A machine learning based power distribution network fault self-healing method and system

Abstract

The application provides a power distribution network fault self-healing method and system based on machine learning, and belongs to the technical field of power distribution network fault self-healing.The application collects line information and equipment information in a power distribution network before and after a fault occurs, extracts main line information, and then performs standardization processing on the main line information, so that different equipment has the same standardized information, then a fault identification model based on machine learning is trained by using a large amount of standardized equipment information and fault point information on the line, and finally, when the main line is restored by the traditional self-healing mode and fails, the fault identification model is used to identify the fault point position and the corresponding tie switch information to restore power again.The application adopts the identification model based on machine learning to supplement the traditional self-healing restoration failure, and improves the reliability of the power distribution network fault self-healing.

Description

Technical Field

[0001] This invention belongs to the field of distribution network fault self-healing technology, specifically relating to a distribution network fault self-healing method and system based on machine learning. Background Technology

[0002] Self-healing in distribution networks involves using automated devices or systems to monitor the operation of distribution lines, promptly detect line faults, diagnose and isolate faulty sections, and restore power supply to non-faulty sections. Based on different implementation methods, it can be mainly divided into centralized control type, local control type, and master station-local collaborative type. The master station centralized type has significant advantages; therefore, considering construction and renovation costs and the workload of on-site terminal maintenance, the master station centralized type and the master station-local collaborative type, which involves comprehensive analysis of power transfer schemes by the master station, have become the mainstream self-healing strategies for distribution networks.

[0003] Current self-healing methods for power distribution networks mainly involve collecting fault information from terminals, determining the fault location based on certain logical judgments, and finally determining a fault recovery plan to achieve self-healing. However, due to the generally low quality of power distribution terminals, signal loss and mistransmission often occur, and the fixed logic of the self-healing strategy frequently leads to incomplete signals causing the self-healing action to fail. Summary of the Invention

[0004] In view of this, the present invention aims to solve the problem that the existing self-healing method of power distribution network may fail due to incomplete signals caused by the low quality of power distribution terminals.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides a self-healing method for distribution network faults based on machine learning, comprising the following steps:

[0007] Collect line information, equipment operation information, and interconnection switch information within a set time period before and after a power distribution network fault.

[0008] Based on the line information of the distribution network, extract the information of the main lines to be monitored in the distribution network, and associate it with the fault information and tie switch information of the equipment on the main lines;

[0009] The information collected on the main line is standardized to obtain the associated equipment on the main line. For each associated equipment, all fault information collected during the fault is standardized into information before the equipment, information of the equipment itself, and information after the equipment. Associated equipment is equipment that has a power supply connection with each other.

[0010] Using standardized associated equipment information on the main line as training data and corresponding fault point information as labels, a fault identification model based on machine learning is trained.

[0011] In response to the failure signal of the current main line power restoration, power is restored again based on the fault location identified by the fault identification model and the corresponding tie switch information.

[0012] Furthermore, the specific equipment operation information includes:

[0013] The equipment's alarm information, telemetry information, location information, and waveform recording information.

[0014] Furthermore, the information collected along the main lines will be standardized, specifically including:

[0015] All equipment on the main line is divided according to its location, and for each piece of equipment, the related upstream and downstream equipment are obtained;

[0016] The operation information of the associated devices is preprocessed to obtain information about the device itself, the information before the device, and the information after the device for each device.

[0017] The fault point information is obtained based on whether the fault point occurs before or after the equipment.

[0018] Furthermore, the associated devices specifically include:

[0019] The upstream device and the downstream device of the current device, where the upstream device is the device that would affect the current device's electrical state when separated from it, and the downstream device is the device that would affect the current device's electrical state when separated from it.

[0020] Furthermore, the preprocessing specifically includes:

[0021] For each device, extract some or all of the alarm information, telemetry information, location information, and waveform information before and after the device. When alarm information is repeated, deduplication is performed. When alarm information conflicts, it is selected based on the weight of the conflict or mutual cancellation. The average value is taken for telemetry information, and waveform information is converted into overcurrent information for processing.

[0022] Furthermore, the fault identification model identifies the location of the fault point, specifically as follows:

[0023] The trained fault identification model is used to identify the probability that the fault point on the current trunk line is located after a certain device. When the probability exceeds a set threshold, a fault is determined to have occurred, and the location of the fault point is obtained.

[0024] Furthermore, when power is restored, if there are multiple power restoration tie switches, the tie switch with the lowest load should be selected for power restoration.

[0025] Secondly, the present invention provides a machine learning-based self-healing system for distribution network faults, comprising:

[0026] The information acquisition unit is used to collect line information, equipment operation information, and interconnection switch information within a set time period before and after a power distribution network fault.

[0027] The information extraction unit is used to extract information on the main lines to be monitored in the distribution network based on the line information of the distribution network, and to associate the fault information and tie switch information of the equipment on the main lines.

[0028] The information processing unit is used to standardize the information collected on the main line to obtain the associated equipment on the main line. For each associated equipment, all fault information collected during a fault is standardized into information before the equipment, information of the equipment itself, and information after the equipment. The associated equipment are devices that have power supply connections with each other.

[0029] The model training unit is used to train a fault identification model based on machine learning, using standardized associated equipment information on the main line as training data and corresponding fault point information as labels.

[0030] The backup power restoration unit, in response to the current main line power restoration failure signal, is used to restore power again based on the fault location identified by the fault identification model and the corresponding tie switch information.

[0031] Thirdly, the present invention provides a self-healing device for power distribution network faults based on machine learning, the device including a processor and a memory:

[0032] The memory is used to store computer programs and send the instructions of the computer programs to the processor;

[0033] The processor executes, according to the instructions of the computer program, a machine learning-based self-healing method for power distribution network faults, as described in the first aspect.

[0034] Fourthly, the present invention provides a computer storage medium storing a computer program, which, when executed by a processor, implements a machine learning-based self-healing method for power distribution network faults as described in the first aspect.

[0035] In summary, this invention provides a machine learning-based method and system for self-healing of distribution network faults. It collects information on several lines and equipment in the distribution network before and after a fault occurs, extracts the main line information, and then standardizes this main line information to ensure that different devices have the same standardized information. Then, it uses a large amount of standardized equipment information and fault point information on the lines to train a machine learning-based fault identification model. Finally, when the main line fails to be restored using traditional self-healing methods, it uses the fault location identified by the fault identification model and the corresponding tie switch information to restore power again. This invention, based on traditional self-healing, uses a machine learning-based identification model to supplement the solution for failed traditional self-healing restorations, thus improving the reliability of distribution network fault self-healing. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart of a machine learning-based self-healing method for power distribution network faults, provided as an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0039] Please see Figure 1 This invention provides a machine learning-based method for self-healing of distribution network faults, comprising the following steps:

[0040] S1: Collect line information, equipment operation information, and interconnection switch information within a set time period before and after a power distribution network fault.

[0041] It should be noted that traditional distribution network fault self-healing relies on comprehensive analysis and judgment of information from a single terminal and its adjacent terminals within the distribution network to quickly locate and isolate faulty sections. However, due to the generally poor quality of distribution terminals in some areas, signal loss and false transmissions are common, and the fixed logic of the self-healing strategy often leads to incomplete signals causing self-healing failures. Therefore, this embodiment collects more diverse and comprehensive information from the distribution network to provide reliable data for subsequent identification of line fault points.

[0042] S2: Extract information on the main lines to be monitored in the distribution network based on the line information of the distribution network, and associate it with the fault information and tie switch information of the equipment on the main lines.

[0043] Because distribution networks have a specific topology, the affected power supply range varies depending on the location of the fault, and the fault recovery plan also differs. Not all faults have a solution for transferring power (restoring power). Distribution network faults can be categorized by location into main line faults and branch line faults. Branch line faults generally do not have a solution for transferring power, while main line faults usually do have a solution for transferring power.

[0044] Therefore, based on the aforementioned collection of information from the distribution network, it is necessary to further obtain relevant information about the main lines. For main line information, the main line diagram can be obtained by using a branch-and-leaf pruning method based on the single-line diagram topology. Through the main lines, the equipment on the main lines can be identified, and then, through the main line equipment, relevant information about the equipment within a specified time period can be obtained. Additionally, the main line diagram information can also be used to identify the related tie switches on the main lines.

[0045] S3: Standardize the information collected on the main line to obtain the associated devices on the main line. For each associated device, standardize all fault information collected during the fault into information before the device, information of the device itself, and information after the device. Associated devices are devices that have power supply connections with each other.

[0046] S4: Using standardized associated equipment information on the main line as training data and corresponding fault point information as labels, a fault identification model based on machine learning is trained.

[0047] Machine learning learns a function mapping from input variable X to input variable Y based on given labeled training data. In this embodiment, a fault identification model is trained using information from several devices obtained from the main distribution network lines as training data, and whether a fault occurs after a certain device as the label value. Multiple model training samples can be generated for a single fault. The training method for the model is not limited here; methods such as logistic regression and neural networks can be used for training.

[0048] S5: In response to the current main line power restoration failure signal, restore power again based on the fault location identified by the fault identification model and the corresponding tie switch information.

[0049] It should be noted that this fault recovery scheme, due to the need to collect more information, may be slower than the traditional logic-based self-healing method. However, in practical applications, it can serve as a supplement to the traditional logic-based self-healing method. For the current backbone line (i.e., the backbone line to be monitored), this self-healing method can be initiated and analyzed simultaneously with the traditional self-healing method. The traditional self-healing method will prioritize power restoration, and if the traditional self-healing method fails to restore power, this self-healing method will be used as a supplement.

[0050] This self-healing solution requires no maintenance of self-healing parameters by the user and does not affect the user's original self-healing strategy after deployment.

[0051] This embodiment provides a machine learning-based method for self-healing of distribution network faults. It collects information on several lines and equipment in the distribution network before and after a fault occurs, extracts the main line information, and then standardizes this main line information to ensure that different devices have the same standardized information. A machine learning-based fault identification model is then trained using a large amount of standardized equipment information and fault point information on the lines. Finally, when the main line fails to be restored using traditional self-healing methods, the fault location identified by the fault identification model and the corresponding tie switch information are used to restore power again. This invention, based on traditional self-healing, uses a machine learning-based identification model to supplement the solution for failed traditional self-healing restorations, thus improving the reliability of distribution network fault self-healing.

[0052] In one embodiment of the present invention, for machine learning samples, the main data collected includes the single-line diagram topology structure, single-line diagram line parameters, attributes of each device on the single-line diagram (including main line equipment / branch equipment), alarm information of each device (such as protection actions, overcurrent signals, blocking signals, etc.), telemetry information (i.e., active power, reactive power, current, etc.), location information (i.e., switch opening and closing status), waveform recording information (waveform recording information can be analyzed to determine whether there is an overcurrent signal in the current device, the approximate location of the fault, etc.), and tie switch information. It is also necessary to collect the location of the fault point or the main line equipment before and after the fault point, and the power restoration tie switch.

[0053] Since the relevant line information and equipment information obtained in the aforementioned embodiments are information about all equipment on the backbone line, different backbone maps may require different training data. Therefore, to reduce the dataset required for training, the data needs to be standardized.

[0054] 1. Divide all equipment on the main line: According to the location of the equipment, take each equipment as the core and classify all equipment on the main line diagram to obtain the current equipment itself, the upstream equipment of the current equipment (i.e., when a certain equipment is separated, it will affect the power supply status of this equipment) and the downstream equipment of the current equipment (i.e., when this equipment is separated, it will affect the power supply equipment). Since the operation of the distribution network ring network is relatively rare, the operation of the ring network is not considered.

[0055] 2. Classify the characteristics of the equipment: the characteristic value of the equipment itself, the characteristic value of the upstream equipment of the current equipment. Since there are multiple upstream equipment of the current equipment, there may be multiple duplicate alarm signal characteristic values. Take one of them. For telemetry or tele-signaling information, the average value can be taken. For waveform recording information, it can be converted into overcurrent information. The downstream equipment of the current equipment is also processed in the same way.

[0056] 3. Tag value: The Boolean value indicating whether the fault point is downstream of the device itself is used as the tag value.

[0057] The above is a detailed description of an embodiment of a machine learning-based self-healing method for distribution network faults according to the present invention. The following will provide a detailed description of an embodiment of a machine learning-based self-healing system for distribution network faults and other related devices according to the present invention.

[0058] This embodiment provides a machine learning-based self-healing system for distribution network faults, including:

[0059] The information acquisition unit is used to collect line information, equipment operation information, and interconnection switch information within a set time period before and after a power distribution network fault.

[0060] The information extraction unit extracts information on the main lines to be monitored in the distribution network based on the line information of the distribution network, and associates it with the fault information and tie switch information of the equipment on the main lines.

[0061] The information processing unit is used to standardize the information collected on the main line to obtain the associated equipment on the main line. For each associated equipment, all fault information collected during a fault is standardized into information before the equipment, information of the equipment itself, and information after the equipment. The associated equipment are devices that have power supply connections with each other.

[0062] The model training unit is used to train a fault identification model based on machine learning, using standardized associated equipment information on the main line as training data and corresponding fault point information as labels.

[0063] The backup power restoration unit, in response to the current main line power restoration failure signal, is used to restore power again based on the fault location identified by the fault identification model and the corresponding tie switch information.

[0064] In another embodiment of the present invention, a self-healing device for power distribution network faults based on machine learning is provided. The device includes a processor and a memory.

[0065] The memory is used to store computer programs and send the instructions of the computer programs to the processor;

[0066] The processor executes a machine learning-based self-healing method for power distribution network faults, as provided in the foregoing embodiments, according to the instructions of the computer program.

[0067] In another embodiment of the present invention, a computer storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, it implements a machine learning-based self-healing method for power distribution network faults as provided in the foregoing embodiments.

[0068] It should be noted that the distribution network fault self-healing system provided in the above embodiments is used to implement the distribution network fault self-healing method provided in the aforementioned embodiments. The specific settings of each unit are based on their mutual cooperation to fully implement the method, and will not be elaborated here.

[0069] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A machine learning-based method for self-healing faults in distribution networks, characterized in that, Includes the following steps: Collect line information, equipment operation information, and interconnection switch information within a set time period before and after a power distribution network fault. Based on the line information of the distribution network, extract the information of the main line to be monitored in the distribution network, and associate the fault information of the equipment on the main line with the information of the tie switch; The information collected on the main line is standardized to obtain the associated devices on the main line. For each associated device, all fault information collected during a fault is standardized into information before the device, information of the device itself, and information after the device. The associated devices are devices that have power supply connections with each other. Using standardized associated equipment information on the main line as training data and corresponding fault point information as labels, a fault identification model based on machine learning is trained. In response to the failure signal of power restoration of the current main line, power is restored again based on the fault location identified by the fault identification model and the corresponding tie switch information; The information collected from the main network is standardized, specifically including: All equipment on the main line is divided according to its location, and for each piece of equipment, the related upstream and downstream equipment are obtained; The operating information of the associated devices is preprocessed to obtain, for each device, information about the device itself, information before the device, and information after the device; The fault point information is obtained based on whether the fault point occurs before or after the equipment. The associated devices specifically include: The upstream device and the downstream device of the current device, wherein the upstream device refers to the device that would affect the electrical state of the current device when it is separated, and the downstream device refers to the device that would affect the electrical state of the current device when it is separated from the current device.

2. The machine learning-based self-healing method for distribution network faults according to claim 1, characterized in that, The equipment operation information specifically includes: The equipment's alarm information, telemetry information, location information, and waveform recording information.

3. The machine learning-based self-healing method for distribution network faults according to claim 1, characterized in that, The preprocessing specifically includes: For each device, some or all of the alarm information, telemetry information, location information, and waveform recording information before and after the device are extracted. When the alarm information is repeated, it is deduplicated. When the alarm information is conflicting, it is selected according to the weight of the conflict or mutual cancellation. The telemetry information is averaged and the waveform recording information is converted into overcurrent information for processing.

4. The machine learning-based self-healing method for distribution network faults according to claim 1, characterized in that, The fault identification model identifies the location of the fault point, specifically as follows: The trained fault identification model is used to identify the probability that the fault location of the current trunk line is behind a certain device. When the probability exceeds a set threshold, a fault is determined to have occurred, and the fault location is obtained.

5. The machine learning-based self-healing method for distribution network faults according to claim 1, characterized in that, When power is restored, if there are multiple power restoration tie switches, select the tie switch with the lowest load for power restoration.

6. A machine learning-based self-healing system for distribution network faults, characterized in that, include: The information acquisition unit is used to collect line information, equipment operation information, and interconnection switch information within a set time period before and after a power distribution network fault. The information extraction unit is used to extract the information of the main line to be monitored in the distribution network based on the line information of the distribution network, and associate the fault information of the equipment on the main line with the tie switch information; The information processing unit is used to standardize the information collected on the main line to obtain the associated devices on the main line. For each associated device, all fault information collected during a fault is standardized into information before the device, information of the device itself, and information after the device. The associated devices are devices that have power supply connections with each other. The model training unit is used to train a fault identification model based on machine learning, using standardized associated equipment information on the main line as training data and corresponding fault point information as labels. The backup power restoration unit, in response to the current main line power restoration failure signal, is used to restore power again based on the fault location identified by the fault identification model and the corresponding tie switch information; The information collected from the main network is standardized, specifically including: All equipment on the main line is divided according to its location, and for each piece of equipment, the related upstream and downstream equipment are obtained; The operating information of the associated devices is preprocessed to obtain, for each device, information about the device itself, information before the device, and information after the device; The fault point information is obtained based on whether the fault point occurs before or after the equipment. The associated devices specifically include: The upstream device and the downstream device of the current device, wherein the upstream device refers to the device that would affect the electrical state of the current device when it is separated, and the downstream device refers to the device that would affect the electrical state of the current device when it is separated from the current device.

7. A machine learning-based self-healing device for distribution network faults, characterized in that, The device includes a processor and a memory: The memory is used to store computer programs and send the instructions of the computer programs to the processor; The processor executes a machine learning-based self-healing method for power distribution network faults as described in any one of claims 1-5 according to the instructions of the computer program.

8. A computer storage medium, characterized in that, The computer storage medium stores a computer program, which, when executed by a processor, implements a machine learning-based self-healing method for power distribution network faults as described in any one of claims 1-5.

Images