables), transformer 12, switching equipment (including switchgear 13, circuit breakers, disconnectors, load switches, fuses, relays, etc.), protection devices, and metering equipment (such as in-house meters 14, current transformers, voltage transformers, power factor measurement devices, etc.), and is responsible for converting and distributing electric power from the high-voltage power grid to end-users. A distribution transformer area, also known as a "distribution transformer area" or simply "section", is a basic unit in the distribution network, usually referring to the scope or area powered by a single distribution transformer 12. Each distribution transformer area covers a certain geographical range and provides power services to users within that range. In urban areas, a distribution transformer area may cover several blocks or even a smaller area; while in rural or remote areas, the coverage area of a distribution transformer area may be larger. The distribution transformer area is an important part of the distribution network, and its main function is to convert electric power at a higher voltage level (such as 10 kV) into a lower voltage level (such as 220 V / 380 V) suitable for households and enterprises through the distribution transformer 12. Each distribution transformer area is equipped with corresponding monitoring and protection measures to ensure the safety and reliability of power supply.

[0024] As an important part of the power system, the safe and stable operation of the distribution network directly affects the power consumption quality of users. Common fault types in the distribution network include short-circuit faults, open-circuit faults, grounding faults, partial discharges, overloads, and overvoltages. Short-circuit faults are one of the most common faults, usually caused by equipment aging, insulation damage, or external factors (such as lightning strikes, animal contact, etc.). A short circuit will cause a sharp increase in current, which may damage electrical equipment and even cause a fire in severe cases. An open-circuit fault occurs when a point in the circuit is disconnected, which may be caused by wire breakage or loose connections. This situation will cause the load to not receive normal power supply and affect the normal use of users.

[0025] A grounding fault means that one or more points in the circuit are accidentally connected to the ground. This kind of fault not only may cause an electric shock hazard but also will change the voltage distribution of the system and affect other normally operating equipment. Partial discharges mainly occur inside high-voltage equipment, such as transformer 12, cables, etc. Long-term partial discharges will gradually damage the insulating material and may eventually lead to equipment failure. Overload means that the current carried by the circuit exceeds its design value, and overvoltage means that the voltage borne by the circuit is higher than the rated value. Both of these situations may cause equipment overheating or insulation damage, thereby shortening the equipment life or directly causing equipment damage.

[0026] If these faults are not handled in a timely manner, a series of hazards will ensue. The damage of electrical equipment requires expensive repair costs, and in severe cases, it may even lead to the paralysis of the entire power distribution system. In particular, grounding faults and partial discharge problems may pose a risk of electric shock to on-site workers. Faults can cause power outages, affecting the quality of life of residents and the production efficiency of enterprises. Fires caused by electrical equipment failures may pollute the surrounding environment.

[0027] Therefore, this specification discloses a distribution network fault diagnosis method based on AI technology. Please refer to the appendix Figure 2 , including the steps: Step S1) Obtain the periodically collected current monitoring data, voltage monitoring data, oil sample monitoring data, operating noise of transformer 12, ultrasonic data collected from the transformer shell, cable inspection images, and electromagnetic radiation data.

[0028] For the current monitoring data and voltage monitoring data, they are usually collected at the circuit entrance in the substation 11 or distribution box, or directly near the critical load points. Intelligent meters, multi-functional meters, or dedicated current / voltage sensors are used as the collection devices, and these devices can monitor and record the changes in current and voltage in real time. Set the sampling frequency according to actual needs, such as once per second or a higher frequency, in order to capture transient phenomena. The oil sample monitoring data mainly targets the insulating oil used inside oil-immersed transformers, and is collected at the sampling valve at the bottom of the transformer. Special oil sample collection tools are used to ensure that the samples are not contaminated. Analyze regularly, such as monthly or quarterly, and check parameters such as the moisture content, acid value, and dielectric strength of the oil sample.

[0029] The operating noise of transformer 12 can be collected at positions close to the transformer shell, especially those parts that are prone to vibration or noise. High-sensitivity microphones or sound level meters are used as the collection devices. By continuously monitoring or regularly checking according to a plan, the operating noise of transformer 12 is obtained. The ultrasonic data collected from the transformer shell is also located on the transformer shell, especially where there may be partial discharge or other abnormal activities. Portable ultrasonic detectors or fixed-mounted ultrasonic sensors are used as the collection devices. It is collected during routine inspections, or set to automatically collect at regular intervals to help detect internal defects.

[0030] The collection positions of the cable inspection images include vulnerable parts such as along the power cables and their connection points (joints). UAVs equipped with high-definition camera robot inspection systems or manual hand-held cameras are used as the collection devices. Arrange inspections according to the maintenance plan, or use an automated system to achieve daily monitoring. The electromagnetic radiation data is collected in places close to electrical equipment, such as around high-voltage cables in the substation 11, etc. Electromagnetic field strength measuring instruments or spectrum analyzers are used as the collection devices. Measure regularly or continuously monitor, especially after new equipment is installed.

[0031] Step S2) Extract the features of the current monitoring data and voltage monitoring data to obtain current features and voltage features, and input the current features and voltage features into a pre-configured first fault prediction model to obtain a first fault prediction result.

[0032] The method for pre-configuring the first fault prediction model includes: Obtain current monitoring data and voltage monitoring data marked with fault labels; Extract the features of the current monitoring data and voltage monitoring data, and associate the features with the fault labels as sample data; Establish a machine learning model, and use the sample data to train the machine learning model, and obtain the first fault prediction model according to the trained machine learning model; After obtaining the current features and voltage features, group the current features and voltage features according to the acquisition location, input the grouped current features and voltage features into the first fault prediction model respectively, and take the outputs of the first fault prediction models corresponding to all the grouped current features and voltage features as the first fault prediction result.

[0033] The features extracted from the current monitoring data and voltage monitoring data, for example, the features are the mean, variance, peak value, frequency components and frequency amplitudes of the current and voltage. These features can obtain key information reflecting the health status of the power system.

[0034] When pre-configuring the first fault prediction model, it is necessary to collect historical current monitoring data and voltage monitoring data marked with fault labels. The fault label refers to identifying whether each group of data corresponds to a specific type of fault. Extract features from the current monitoring data and voltage monitoring data, and associate the extracted features with the corresponding fault labels to form sample data. Use the sample data to train the machine learning model. For example, the machine learning model uses support vector machine (SVM), decision tree, random forest or neural network, etc. After training, the obtained model is the first fault prediction model, which is used to predict whether there are similar faults in the newly collected data.

[0035] After new current features and voltage features are obtained, they are grouped according to the acquisition location. Data at different locations may reflect the states of different devices or circuit segments. Input each grouped feature into the first fault prediction model respectively, and the model will give an evaluation of the fault possibility corresponding to this group of data. Finally, the results of all groups are summarized to form a comprehensive first fault prediction result.

[0036] The types of faults that the first fault prediction model can predict include short - circuit faults, open - circuit faults, overload faults, and power quality problems. A short - circuit fault is a situation where the current increases sharply due to accidental contact between conductors. It may cause overheating and even damage to electrical equipment. An open - circuit fault occurs when a point in the circuit is disconnected, resulting in abnormal current flow and affecting the normal operation of the load. At this time, the current will be detected as zero. When the current in the circuit exceeds the designed capacity, an overload fault occurs. Long - term overload faults will accelerate the aging of insulating materials and increase the risk of fire. Power quality problems, such as voltage fluctuations and harmonic distortions, will affect the normal operation of sensitive electronic equipment or cause equipment damage.

[0037] By deeply analyzing the current monitoring data and voltage monitoring data and combining advanced machine - learning techniques, the first fault prediction model can effectively identify the aforementioned faults, helping maintenance personnel take timely measures to avoid risks and reduce losses.

[0038] Step S3): Compare the oil - sample monitoring data with the reference oil - sample data to obtain the second fault prediction result.

[0039] The oil - sample monitoring data includes one or more of the moisture content, acid value, dielectric strength, dissolved gas content in oil, oil - liquid color, and oil - level of transformer oil. The reference oil - sample data includes multiple oil - sample data. The oil - sample data includes the oil - sample monitoring data and associated fault labels. The method of comparing the oil - sample monitoring data with the reference oil - sample data to obtain the second fault prediction result includes: Represent both the oil - sample monitoring data and the reference oil - sample data in vector form; Obtain the oil - sample data and associated fault labels in the reference oil - sample data that are closest to the oil - sample monitoring data according to the vector distance, and obtain the second fault prediction result according to the fault labels.

[0040] Comparing the oil - sample monitoring data with the reference oil - sample data can obtain the second fault prediction result, which can be used to evaluate the operating state of transformer 12. The oil - sample monitoring data includes one or more of the parameters such as the moisture content, acid value, dielectric strength, dissolved gas content in oil, oil - liquid color, and oil - level of transformer oil. The reference oil - sample data contains multiple oil - sample data, and each sample data not only records the above - mentioned monitoring parameters but also includes associated fault labels.

[0041] Convert both the currently acquired oil sample monitoring data and the reference oil sample data into vector form. Each sample can be represented as a multi-dimensional vector, where each dimension corresponds to the value of a specific detection parameter. Calculate the distances between these vectors to find one or more samples in the reference oil sample data that are closest to the currently acquired oil sample monitoring data. Determine the second fault prediction result based on the fault labels associated with the closest samples. Please refer to the appendix Figure 3 , it is necessary to pre-establish a reference database containing various typical fault situations in advance. When new oil sample monitoring data shows characteristics similar to a certain known fault mode, it can quickly identify and give corresponding fault warnings.

[0042] By analyzing the oil sample monitoring data, faults such as insulation oil aging problems, internal partial discharges, overheating faults, arc discharges, and abnormal oil levels can be obtained. For insulation oil aging problems, due to the influence of long-term operation or high-temperature environments, the oil will undergo chemical changes, resulting in a decline in its performance. The specific manifestations are an increase in acid value and a decrease in dielectric strength. If no measures are taken in a timely manner, this situation will seriously affect the insulation performance of transformer 12 and increase the risk of faults. Internal partial discharge phenomena usually occur when there are small defects or voids inside the transformer. This will cause a significant increase in the content of gases such as hydrogen and methane in the oil. Persistent partial discharges will not only accelerate the aging process of the insulation material but may also lead to more serious insulation breakdown accidents.

[0043] Overheating faults are generally caused by reasons such as poor winding contact or core short circuits, resulting in local temperature increases. The concentrations of gases such as ethane and ethylene in the oil will show abnormalities. If the overheating fault is not controlled, it may further evolve into a more serious fault type. Arc discharge is a high-energy discharge phenomenon, commonly seen in situations such as short circuits or severe insulation failures. The typical sign is a sharp increase in the content of acetylene gas in the oil. Arc discharge has great destructive power to the equipment, and immediate action must be taken to prevent the loss from expanding. Abnormal oil level refers to the situation where the oil level is too high or too low. A too-low oil level will affect the cooling effect of transformer 12 or reduce its insulation performance; while a too-high oil level may cause seal damage due to expansion. Either situation poses a threat to the safe operation of transformer 12. Through the oil sample monitoring data, various potential fault hazards of transformer 12 can be effectively identified, improving the speed and accuracy of fault diagnosis.

[0044] Step S4) Extract the frequency-domain characteristics of the working noise, denoted as noise frequency-domain characteristics, and input the noise frequency-domain characteristics into a preset second fault prediction model to obtain a third fault prediction result.

[0045] The method for presetting the second fault prediction model includes: Read the working noise with fault annotations and the working noise in several normal working states, and extract the frequency-domain features of the working noise, denoted as noise frequency-domain features; Associate the fault annotations and normal working states as associated labels with the noise frequency-domain features to form the second sample data; Build a second machine learning model and use the second sample data to train the second machine learning model; Obtain a second fault prediction model according to the trained second machine learning model.

[0046] The frequency-domain features of the working noise refer to the data obtained by converting the time-domain signal to the frequency domain. The frequency-domain features are used to reveal the state information of the internal mechanical structure or electrical components of the device. Fourier transform and wavelet transform are used to extract the frequency-domain features.

[0047] Obtain the working noise with fault annotations and the working noise in multiple normal working states. The fault annotation refers to whether each segment of the working noise corresponds to a specific type of fault. Extract the frequency-domain features from the working noise and label them as noise frequency-domain features. Associate the faults and normal working states as labels with the noise frequency-domain features to form the second sample data, as shown in the appendix Figure 4 shown. Use the second sample data to build and train a machine learning model. By repeatedly iterating and training the model, it can learn the differences between the normal state and various fault states. The trained model is used as the second fault prediction model to predict whether there are abnormalities in the new noise frequency-domain features. After the new noise frequency-domain features are extracted, input them into the second fault prediction model, and the model will output a probability value or directly give the most likely fault type as the third fault prediction result.

[0048] Through the working noise, the fault types that can be identified include wear or looseness of the mechanical components of the transformer 12, cooling system faults, etc. For the case of wear or looseness of the mechanical components, long-term operation may cause the fixing screws to loosen or gaps to appear in other connection parts, which will introduce sounds of specific frequencies in the working noise. Over time, this phenomenon may lead to more serious mechanical faults such as bearing damage or rotor imbalance. Cooling system faults can also be discovered by analyzing the working noise. If the fan blades are damaged or the radiator is blocked, the cooling efficiency will decrease, and abnormal low-frequency noise may appear in the noise spectrum. If this situation is not resolved, it will affect the overall performance of the device and even cause overheating damage. When the transformer 12 has abnormal vibrations, it also means some potential problems such as unstable foundation or improper installation. These problems often generate vibration noises of specific patterns.

[0049] Step S5) Extract the frequency-domain features of the ultrasonic data, denoted as ultrasonic frequency-domain features, and compare the ultrasonic frequency-domain features with a preset reference baseline to obtain a first comparison result.

[0050] The method for presetting the reference baseline includes: Obtain ultrasonic data of a plurality of transformers 12 under normal operating conditions and fault conditions respectively, denoted as reference ultrasonic data; Extract the frequency-domain features of the reference ultrasonic data, denoted as reference frequency-domain features, and associate the reference frequency-domain features with the state of the transformer 12; Obtain a reference baseline based on all the reference frequency-domain features after association with the state; The method for comparing the ultrasonic frequency-domain features with the preset reference baseline includes: Compare the ultrasonic frequency-domain features with each reference frequency-domain feature in the reference baseline according to the frequency composition respectively to obtain the similarity of each frequency composition, and obtain the total similarity according to the similarities of all frequency compositions; Take the reference frequency-domain feature with the highest total similarity as the matching reference frequency-domain feature, and obtain the first comparison result according to the state associated with the matching reference frequency-domain feature.

[0051] Collect ultrasonic data on the outer shell of the transformer 12 under normal operating conditions and various typical fault conditions respectively, denoted as reference ultrasonic data. Then, perform frequency-domain transformation on these reference ultrasonic data, such as using the fast Fourier transform (FFT), to extract the frequency composition information of each sample and form reference frequency-domain features. Establish an association between these reference frequency-domain features and the corresponding device states (such as normal, insulation aging, partial discharge, winding deformation, etc.) to obtain a reference baseline.

[0052] When new ultrasonic data is collected, perform frequency-domain analysis on it as well to extract the current ultrasonic frequency-domain features. Compare this feature with each reference frequency-domain feature in the reference baseline one by one according to the frequency components, and calculate the similarity between each frequency point. Usually, the cosine similarity, correlation coefficient, or the reciprocal of the Euclidean distance can be used to measure the degree of similarity. Combine the similarities of all frequency points to obtain the total similarity between this ultrasonic frequency-domain feature and each reference frequency-domain feature. Select the reference frequency-domain feature with the highest total similarity as the matching result, and obtain the first comparison result according to the associated state label. If the matching is a sample known to have a partial discharge fault, it is considered that the current device may have a partial discharge problem.

[0053] Step S6) Extract the features of the electromagnetic radiation data, denoted as radiation features, and compare the radiation features with a preset reference radiation feature to obtain a second comparison result.

[0054] The method for presetting the reference radiation feature includes: Collect electromagnetic radiation data near multiple devices under normal operating conditions, and extract the characteristics of the electromagnetic radiation data respectively; The set of characteristics of all the electromagnetic radiation data is used as the reference radiation characteristics; The method of comparing the radiation characteristics with the preset reference radiation characteristics includes: Represent the radiation characteristics and the reference radiation characteristics in vector form; Based on the vector distance, obtain the vector distance closest to the radiation characteristics and the reference radiation characteristics; According to the result of comparing the vector distance with the preset threshold, obtain the second comparison result.

[0055] The processing method for electromagnetic radiation data is slightly different from that of ultrasonic data. Under the normal operating conditions of the device, collect the electromagnetic radiation data around it at multiple time points, and extract the characteristics of each group of data, including but not limited to parameters such as signal intensity, frequency distribution, and energy density. Combine all these characteristics into a set as the preset reference radiation characteristics. After obtaining new electromagnetic radiation data, also extract its characteristics and represent them in vector form. Convert each group of characteristics in the reference radiation characteristics into vector form. Then calculate the distance between the new feature vector and the reference vector. Common methods include Euclidean distance, Mahalanobis distance, or Manhattan distance, etc. If the minimum vector distance is less than the set threshold, it is considered that the current device operating state is consistent with the reference state and belongs to the normal situation; otherwise, it may be in an abnormal state, such as the risk of partial discharge.

[0056] Step S7) Obtain the partial discharge fault prediction result according to the first comparison result and the second comparison result.

[0057] The first comparison result (from ultrasonic data) and the second comparison result (from electromagnetic radiation data) are respectively applicable to the detection of partial discharge in different devices or positions.

[0058] Partial discharge refers to the non-penetrating discharge phenomenon that occurs in certain regions under the action of an electric field in the insulation system of electrical equipment. It will not immediately cause the device to fail, but will gradually damage the insulating material and may eventually lead to insulation breakdown, causing serious accidents.

[0059] The first comparison result is mainly used for the detection of partial discharge faults in the transformer 12. It includes oil-immersed transformer 12 and dry-type transformer 12. The oil-immersed transformer 12 is a typical high-voltage electrical equipment with a complex internal structure and is prone to partial discharge due to reasons such as insulation aging, bubbles, and cracks. Although the dry-type transformer 12 does not have insulating oil, its solid insulating material may also cause partial discharge due to moisture, pollution, or mechanical damage.

[0060] In transformer 12, if there are tiny voids in the insulating cardboard or there is a trace amount of moisture in the oil, partial discharge may occur in the area where the electric field is concentrated. This kind of discharge will generate trace gases (such as hydrogen, methane, etc.), emit ultrasonic waves at a specific frequency, and be accompanied by electromagnetic radiation within a certain frequency range. By detecting the changes in these physical signals, potential insulation defects can be detected early.

[0061] The second comparison result is mainly used for detecting partial discharge problems in equipment such as cables and switchgear 13. For cable joints and terminations at high voltage levels, if the connection is not tight or the insulation layer is damaged, it is likely to become the main part where partial discharge occurs. During the operation of the circuit breaker and disconnector in switchgear 13, due to contact wear or poor contact, partial discharge may also occur.

[0062] Although the GIS gas-insulated switchgear has high insulation performance, if there is gas leakage due to poor sealing or foreign objects enter the interior, it may also trigger partial discharge. In cable terminations, if the shielding layer is broken due to improper installation or the insulation layer becomes thinner due to external extrusion during operation, it will also cause an increase in the local electric field at the weak point, inducing partial discharge. At this time, the electromagnetic radiation signal will change significantly, and the ultrasonic signal will also show an increase in high-frequency components. Through the multi-source signal fusion and comparison analysis method described in steps S5 to S7, it is possible to effectively identify whether there are potential partial discharge hazards in the equipment, achieve early warning and accurate positioning, provide a scientific decision-making basis for maintenance personnel, and improve the safety and reliability of the distribution network system.

[0063] Step S8) Use a preset image recognition model to identify the surface defects of the cable inspection image, and obtain a cable fault prediction result based on the surface defects.

[0064] When constructing the image recognition model, it is trained through a large number of images marked with different types of cable surface defects. The training data set should cover as many actual working conditions and defect types as possible, such as corrosion, cracks, breakage, aging, joint looseness, etc. Commonly used deep learning architectures include convolutional neural networks (CNNs), which can automatically extract complex features from images for classification or detection tasks.

[0065] During the cable inspection process, high-resolution cable images are taken using drones, robots or professional equipment carried by humans. The images are then input into a preset image recognition model. The model will process each picture and identify whether there are known types of surface defects. Specifically, the model will output information such as the location, size and category of each detected defect.

[0066] Based on the identified surface defects, the system can further generate cable fault prediction results. Exemplarily, if severe corrosion is found in the cable, it may be due to the long-term exposure to harsh environments, resulting in the loss of protection of the metal sheath. In this case, the mechanical strength of the cable may be reduced, increasing the risk of breakage. If obvious cracks or damages are detected on the cable outer sheath, it indicates that the internal conductor may have been damaged or is at risk of moisture intrusion, which will seriously affect the safety and efficiency of power transmission. For aging cables, the decline in insulation performance may lead to leakage or even short-circuit accidents. Timely replacement of severely aged parts is crucial for preventing major faults. If there are signs of looseness at the joints, it will not only affect the normal transmission of current but also may cause local overheating problems, thereby accelerating the aging rate of materials and increasing the fire risk.

[0067] Step S9): Generate a distribution network fault diagnosis result based on the first fault prediction result, the second fault prediction result, the third fault prediction result, the partial discharge fault prediction result, and the cable fault prediction result.

[0068] After summarizing the fault prediction results obtained by analyzing the current monitoring data and voltage monitoring data (corresponding to obtaining the first fault prediction result), the oil sample monitoring data (corresponding to obtaining the second fault prediction result), the working noise frequency domain characteristics (corresponding to obtaining the third fault prediction result), the ultrasonic data and electromagnetic radiation data (corresponding to obtaining the partial discharge fault prediction result), and the cable inspection images (corresponding to obtaining the cable fault prediction result), a distribution network fault diagnosis result is generated. The distribution network fault diagnosis result can reflect faults such as short-circuit faults, open-circuit faults, grounding faults, partial discharges, overloads, and overvoltages, and realizes accurate positioning and timely handling of possible faults in the substation area.

[0069] On the other hand, this specification provides a distribution network fault diagnosis system based on AI technology. Please refer to the appendix Figure 5 , including: The acquisition module 100 acquires periodically collected current monitoring data, voltage monitoring data, oil sample monitoring data, the working noise of the transformer 12, ultrasonic data collected from the transformer shell, cable inspection images, and electromagnetic radiation data; The first prediction module 200 extracts the characteristics of the current monitoring data and voltage monitoring data to obtain current characteristics and voltage characteristics, and inputs the current characteristics and voltage characteristics into a pre-configured first fault prediction model to obtain a first fault prediction result; The first comparison module 300 compares the oil sample monitoring data with the reference oil sample data to obtain a second fault prediction result; The second prediction module 400 extracts the frequency-domain features of the operating noise, denoted as noise frequency-domain features, and inputs the noise frequency-domain features into a preset second fault prediction model to obtain a third fault prediction result; The second comparison module 500 extracts the frequency-domain features of the ultrasonic data, denoted as ultrasonic frequency-domain features, and compares the ultrasonic frequency-domain features with a preset reference baseline to obtain a first comparison result; The third comparison module 600 extracts the features of the electromagnetic radiation data, denoted as radiation features, and compares the radiation features with preset reference radiation features to obtain a second comparison result; The third prediction module 700 obtains a partial discharge fault prediction result based on the first comparison result and the second comparison result; The recognition module 800 uses a preset image recognition model to recognize the surface defects of the cable inspection image, and obtains a cable fault prediction result based on the surface defects; The diagnosis module 900 generates a distribution network fault diagnosis result based on the first fault prediction result, the second fault prediction result, the third fault prediction result, the partial discharge fault prediction result, and the cable fault prediction result.

[0070] Please refer to Figure 6 the structural schematic diagram of an electronic device provided by the embodiment of the present specification shown.

[0071] As Figure 6As shown, the electronic device 1100 may include: at least one processor 1101, at least one network interface 1104, a user interface 1103, a memory 1105, and at least one communication bus 1102. Among them, the communication bus 1102 can be used to realize the connection and communication of the above-mentioned various components. Among them, the user interface 1103 may include buttons, and the optional user interface may further include a standard wired interface and a wireless interface. Among them, the network interface 1104 may, but is not limited to, include a Bluetooth module, an NFC module, a Wi-Fi module, etc. Among them, the processor 1101 may include one or more processing cores. The processor 1101 connects various parts within the entire electronic device 1100 through various interfaces and lines, and by running or executing instructions, programs, code sets, or instruction sets stored in the memory 1105, and by calling the data stored in the memory 1105, it executes various functions of the routing device 1100 and processes data. Optionally, the processor 1101 may be implemented in at least one of the hardware forms of DSP, FPGA, and PLA. The processor 1101 may integrate one or several combinations of a CPU, a GPU, and a modem, etc. Among them, the CPU mainly processes the operating system, the user interface, application programs, etc.; the GPU is responsible for the rendering and drawing of the content to be displayed on the display screen; the modem is used to process wireless communication.

[0072] It can be understood that the above-mentioned modem may not be integrated into the processor 1101 and may be implemented separately by a single chip.

[0073] Among them, the memory 1105 may include RAM and may also include ROM. Optionally, the memory 1105 includes a non-transitory computer-readable medium. The memory 1105 can be used to store instructions, programs, codes, code sets, or instruction sets. The memory 1105 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store the data involved in the above-mentioned various method embodiments. Optionally, the memory 1105 may also be at least one storage device located far from the aforementioned processor 1101. The memory 1105, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and application programs. The processor 1101 can be used to call the application programs stored in the memory 1105 and execute the methods in the above-mentioned multiple embodiments.

[0074] The embodiments of this specification also provide a computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer or a processor, the computer or the processor is caused to execute multiple steps in the above embodiments. If each component module of the above electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in the computer-readable storage medium.

[0075] The embodiments of this specification also provide a computer program product, including a computer program. When the computer program is executed by a processor, multiple steps in the above embodiments are implemented.

[0076] Without conflict, the technical features in this embodiment and the implementation solutions can be combined arbitrarily.

[0077] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes multiple computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this specification are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that integrates multiple available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a Digital Versatile Disc (DVD)), or a semiconductor medium (for example, a Solid State Disk (SSD)), etc.

[0078] When implemented through hardware or firmware, the foregoing method flow is programmed into a hardware circuit to obtain a corresponding hardware circuit structure and implement corresponding functions. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is an integrated circuit whose logic function is determined by a user's programming of the device. A designer can program a digital system "integrated" on a PLD by themselves, without having to ask a chip manufacturer to design and fabricate a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, which is similar to the software compiler used in program development and writing. The original code before compilation also has to be written in a specific programming language, which is called a hardware description language (HDL), and there are not only one but many kinds of HDLs. Those skilled in the art should also be clear that as long as the method flow is slightly logically programmed in the above-mentioned several hardware description languages and programmed into an integrated circuit, it is easy to obtain a hardware circuit that implements the logical method flow.

[0079] The foregoing embodiments are merely described as preferred implementation manners of this specification, and do not limit the scope of this specification. Without departing from the design spirit of this specification, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of this specification shall fall within the protection scope determined by the claims of this specification.

Claims

1. A distribution network fault diagnosis method based on AI technology, characterized in that, Including the steps of: Obtaining periodically collected current monitoring data, voltage monitoring data, oil sample monitoring data, the operating noise of the transformer, ultrasonic data collected from the transformer casing, cable inspection images, and electromagnetic radiation data; Extracting the features of the current monitoring data and voltage monitoring data to obtain current features and voltage features, and inputting the current features and voltage features into a pre-configured first fault prediction model to obtain a first fault prediction result; Comparing the oil sample monitoring data with reference oil sample data to obtain a second fault prediction result; Extracting the frequency domain features of the operating noise, denoted as noise frequency domain features, and inputting the noise frequency domain features into a preset second fault prediction model to obtain a third fault prediction result; Extracting the frequency domain features of the ultrasonic data, denoted as ultrasonic frequency domain features, and comparing the ultrasonic frequency domain features with a preset reference baseline to obtain a first comparison result; Extracting the features of the electromagnetic radiation data, denoted as radiation features, and comparing the radiation features with preset reference radiation features to obtain a second comparison result; Obtaining a partial discharge fault prediction result based on the first comparison result and the second comparison result; Using a preset image recognition model to identify surface defects in the cable inspection images, and obtaining a cable fault prediction result based on the surface defects; Generating a distribution network fault diagnosis result based on the first fault prediction result, the second fault prediction result, the third fault prediction result, the partial discharge fault prediction result, and the cable fault prediction result.

2. The distribution network fault diagnosis method based on AI technology according to claim 1, wherein The method for pre-configuring the first fault prediction model includes: Obtaining current monitoring data and voltage monitoring data marked with fault labels; Extracting the features of the current monitoring data and voltage monitoring data, associating the features with the fault labels, and using them as sample data; Establishing a machine learning model, training the machine learning model using the sample data, and obtaining the first fault prediction model based on the trained machine learning model; After obtaining the current features and voltage features, grouping the current features and voltage features according to the acquisition location, inputting the grouped current features and voltage features into the first fault prediction model respectively, and using the outputs of the first fault prediction models corresponding to all the grouped current features and voltage features as the first fault prediction result.

3. The distribution network fault diagnosis method based on AI technology according to claim 1 or 2, wherein The oil sample monitoring data includes one or more of the moisture content, acid value, dielectric strength, dissolved gas content in oil table, oil liquid color, and oil liquid level of the transformer oil, and the reference oil sample data includes multiple oil sample data, and the oil sample data includes oil sample monitoring data and associated fault labels. The method for comparing the oil sample monitoring data with the reference oil sample data to obtain a second fault prediction result includes: Representing both the oil sample monitoring data and the reference oil sample data in vector form; Obtaining the oil sample data and associated fault labels in the reference oil sample data that are closest to the oil sample monitoring data according to the vector distance, and obtaining the second fault prediction result according to the fault labels.

4. The power distribution network fault diagnosis method based on AI technology according to claim 1 or 2, characterized in that The method for presetting the second fault prediction model includes: Read the working noise with fault labels and the working noise in several normal working states, and extract the frequency domain features of the working noise, denoted as noise frequency domain features; Associate the fault label and the normal working state as associated labels with the noise frequency domain features to form the second sample data; Establish a second machine learning model and use the second sample data to train the second machine learning model; Obtain the second fault prediction model according to the trained second machine learning model.

5. The power distribution network fault diagnosis method based on AI technology according to claim 1 or 2, characterized in that The method for presetting the reference baseline includes: Obtain ultrasonic data of a plurality of transformers in normal working states and fault states respectively, denoted as reference ultrasonic data; Extract the frequency domain features of the reference ultrasonic data, denoted as reference frequency domain features, and associate the reference frequency domain features with the state of the transformer; Obtain the reference baseline according to all the reference frequency domain features after the association state; The method for comparing the ultrasonic frequency domain features with the preset reference baseline includes: Compare the ultrasonic frequency domain features with each reference frequency domain feature in the reference baseline according to the frequency composition respectively to obtain the similarity of each frequency composition, and obtain the total similarity according to the similarities of all frequency compositions; The reference frequency domain feature with the highest total similarity is used as the matching reference frequency domain feature, and the first comparison result is obtained according to the state associated with the matching reference frequency domain feature.

6. The power distribution network fault diagnosis method based on AI technology according to claim 1 or 2, characterized in that The method for presetting the reference radiation feature includes: Collect electromagnetic radiation data near a plurality of devices in normal working states, and extract the features of the electromagnetic radiation data respectively; The set of features of all the electromagnetic radiation data is used as the reference radiation feature; The method for comparing the radiation feature with the preset reference radiation feature includes: Represent the radiation feature and the reference radiation feature in vector form; Obtain the vector distance closest to the radiation feature and the reference radiation feature based on the vector distance; Obtain the second comparison result according to the result of comparing the vector distance with the preset threshold.

7. A distribution network fault diagnosis system based on AI technology, characterized in that, It includes: A collection module that acquires periodically collected current monitoring data, voltage monitoring data, oil sample monitoring data, the working noise of the transformer, ultrasonic data collected from the transformer shell, cable inspection images, and electromagnetic radiation data; A first prediction module that extracts the features of the current monitoring data and the voltage monitoring data to obtain current features and voltage features, and inputs the current features and voltage features into a pre-configured first fault prediction model to obtain a first fault prediction result; A first comparison module that compares the oil sample monitoring data with the reference oil sample data to obtain a second fault prediction result; A second prediction module that extracts the frequency domain features of the working noise, denoted as noise frequency domain features, and inputs the noise frequency domain features into a preset second fault prediction model to obtain a third fault prediction result; A second comparison module extracts the frequency-domain features of the ultrasonic data, denoted as ultrasonic frequency-domain features, and compares the ultrasonic frequency-domain features with a preset reference baseline to obtain a first comparison result; A third comparison module extracts the features of the electromagnetic radiation data, denoted as radiation features, and compares the radiation features with a preset reference radiation feature to obtain a second comparison result; A third prediction module obtains a partial discharge fault prediction result according to the first comparison result and the second comparison result; An identification module uses a preset image recognition model to identify the surface defects of the cable inspection image, and obtains a cable fault prediction result according to the surface defects; A diagnosis module generates a distribution network fault diagnosis result according to the first fault prediction result, the second fault prediction result, the third fault prediction result, the partial discharge fault prediction result and the cable fault prediction result; 8. An electronic device, characterized in that, It includes a processor and a memory; The processor is connected to the memory; The memory is used to store executable program codes; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to execute the method according to any one of claims 1-6; 9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program, when executed by the processor, implements the method according to any one of claims 1-6; 10. A computer program product comprising a computer program, characterized in that, The computer program, when executed by the processor, implements the method according to any one of claims 1-6.

Images