Self-healing control method, device and equipment for medium and low voltage power grid, and storage medium

Abstract

The application relates to a self-recovery control method, device and equipment of a medium and low voltage power grid, and a storage medium. The method comprises the following steps: collecting electrical and non-electrical multi-modal operation data before and after a fault to construct reference and monitoring data sets, extracting electrical and non-electrical characteristics, identifying abnormal mutation characteristics through comparison and establishing a fault characteristic set, inputting the fault characteristic set into a trained fault classification model to output fault information; constructing a mixed integer programming model with the minimum power loss and the maximum load recovery as the targets, substituting the fault information and a power grid network model to obtain a cross-level recovery strategy, combining the fault characteristic set and the fault information to complete fault isolation, generating fusion control instructions with the maximum load recovery and the most stable voltage as the targets, and finally executing the cross-level recovery strategy and the fusion control instructions to realize power supply recovery. The method can realize fast recovery and stable operation of the power grid, and simultaneously considers preventive maintenance and life cycle management.

Description

Technical Field

[0001] This application relates to the field of smart grid technology, and in particular to a self-healing control method, device, computer equipment, computer-readable storage medium, and computer program product for medium and low voltage power grids. Background Technology

[0002] With the development of smart grid technology, a self-healing control method for power grids has emerged. This method collects data information from the distribution network, analyzes the current operating status of the distribution network, and selects the corresponding control strategy.

[0003] However, traditional methods mainly focus on the post-fault handling phase, with relatively insufficient research on preventative maintenance and life-cycle management of power grid equipment. This leads to a situation where, when the power grid faces potential fault risks, traditional methods suffer from untimely responses, making it difficult to fully control the impact on overall power grid stability. Summary of the Invention

[0004] Therefore, it is necessary to provide a self-healing control method, device, computer equipment, computer-readable storage medium, and computer program product for medium and low voltage power grids that can take into account both fault handling and preventive maintenance, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a self-healing control method for medium- and low-voltage power grids, including:

[0006] Acquire a baseline dataset consisting of multimodal operating data collected before the fault, and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0007] Electrical and non-electrical features are extracted from multimodal operation data. Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0008] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0009] To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. Fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0010] Fault isolation is performed based on fault feature sets and fault information, and fusion control commands are generated with the goal of restoring the most load and the most stable voltage. Cross-level recovery strategies and fusion control commands are executed to restore power supply.

[0011] In one embodiment, electrical data includes three-phase current and three-phase voltage, and non-electrical data includes equipment temperature, partial discharge pulse signals, and mechanical vibration signals; extracting electrical and non-electrical features from the multimodal operation data includes:

[0012] Based on continuously sampled three-phase current and three-phase voltage, the current and voltage differences between adjacent sampling points and the quadratic difference between adjacent differences are calculated; the root mean square values ​​of current and voltage within a time window are calculated by sliding through a time window; the energy values ​​of current and voltage, and the approximate entropy of current and voltage used to characterize the degree of numerical disorder are calculated; the differences, quadratic differences, root mean square values, energy values, and approximate entropy of current and voltage are used as electrical characteristics.

[0013] Based on continuously sampled device temperatures, the temperature difference and root mean square (RMS) value are calculated; based on partial discharge pulse signals, the maximum pulse amplitude, the number of pulses per second, and the sum of squares of pulse amplitude are obtained; based on mechanical vibration signals, wavelet transform is performed to obtain the decomposed high-frequency peak value, the sum of squares of vibration signals, and the approximate entropy of vibration signals; the temperature difference and RMS value, the maximum pulse amplitude, the number of pulses per second, the sum of squares of pulse amplitude, the high-frequency peak value, the sum of squares of vibration signals, and the approximate entropy of vibration signals are used as non-electrical features.

[0014] In one embodiment, based on the comparison results between electrical features in the monitoring dataset and electrical features in the benchmark dataset, and the comparison results between non-electrical features in the monitoring dataset and non-electrical features in the benchmark dataset, anomalous mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset, including:

[0015] For each feature, obtain the monitoring mean of the corresponding type of feature in the monitoring dataset and the baseline mean of the corresponding type of feature in the benchmark dataset; calculate the difference between the monitoring mean and the benchmark mean; obtain the product of the standard deviation of the corresponding type of feature in the benchmark dataset and a preset multiple as a preset threshold; if the difference is greater than the preset threshold, the corresponding type of feature in the monitoring dataset is regarded as an abnormal mutation feature.

[0016] In one embodiment, fault isolation is performed based on a fault feature set and fault information, including:

[0017] Based on the fault location in the fault feature set and fault information, the isolated forest algorithm is used to analyze the time-series power grid operation dataset composed of the benchmark dataset and the monitoring dataset to obtain the fault feeder or fault node.

[0018] To isolate the fault, the circuit breaker of the faulty feeder or faulty node is tripped.

[0019] In one embodiment, a fusion control command is generated with the goal of maximizing load recovery and achieving the most stable voltage, including:

[0020] Calculate the fault probability based on the fault feature set and the fault type in the fault information;

[0021] Based on the failure probability, the weight coefficients are calculated; the quantization parameters corresponding to the reinforcement learning instructions for load recovery and the quantization parameters corresponding to the model predictive control instructions for voltage stabilization are obtained; the quantization parameters include output power and switch closure degree.

[0022] Based on the weight coefficients, the quantization parameters corresponding to the reinforcement learning instructions and the quantization parameters corresponding to the model prediction control instructions are weighted and summed to obtain the fused quantization parameters.

[0023] Based on the fused quantization parameters, a fused control command is generated.

[0024] In one embodiment, a cross-level recovery strategy and fusion control commands are executed to restore power supply, including:

[0025] Close the circuit breaker of the medium-voltage standby line to reconfigure the medium- and low-voltage power grid topology;

[0026] Power supply restoration is carried out according to the quantization parameters in the fusion control command, the switching operation sequence in the cross-level recovery strategy, the output limit of the distributed power source, and the order of load restoration.

[0027] In one embodiment, the method further includes:

[0028] From a time-series power grid operation dataset consisting of a benchmark dataset and a monitoring dataset, we extract equipment temperature time-series, partial discharge cumulative energy, and vibration frequency domain features, and perform preprocessing.

[0029] Based on the pre-processed equipment temperature time series, partial discharge cumulative energy, and vibration frequency domain characteristics, the remaining lifespan of the equipment is predicted, and corresponding levels of early warning are issued based on the prediction results.

[0030] Secondly, this application also provides a self-healing control device for medium and low voltage power grids, comprising:

[0031] The acquisition module is used to acquire a baseline dataset consisting of multimodal operating data collected before the fault and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0032] The extraction module is used to extract electrical and non-electrical features from multimodal operation data; based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set;

[0033] The input and output module is used to input the fault feature set into the trained fault classification model and output the corresponding fault information, including fault type, fault location and fault impact range.

[0034] The construction and solution module is used to construct a mixed-integer programming model with the goal of minimizing power outage losses and maximizing load recovery. The fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0035] The control module is used to isolate faults based on fault feature sets and fault information, and generate fusion control commands aimed at restoring the maximum load and the most stable voltage; it executes cross-level recovery strategies and fusion control commands to restore power supply.

[0036] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0037] Acquire a baseline dataset consisting of multimodal operating data collected before the fault, and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0038] Electrical and non-electrical features are extracted from multimodal operation data. Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0039] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0040] To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. Fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0041] Fault isolation is performed based on fault feature sets and fault information, and fusion control commands are generated with the goal of restoring the most load and the most stable voltage. Cross-level recovery strategies and fusion control commands are executed to restore power supply.

[0042] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0043] Acquire a baseline dataset consisting of multimodal operating data collected before the fault, and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0044] Electrical and non-electrical features are extracted from multimodal operation data. Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0045] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0046] To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. Fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0047] Fault isolation is performed based on fault feature sets and fault information, and fusion control commands are generated with the goal of restoring the most load and the most stable voltage. Cross-level recovery strategies and fusion control commands are executed to restore power supply.

[0048] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0049] Acquire a baseline dataset consisting of multimodal operating data collected before the fault, and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0050] Electrical and non-electrical features are extracted from multimodal operation data. Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0051] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0052] To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. Fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0053] Fault isolation is performed based on fault feature sets and fault information, and fusion control commands are generated with the goal of restoring the most load and the most stable voltage. Cross-level recovery strategies and fusion control commands are executed to restore power supply.

[0054] The aforementioned self-healing control methods, devices, computer equipment, computer-readable storage media, and computer program products for medium- and low-voltage power grids collect electrical and non-electrical multimodal operating data before and after a fault to construct benchmark and monitoring datasets, respectively. Electrical and non-electrical features are extracted from these two datasets. Abnormal mutation features are identified by comparing the corresponding features of the two datasets, and a fault feature set is constructed. This feature set is then input into a trained fault classification model to output fault information such as fault type, location, and impact range. Accurate fault identification and comprehensive characterization are achieved by fusing electrical and non-electrical multimodal features, and the accuracy of fault information output is ensured by the trained model. A mixed-integer programming model is then constructed with the objectives of minimizing power outage losses and maximizing restored load. By substituting fault information and a power grid network model containing medium- and low-voltage power grid topology, line parameters, and node load levels, a cross-level restoration strategy is obtained. Simultaneously, fault isolation is achieved by combining the fault feature set and fault information, generating a fusion control command aimed at maximizing restored load and voltage stability. Finally, the cross-level restoration strategy and fusion control command are executed to restore power supply. By leveraging the synergy of mixed-integer programming models and multi-objective fusion control commands, fault isolation and power restoration across medium and low voltage levels can be achieved. This takes into account multiple objectives such as power outage losses, load restoration, and voltage stability, making the power grid's self-healing decisions more scientific and the restoration strategies more optimized. It can effectively improve the accuracy of fault handling in medium and low voltage power grids and the efficiency and stability of power restoration, maximizing the continuity of power supply. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is an application environment diagram of a self-healing control method for a medium- and low-voltage power grid in one embodiment;

[0057] Figure 2 This is a flowchart illustrating a self-healing control method for a medium- and low-voltage power grid in one embodiment.

[0058] Figure 3 This is a flowchart illustrating a self-healing control method for a medium- and low-voltage power grid in another embodiment.

[0059] Figure 4 This is a structural block diagram of a self-healing control device for a medium- and low-voltage power grid in one embodiment;

[0060] Figure 5 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0062] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0063] The self-healing control method for medium and low voltage power grids provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on the cloud or other network servers. First, electrical and non-electrical multimodal operating data before and after the fault are collected to construct benchmark and monitoring datasets. Electrical and non-electrical features are extracted from the two datasets. Abnormal mutation features are identified by comparing the corresponding features of the two datasets, and a fault feature set is constructed. This set is input into the trained fault classification model to output fault information such as fault type, location, and impact range. Then, a mixed-integer programming model is constructed with the objectives of minimizing power outage losses and maximizing restored load. The fault information and a power grid network model containing medium- and low-voltage grid topology, line parameters, and node load levels are substituted to solve for a cross-level restoration strategy. Simultaneously, fault isolation is completed by combining the fault feature set and fault information. A fusion control command is generated with the objective of restoring the most load and achieving the most stable voltage. Finally, the cross-level restoration strategy and fusion control command are executed to restore power supply.

[0064] Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, and projection equipment. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted displays. Head-mounted displays can be virtual reality (VR) devices, augmented reality (AR) devices, and smart glasses. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0065] In one exemplary embodiment, such as Figure 2 As shown, a self-healing control method for medium and low voltage power grids is provided, which is then applied to... Figure 1 Taking server 104 as an example, the explanation includes the following steps 202 to 210. Wherein:

[0066] Step 202: Obtain a baseline dataset consisting of multimodal operating data collected before the fault and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0067] Electrical data directly reflects the electrical operating characteristics of the power grid, such as voltage, current, power, frequency, power factor, and switch status. Non-electrical data indirectly reflects the operating environment and equipment status of the power grid, such as ambient temperature and humidity, equipment temperature rise, wind speed, equipment vibration data, image and video data, and communication status data. The benchmark dataset, formed by organizing and integrating multi-modal operating data collected before a power grid fault occurs under normal operating conditions, serves as a reference benchmark for determining whether the power grid's operating status is abnormal. The monitoring dataset, formed by organizing and integrating multi-modal operating data collected after a power grid fault occurs under abnormal operating conditions, is the core data source for analyzing power grid fault status and extracting fault characteristics.

[0068] Step 204: Extract electrical and non-electrical features from the multimodal operation data; based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, identify abnormal mutation features from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set;

[0069] Electrical features refer to key quantitative indicators extracted from the electrical data of the multi-modal operation data of the power grid, which can characterize the electrical operating status of the power grid. They are the characteristic expression of electrical data, such as voltage fluctuation amplitude, current fluctuation frequency, power change rate, and three-phase current imbalance. Non-electrical features refer to key indicators extracted from the non-electrical data of the multi-modal operation data of the power grid, which can reflect changes in the status of power grid equipment or operating environment. They are the characteristic expression of non-electrical data, such as equipment temperature rise rate, peak vibration frequency, sudden changes in ambient temperature and humidity, and characteristic parameters of equipment appearance in images. Abnormal mutation features refer to features in the monitoring dataset that significantly deviate from the normal range and undergo abrupt changes compared with the corresponding features in the benchmark dataset. The fault feature set refers to the feature set formed by integrating and normalizing all abnormal mutation features identified from the electrical and non-electrical features of the monitoring dataset. It centrally reflects the characteristic information of power grid faults and is the core basis for subsequent fault diagnosis and analysis.

[0070] Step 206: Input the fault feature set into the trained fault classification model and output the corresponding fault information, including fault type, fault location and fault impact range.

[0071] The fault information refers to the key information reflecting the core attributes of power grid faults, output by the fault classification model after parsing the fault feature set. It includes three core categories: fault type, fault location, and fault impact range. Fault type refers to the specific category of the fault occurring in the power grid, indicating the fault mode in the power grid's electrical system or equipment, such as single-phase grounding faults, phase-to-phase short-circuit faults, line overload faults, equipment insulation faults, and open-circuit faults. Fault location refers to the specific spatial and equipment location where the power grid fault occurs, accurate to the physical location of specific line sections, switching equipment, transformer nodes, and branch lines in medium and low voltage power grids. Fault impact range refers to the area and equipment range of the power grid directly or indirectly affected by the fault, including outage sections, affected distribution nodes, load users, related power supply lines, and power grid equipment, reflecting the degree to which the fault impacts the power grid operation.

[0072] Step 208: With the goal of minimizing power outage losses and maximizing load recovery, a mixed-integer programming model is constructed; the fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain the cross-level recovery strategy; the power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels;

[0073] Among them, power outage losses refer to the various economic and functional losses caused to power users, grid operators, and society due to power supply interruptions after a grid failure, covering quantitative indicators such as user production and operation losses, grid equipment wear and tear, and power supply reliability assessment losses. Load restoration refers to the load capacity that can be restored to power supply through methods such as power transfer and reconfiguration during grid failure handling.

[0074] Step 210: Based on the fault feature set and fault information, perform fault isolation and generate a fusion control command with the goal of restoring the most load and the most stable voltage; execute the cross-level recovery strategy and fusion control command to restore power supply.

[0075] Fault isolation refers to the process of electrically isolating the faulty line section and equipment location from the normally operating power grid system by operating power distribution equipment such as switches and circuit breakers, based on the characteristic information and accurate diagnostic results of the power grid fault. The purpose is to prevent the fault from spreading further and ensure the safe and stable operation of the power grid in non-faulty areas. Integrated control commands refer to unified control commands that can be directly issued to various execution terminals of the power grid, with the dual optimization objectives of restoring the maximum load and maximizing voltage stability.

[0076] In the aforementioned self-healing control method for medium and low voltage power grids, baseline and monitoring datasets are constructed by collecting electrical and non-electrical multimodal operating data before and after a fault. Electrical and non-electrical features are extracted from these two datasets. By comparing the corresponding features of the two datasets, abnormal mutation features are identified and a fault feature set is constructed. This set is then input into a trained fault classification model to output fault information such as fault type, location, and impact range. Accurate fault identification and comprehensive characterization are achieved by fusing electrical and non-electrical multimodal features, and the accuracy of fault information output is ensured by relying on the trained model. A mixed-integer programming model is then constructed with the objectives of minimizing power outage losses and maximizing restored load. By substituting fault information and a power grid network model containing medium and low voltage power grid topology, line parameters, and node load levels, a cross-level restoration strategy is obtained. Simultaneously, fault isolation is completed by combining the fault feature set and fault information, and a fusion control command is generated with the objective of restoring the most load and achieving the most stable voltage. Finally, the cross-level restoration strategy and fusion control command are executed to restore power supply. By leveraging the synergy of mixed-integer programming models and multi-objective fusion control commands, fault isolation and power restoration across medium and low voltage levels can be achieved. This takes into account multiple objectives such as power outage losses, load restoration, and voltage stability, making the power grid's self-healing decisions more scientific and the restoration strategies more optimized. It can effectively improve the accuracy of fault handling in medium and low voltage power grids and the efficiency and stability of power restoration, maximizing the continuity of power supply.

[0077] In one embodiment, electrical data includes three-phase current and three-phase voltage, and non-electrical data includes equipment temperature, partial discharge pulse signals, and mechanical vibration signals; extracting electrical and non-electrical features from the multimodal operation data includes:

[0078] Based on continuously sampled three-phase current and three-phase voltage, the current and voltage differences between adjacent sampling points and the quadratic difference between adjacent differences are calculated; the root mean square values ​​of current and voltage within a time window are calculated by sliding through a time window; the energy values ​​of current and voltage, and the approximate entropy of current and voltage used to characterize the degree of numerical disorder are calculated; the differences, quadratic differences, root mean square values, energy values, and approximate entropy of current and voltage are used as electrical characteristics.

[0079] Based on continuously sampled device temperatures, the temperature difference and root mean square (RMS) value are calculated; based on partial discharge pulse signals, the maximum pulse amplitude, the number of pulses per second, and the sum of squares of pulse amplitude are obtained; based on mechanical vibration signals, wavelet transform is performed to obtain the decomposed high-frequency peak value, the sum of squares of vibration signals, and the approximate entropy of vibration signals; the temperature difference and RMS value, the maximum pulse amplitude, the number of pulses per second, the sum of squares of pulse amplitude, the high-frequency peak value, the sum of squares of vibration signals, and the approximate entropy of vibration signals are used as non-electrical features.

[0080] Specifically, electrical data includes three-phase current and three-phase voltage, while non-electrical data includes equipment temperature, partial discharge pulse signals, and mechanical vibration signals. For electrical data, based on continuously sampled three-phase current and three-phase voltage, the differences between current and voltage at adjacent sampling points and the quadratic differences between adjacent differences are first calculated. Then, the root mean square (RMS) values ​​of current and voltage within the window are obtained by sliding through a time window. Simultaneously, the energy values ​​of current and voltage and the approximate entropy representing the degree of numerical disorder are calculated. The above differences, quadratic differences, RMS values, energy values, and approximate entropy are integrated into electrical features. For non-electrical data, the differences and RMS values ​​of continuously sampled equipment temperature are calculated. The maximum pulse amplitude, number of pulses per second, and sum of squared pulse amplitudes are extracted from the partial discharge pulse signals. Wavelet transform is performed on the mechanical vibration signals to obtain the decomposed high-frequency peak values, the sum of squared vibration signals, and the approximate entropy of vibration signals. Finally, the temperature-related indicators, partial discharge pulse-related indicators, and wavelet-transformed indicators of the mechanical vibration signals are integrated into non-electrical features, completing the feature extraction of multimodal operation data.

[0081] In the above embodiments, by clearly defining the specific types of electrical and non-electrical data, the collection of multimodal data becomes more targeted, accurately covering the core monitoring dimensions of the power grid's electrical operating status and the equipment's physical operating status. In the feature extraction stage, electrical data is quantitatively analyzed from multiple dimensions, including difference changes, statistical features, energy features, and complexity features. Non-electrical data is specifically extracted by combining dynamic changes in equipment temperature, pulse features of partial discharge, and wavelet decomposition features of mechanical vibration. The resulting electrical and non-electrical features can comprehensively, accurately, and quantitatively characterize the electrical parameter anomalies and equipment physical state anomalies caused by power grid faults, effectively avoiding the one-sidedness of feature extraction. This provides highly identifiable and highly relevant feature basis for subsequent anomaly mutation feature identification and fault diagnosis, significantly improving the accuracy and reliability of fault feature characterization.

[0082] In one embodiment, based on the comparison results between electrical features in the monitoring dataset and electrical features in the benchmark dataset, and the comparison results between non-electrical features in the monitoring dataset and non-electrical features in the benchmark dataset, anomalous mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset, including:

[0083] For each feature, obtain the monitoring mean of the corresponding type of feature in the monitoring dataset and the baseline mean of the corresponding type of feature in the benchmark dataset; calculate the difference between the monitoring mean and the benchmark mean; obtain the product of the standard deviation of the corresponding type of feature in the benchmark dataset and a preset multiple as a preset threshold; if the difference is greater than the preset threshold, the corresponding type of feature in the monitoring dataset is regarded as an abnormal mutation feature.

[0084] Specifically, when identifying anomalous mutation features in multimodal features, a method based on statistical feature threshold determination is adopted. For each type of extracted electrical and non-electrical feature, the monitoring mean in the monitoring dataset and the benchmark mean in the benchmark dataset are calculated respectively, and then the difference between the two is solved. At the same time, the product of the standard deviation of the feature in the benchmark dataset and a preset multiple is calculated, and this product is used as a preset threshold for determining whether the feature is anomalous. Finally, the anomaly determination is completed by comparing the thresholds. If the above difference is greater than the preset threshold, the corresponding electrical or non-electrical feature in the monitoring dataset is identified as an anomalous mutation feature, thereby achieving standardized and quantitative anomaly identification for various features.

[0085] In the above embodiments, a statistical judgment method combining the mean difference with a preset threshold based on the standard deviation of the benchmark is used to identify abnormal mutation features. The personalized judgment threshold is determined by the statistical features of the benchmark dataset, which is adapted to the normal fluctuation characteristics of different features and avoids the one-sidedness and misjudgment of fixed threshold judgment. At the same time, independent quantitative calculation and threshold comparison are performed for each type of electrical and non-electrical feature, realizing the standardized and refined identification of abnormal mutation features, effectively improving the accuracy and objectivity of feature anomaly judgment, and accurately screening feature mutation information caused by power grid faults.

[0086] In one embodiment, fault isolation is performed based on a fault feature set and fault information, including:

[0087] Based on the fault location in the fault feature set and fault information, the isolated forest algorithm is used to analyze the time-series power grid operation dataset composed of the benchmark dataset and the monitoring dataset to obtain the fault feeder or fault node.

[0088] To isolate the fault, the circuit breaker of the faulty feeder or faulty node is tripped.

[0089] Specifically, when carrying out fault isolation operations, the fault location in the fault feature set and fault information is first used as the core analysis basis. The benchmark dataset and monitoring dataset are integrated into a time-series power grid operation dataset. The isolated forest algorithm is used to perform anomaly mining and analysis on this dataset to accurately locate the faulty feeder or faulty node in the power grid. Then, for the sectional circuit breaker corresponding to the identified faulty feeder or faulty node, a tripping operation command is issued. By executing the tripping action of the sectional circuit breaker, the faulty feeder or faulty node is electrically isolated from the normally operating power grid system, thus completing the fault isolation operation process.

[0090] In the above embodiments, fault feature sets, fault location information, and the isolated forest algorithm are combined to carry out fault localization. Relying on the high sensitivity and strong mining ability of the isolated forest algorithm to abnormal data, combined with time-series power grid operation data, it can further accurately locate the faulty feeder or faulty node based on the fusion of multi-source fault information, thereby improving the accuracy and reliability of fault localization. At the same time, the tripping operation is directly performed on the section circuit breaker corresponding to the fault point to achieve precise isolation of the fault point, avoid the expansion of the power outage area caused by indiscriminate isolation, and preserve the power grid connectivity of non-faulty areas to the maximum extent.

[0091] In one embodiment, a fusion control command is generated with the goal of maximizing load recovery and achieving the most stable voltage, including:

[0092] Calculate the fault probability based on the fault feature set and the fault type in the fault information;

[0093] Based on the failure probability, the weight coefficients are calculated; the quantization parameters corresponding to the reinforcement learning instructions for load recovery and the quantization parameters corresponding to the model predictive control instructions for voltage stabilization are obtained; the quantization parameters include output power and switch closure degree.

[0094] Based on the weight coefficients, the quantization parameters corresponding to the reinforcement learning instructions and the quantization parameters corresponding to the model prediction control instructions are weighted and summed to obtain the fused quantization parameters.

[0095] Based on the fused quantization parameters, a fused control command is generated.

[0096] Specifically, based on the fault feature set and fault types in the fault information, the corresponding fault probability is calculated through quantitative analysis. Based on machine learning, it is used to measure and receive power grid equipment operation data and medium-voltage fault information in real time. Faulty feeders or nodes are identified through anomaly detection or classification algorithms, and a rapid isolation operation is triggered on the low-voltage side. Reinforcement learning (RL) and model predictive control (MPC) are integrated and fused to learn and generate a recovery strategy for the low-voltage distribution network after the fault is isolated. The recovery strategy includes coordinating the output of distributed generation (DG) and energy storage system (ESS), as well as adjusting the low-voltage switch state to restore power supply.

[0097] The fusion control command of RL and MPC is generated by the following formula:

[0098]

[0099] In the formula, To reinforce the control commands generated by learning, Control commands generated for model predictive control, weighting coefficients The calculation formula is as follows: The adjustment is made dynamically based on the severity of the fault.

[0100]

[0101] In the formula, This is the sensitivity coefficient. This represents the probability of failure.

[0102] In the above embodiments, the fault probability is calculated and the weight coefficient is determined based on the fault feature set and fault type, so that the weight allocation is accurately matched with the actual fault conditions, avoiding the adaptability problem of fixed weights. By integrating the dynamic optimization advantages of reinforcement learning instructions in load restoration and the precise regulation advantages of model predictive control instructions in voltage stability, the core quantitative parameters of the two types of instructions are weighted and integrated. The generated fused control instructions can simultaneously take into account the dual objectives of maximizing load restoration and voltage stability, avoiding the disadvantages of single-objective regulation, effectively improving the regulation accuracy and operational stability in the power grid fault restoration process, and ensuring the overall effect of power supply restoration.

[0103] In one embodiment, a cross-level recovery strategy and fusion control commands are executed to restore power supply, including:

[0104] Close the circuit breaker of the medium-voltage standby line to reconfigure the medium- and low-voltage power grid topology;

[0105] Power supply restoration is carried out according to the quantization parameters in the fusion control command, the switching operation sequence in the cross-level recovery strategy, the output limit of the distributed power source, and the order of load restoration.

[0106] Specifically, by closing the circuit breakers of the medium-voltage backup lines, the topology of the medium- and low-voltage power grid after a fault is reconstructed, establishing an effective physical channel for load transfer and power output, laying the topological foundation for subsequent cross-level power restoration. Subsequently, on the reconstructed power grid topology, the quantitative control parameters such as output power and switch closure degree in the integrated control commands are strictly followed. At the same time, in accordance with the sequence of switching operations specified in the cross-level restoration strategy, the maximum output limit of distributed power sources, and the load restoration priority divided by importance, various power restoration operations are executed step by step and according to rules, so as to achieve coordinated power restoration at the medium-voltage and low-voltage power grid levels, taking into account the efficiency, stability and load restoration maximization goals of power restoration.

[0107] In the above embodiments, by combining the quantized parameters of the fusion control command with the operation rules, output limits, and load priorities of the cross-level restoration strategy to perform power restoration, the operation process has both precise parameter control basis and clear execution sequence and resource constraint specifications, realizing cross-level collaborative restoration of medium and low voltage power grids. This not only ensures the stability of parameters such as voltage and power during the power restoration process, but also maximizes the restoration of load according to priority, effectively reducing power outage losses.

[0108] In one embodiment, the method further includes:

[0109] From a time-series power grid operation dataset consisting of a benchmark dataset and a monitoring dataset, we extract equipment temperature time-series, partial discharge cumulative energy, and vibration frequency domain features, and perform preprocessing.

[0110] Based on the pre-processed equipment temperature time series, partial discharge cumulative energy, and vibration frequency domain characteristics, the remaining lifespan of the equipment is predicted, and corresponding levels of early warning are issued based on the prediction results.

[0111] Specifically, from the time-series power grid operation dataset integrated from the benchmark dataset and the monitoring dataset, three types of equipment status-related features are extracted: equipment temperature time series, partial discharge cumulative energy, and vibration frequency domain features. These features are then preprocessed, including data cleaning, normalization, and noise reduction. The preprocessed features are then used as input to quantify and predict the remaining lifespan of the power grid equipment using a corresponding lifespan prediction model. Finally, based on the predicted remaining lifespan, corresponding level equipment status warnings are issued according to preset warning level standards, thereby achieving full lifespan status monitoring and early warning for the equipment.

[0112] In the above embodiments, the time-series multimodal data of power grid operation is combined with equipment remaining life prediction and hierarchical early warning. Based on the core state characteristics of equipment such as temperature, partial discharge, and vibration, the remaining life is accurately predicted, breaking through the passive mode of traditional post-fault handling and realizing active monitoring and early warning of power grid equipment status. Different levels of early warning allow maintenance personnel to carry out targeted equipment inspection and maintenance work, effectively avoiding power grid faults caused by equipment aging and performance degradation, reducing the probability of fault occurrence from the source, and improving the planning and scientific nature of equipment operation and maintenance, reducing the number of unplanned power outages.

[0113] In one embodiment, such as Figure 3 The image shows a self-healing control method for a medium- and low-voltage power grid in a specific embodiment, including:

[0114] A baseline dataset consisting of multimodal operational data collected before the fault and a monitoring dataset consisting of multimodal operational data collected after the fault are acquired. The multimodal operational data includes electrical and non-electrical data, such as current, voltage, temperature, partial discharge, and mechanical vibration. Electrical and non-electrical features are extracted from the multimodal operational data. Based on comparisons between the electrical features in the monitoring dataset and those in the baseline dataset, and between the non-electrical features in the monitoring dataset and those in the baseline dataset, abnormal abrupt changes are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0115] Fault characteristic analysis includes time-domain analysis, frequency-domain analysis, energy analysis, and entropy analysis. In time-domain analysis, the first and second derivatives of voltage and current are calculated to quickly capture abrupt changes in information when a fault occurs. It is assumed that the rate of rise of fault current (di / dt) and the rate of drop of voltage (dv / dt) are important fault characteristics. The formula for calculating the rate of change of current is as follows:

[0116] (1)

[0117] In the formula, Sampling time The current value;

[0118] The second derivative of the current is expressed as follows:

[0119] (2)

[0120] RMS value analysis: Calculates the RMS values ​​of voltage and current within a short time window, detecting abnormal drops or increases in the RMS value, as shown below:

[0121] (3)

[0122] In the formula, This represents the number of samples within the sampling window. For the first time within the short sampling window The instantaneous voltage value collected at each sampling point.

[0123] Monitor the peak values ​​of voltage and current; the peak values ​​will change significantly during a fault.

[0124] Frequency domain analysis uses wavelet transform to extract fault-related transient components. Energy analysis is used to calculate the energy of the fault signal within a specific time window to determine whether a fault has occurred. Entropy analysis measures the randomness or disorder of a signal; when a fault occurs, the entropy of the signal increases significantly.

[0125] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0126] This paper analyzes the centrality, connectivity, reachability, and load flow of power grid nodes, constructs a power grid network model including the topology, node and edge attributes of medium- and low-voltage power grids, and analyzes the network attributes. Based on the power grid network model, fault information, and preset optimization objectives (including but not limited to minimizing power outage losses and maximizing restored load), a mixed-integer programming (MIP) model is used. The variables of the MIP model include switch state variables, load redistribution variables, distributed generation (DG) / energy storage system (ESS) power dispatch variables, and restoration sequence variables, and include KCL / KVL constraints, line capacity constraints, node voltage constraints, DG / ESS output constraints, and restoration logic constraints. An optimal restoration strategy spanning medium- and low-voltage levels is generated. This strategy includes switch operation sequences, power dispatch of distributed generation (DG) and energy storage system (ESS), and load redistribution schemes, driving the power grid's automated equipment to perform fault isolation and restoration operations. The objective function of the MIP model is expressed as follows, where the constraints include radial topology constraints, distributed generation capacity limits, and switch action time limits:

[0127] (4)

[0128] In the formula, For the line The operational status (1 is closed); For load Priority weights; Recovery time (positively correlated with the number of switching actions); For distributed power sources The startup status; For distributed power sources Startup costs, All are weighting coefficients. It is the collection of all load lines in the power grid. It is the set of all distributed power sources in the power grid.

[0129] Generating an optimal recovery strategy across medium and low voltage levels further includes first isolating the level where the fault occurs, then generating a reconfiguration strategy at the medium voltage level, and coordinating distributed DG / ESS scheduling and feeder reconfiguration at the low voltage level to achieve rapid recovery of the entire power grid.

[0130] Based on machine learning, it is used to measure and receive power grid equipment operation data and medium-voltage fault information in real time. It identifies faulty feeders or nodes through anomaly detection or classification algorithms and triggers rapid isolation operations on the low-voltage side. It integrates and fuses reinforcement learning (RL) and model predictive control (MPC) to learn and generate recovery strategies for the low-voltage distribution network after fault isolation. The recovery strategies include coordinating the output of distributed generation (DG) and energy storage system (ESS), and adjusting the low-voltage switch states to restore power supply.

[0131] The fusion control command of RL and MPC is generated by the following formula:

[0132] (5)

[0133] In the formula, To reinforce the control commands generated by learning, Control commands generated for model predictive control, weighting coefficients The calculation formula is as follows: The adjustment is made dynamically based on the severity of the fault.

[0134] (6)

[0135] In the formula, This is the sensitivity coefficient. This represents the probability of failure.

[0136] Based on Temporal-CNN-LSTM, the remaining lifespan of equipment is predicted and preventative maintenance is triggered. The predicted remaining lifespan information and early warning signals are provided to collaborative self-healing control to assess future fault risks and optimize grid operation. The grid topology is updated in real time based on a graph neural network (GNN), with node embeddings as shown below, and the control strategy is dynamically adjusted according to topology changes.

[0137] (7)

[0138] In the formula, For activation function, For the first Layer nodes Embedded vector, For layer weights, For nodes The set of adjacent nodes, These are the adjacent nodes of the node.

[0139] The remaining life prediction process collects a time-series power grid operation dataset comprised of a baseline dataset and a monitoring dataset. This dataset includes, but is not limited to, data on temperature, vibration, electrical parameters, partial discharge signals, historical equipment maintenance records, and equipment specifications. Temporal temperature characteristics, cumulative partial discharge energy, and frequency domain vibration features are extracted and preprocessed. Preprocessing includes data cleaning, feature engineering (such as time window statistical features, frequency domain features, and time-frequency domain features), and data normalization. Based on the preprocessed data, a Temporal-CNN-LSTM model is trained to predict the remaining lifespan of the equipment. The predicted remaining lifespan is compared to a preset warning threshold. When the predicted lifespan falls below the threshold, a preventative maintenance warning is triggered. The warning threshold is dynamically set based on the equipment type, importance, and acceptable risk level. The preventative maintenance warning information includes the predicted remaining lifespan, prediction reliability, and abnormal features leading to shortened lifespan.

[0140] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0141] Based on the same inventive concept, this application also provides a self-healing control device for medium and low voltage power grids to implement the self-healing control method for medium and low voltage power grids described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the self-healing control device for medium and low voltage power grids provided below can be found in the limitations of the self-healing control method for medium and low voltage power grids described above, and will not be repeated here.

[0142] In one exemplary embodiment, such as Figure 4 As shown, a self-healing control device for medium and low voltage power grids is provided, comprising: an acquisition module 402, an extraction module 404, an input and output module 406, a construction and solution module 408, and a control module 410, wherein:

[0143] The acquisition module 402 is used to acquire a baseline dataset consisting of multimodal operating data collected before the fault and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0144] Extraction module 404 is used to extract electrical and non-electrical features from multimodal operation data; based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set;

[0145] The input and output module 406 is used to input the fault feature set into the trained fault classification model and output the corresponding fault information, including fault type, fault location and fault impact range.

[0146] The construction and solution module 408 is used to construct a mixed integer programming model with the goal of minimizing power outage losses and maximizing restored load. The fault information and the power grid network model are substituted into the mixed integer programming model for solution to obtain the cross-level restoration strategy. The power grid network model includes the topology of medium and low voltage power grids, line parameters and node load levels.

[0147] The control module 410 is used to isolate faults based on fault feature sets and fault information, and generate fusion control commands aimed at restoring the maximum load and the most stable voltage; it executes cross-level recovery strategies and fusion control commands to restore power supply.

[0148] In one embodiment, the acquisition module 402 is further configured to:

[0149] Based on continuously sampled three-phase current and three-phase voltage, the current and voltage differences between adjacent sampling points and the quadratic difference between adjacent differences are calculated; the root mean square values ​​of current and voltage within a time window are calculated by sliding through a time window; the energy values ​​of current and voltage, and the approximate entropy of current and voltage used to characterize the degree of numerical disorder are calculated; the differences, quadratic differences, root mean square values, energy values, and approximate entropy of current and voltage are used as electrical characteristics.

[0150] Based on continuously sampled device temperatures, the temperature difference and root mean square (RMS) value are calculated; based on partial discharge pulse signals, the maximum pulse amplitude, the number of pulses per second, and the sum of squares of pulse amplitude are obtained; based on mechanical vibration signals, wavelet transform is performed to obtain the decomposed high-frequency peak value, the sum of squares of vibration signals, and the approximate entropy of vibration signals; the temperature difference and RMS value, the maximum pulse amplitude, the number of pulses per second, the sum of squares of pulse amplitude, the high-frequency peak value, the sum of squares of vibration signals, and the approximate entropy of vibration signals are used as non-electrical features.

[0151] In one embodiment, the extraction module 404 is further configured to:

[0152] For each feature, obtain the monitoring mean of the corresponding type of feature in the monitoring dataset and the baseline mean of the corresponding type of feature in the benchmark dataset; calculate the difference between the monitoring mean and the benchmark mean; obtain the product of the standard deviation of the corresponding type of feature in the benchmark dataset and a preset multiple as a preset threshold; if the difference is greater than the preset threshold, the corresponding type of feature in the monitoring dataset is regarded as an abnormal mutation feature.

[0153] In one embodiment, the control module 410 is further configured to:

[0154] Based on the fault location in the fault feature set and fault information, the isolated forest algorithm is used to analyze the time-series power grid operation dataset composed of the benchmark dataset and the monitoring dataset to obtain the fault feeder or fault node.

[0155] To isolate the fault, the circuit breaker of the faulty feeder or faulty node is tripped.

[0156] In one embodiment, the control module 410 is further configured to:

[0157] Calculate the fault probability based on the fault feature set and the fault type in the fault information;

[0158] Based on the failure probability, the weight coefficients are calculated; the quantization parameters corresponding to the reinforcement learning instructions for load recovery and the quantization parameters corresponding to the model predictive control instructions for voltage stabilization are obtained; the quantization parameters include output power and switch closure degree.

[0159] Based on the weight coefficients, the quantization parameters corresponding to the reinforcement learning instructions and the quantization parameters corresponding to the model prediction control instructions are weighted and summed to obtain the fused quantization parameters.

[0160] Based on the fused quantization parameters, a fused control command is generated.

[0161] In one embodiment, the control module 410 is further configured to:

[0162] Close the circuit breaker of the medium-voltage standby line to reconfigure the medium- and low-voltage power grid topology;

[0163] Power supply restoration is carried out according to the quantization parameters in the fusion control command, the switching operation sequence in the cross-level recovery strategy, the output limit of the distributed power source, and the order of load restoration.

[0164] In one embodiment, the method further includes:

[0165] From a time-series power grid operation dataset consisting of a benchmark dataset and a monitoring dataset, we extract equipment temperature time-series, partial discharge cumulative energy, and vibration frequency domain features, and perform preprocessing.

[0166] Based on the pre-processed equipment temperature time series, partial discharge cumulative energy, and vibration frequency domain characteristics, the remaining lifespan of the equipment is predicted, and corresponding levels of early warning are issued based on the prediction results.

[0167] Each module in the aforementioned self-healing control device for medium and low voltage power grids can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0168] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a self-healing control method for medium- and low-voltage power grids. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0169] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0170] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0171] Acquire a baseline dataset consisting of multimodal operating data collected before the fault, and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0172] Electrical and non-electrical features are extracted from multimodal operation data. Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0173] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0174] To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. Fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0175] Fault isolation is performed based on fault feature sets and fault information, and fusion control commands are generated with the goal of restoring the most load and the most stable voltage. Cross-level recovery strategies and fusion control commands are executed to restore power supply.

[0176] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0177] Acquire a baseline dataset consisting of multimodal operating data collected before the fault, and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0178] Electrical and non-electrical features are extracted from multimodal operation data. Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0179] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0180] To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. Fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0181] Fault isolation is performed based on fault feature sets and fault information, and fusion control commands are generated with the goal of restoring the most load and the most stable voltage. Cross-level recovery strategies and fusion control commands are executed to restore power supply.

[0182] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0183] Acquire a baseline dataset consisting of multimodal operating data collected before the fault, and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data;

[0184] Electrical and non-electrical features are extracted from multimodal operation data. Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set.

[0185] The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range.

[0186] To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. Fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain cross-level recovery strategies. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels.

[0187] Fault isolation is performed based on fault feature sets and fault information, and fusion control commands are generated with the goal of restoring the most load and the most stable voltage. Cross-level recovery strategies and fusion control commands are executed to restore power supply.

[0188] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0189] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0190] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0191] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A self-healing control method for medium and low voltage power grids, characterized in that, The method includes: A baseline dataset consisting of multimodal operating data collected before the fault and a monitoring dataset consisting of multimodal operating data collected after the fault are acquired; the multimodal operating data includes electrical data and non-electrical data. Electrical and non-electrical features are extracted from the multimodal operation data; based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set; The fault feature set is input into the trained fault classification model, and the corresponding fault information is output, including fault type, fault location and fault impact range. To minimize power outage losses and maximize load recovery, a mixed-integer programming model is constructed. The fault information and the power grid network model are substituted into the mixed-integer programming model for solution to obtain a cross-level recovery strategy. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels. Based on the fault feature set and the fault information, fault isolation is performed, and a fusion control command is generated with the goal of restoring the maximum load and the most stable voltage. The cross-level recovery strategy and the fusion control command are executed to restore power supply.

2. The method according to claim 1, characterized in that, The electrical data includes three-phase current and three-phase voltage, and the non-electrical data includes equipment temperature, partial discharge pulse signals, and mechanical vibration signals; the extraction of electrical and non-electrical features from the multimodal operation data includes: Based on continuously sampled three-phase current and three-phase voltage, the current and voltage differences between adjacent sampling points and the quadratic difference between adjacent differences are calculated; the root mean square values ​​of current and voltage within the time window are calculated by sliding through the time window; the energy values ​​of current and voltage and the approximate entropy of current and voltage used to characterize the degree of numerical disorder are calculated; the differences between current and voltage, the quadratic difference, the root mean square value, the energy value, and the approximate entropy are used as electrical characteristics. Based on continuously sampled device temperatures, the temperature difference and root mean square (RMS) value are calculated; based on the partial discharge pulse signal, the maximum pulse amplitude, the number of pulses per second, and the sum of squares of pulse amplitude are obtained; based on the mechanical vibration signal, wavelet transform is performed to obtain the decomposed high-frequency peak value, the sum of squares of the vibration signal, and the approximate entropy of the vibration signal; the temperature difference and RMS value, the maximum pulse amplitude, the number of pulses per second, the sum of squares of pulse amplitude, the high-frequency peak value, the sum of squares of the vibration signal, and the approximate entropy of the vibration signal are used as non-electrical features.

3. The method according to claim 2, characterized in that, Based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset, including: For each feature, the monitoring mean of the corresponding type of feature in the monitoring dataset and the benchmark mean of the corresponding type of feature in the benchmark dataset are obtained; the difference between the monitoring mean and the benchmark mean is calculated; the product of the standard deviation of the corresponding type of feature in the benchmark dataset and a preset multiple is obtained as a preset threshold; if the difference is greater than the preset threshold, the corresponding type of feature in the monitoring dataset is regarded as an abnormal mutation feature.

4. The method according to claim 1, characterized in that, The fault isolation based on the fault feature set and the fault information includes: Based on the fault feature set and the fault location in the fault information, the isolated forest algorithm is used to analyze the time-series power grid operation dataset composed of the benchmark dataset and the monitoring dataset to obtain the fault feeder or fault node. The circuit breaker of the faulty feeder or the faulty node is tripped to isolate the fault.

5. The method according to claim 1, characterized in that, The generation of fusion control commands aimed at maximizing load recovery and achieving the most stable voltage includes: Calculate the fault probability based on the fault feature set and the fault type in the fault information; Based on the fault probability, weighting coefficients are calculated; quantization parameters corresponding to the reinforcement learning instructions for load recovery and the model predictive control instructions for voltage stabilization are obtained; the quantization parameters include output power and switch closure degree. Based on the weight coefficients, the quantization parameters corresponding to the reinforcement learning instructions and the quantization parameters corresponding to the model prediction control instructions are weighted and summed to obtain the fused quantization parameters. Based on the fused quantization parameters, a fused control command is generated.

6. The method according to claim 5, characterized in that, The execution of the cross-level recovery strategy and fusion control instructions for power restoration includes: Close the circuit breaker of the medium-voltage standby line to reconfigure the medium- and low-voltage power grid topology; Power supply restoration is performed according to the quantization parameters in the fusion control command, the switching operation sequence in the cross-level recovery strategy, the output limit of the distributed power source, and the order of load restoration.

7. The method according to claim 1, characterized in that, The method further includes: From the time-series power grid operation dataset composed of the benchmark dataset and the monitoring dataset, extract the equipment temperature time series, partial discharge cumulative energy and vibration frequency domain features, and perform preprocessing. Based on the pre-processed equipment temperature time series, partial discharge cumulative energy, and vibration frequency domain characteristics, the remaining lifespan of the equipment is predicted, and corresponding levels of early warning are issued based on the prediction results.

8. A self-healing control device for medium and low voltage power grids, characterized in that, The device includes: The acquisition module is used to acquire a baseline dataset consisting of multimodal operating data collected before the fault and a monitoring dataset consisting of multimodal operating data collected after the fault; the multimodal operating data includes electrical data and non-electrical data. An extraction module is used to extract electrical and non-electrical features from the multimodal operation data; based on the comparison results between the electrical features in the monitoring dataset and the electrical features in the benchmark dataset, and the comparison results between the non-electrical features in the monitoring dataset and the non-electrical features in the benchmark dataset, abnormal mutation features are identified from the corresponding electrical and non-electrical features in the monitoring dataset to form a fault feature set; The input and output module is used to input the fault feature set into the trained fault classification model and output the corresponding fault information, which includes fault type, fault location and fault impact range. The construction and solution module is used to construct a mixed integer programming model with the goal of minimizing power outage losses and maximizing load recovery. The fault information and the power grid network model are substituted into the mixed integer programming model for solution to obtain a cross-level recovery strategy. The power grid network model includes the topology of medium and low voltage power grids, line parameters, and node load levels. The control module is used to isolate faults based on the fault feature set and the fault information, and generate fusion control commands aimed at restoring the maximum load and the most stable voltage; and execute the cross-level recovery strategy and fusion control commands to restore power supply.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images