Power dispatching method, device and equipment for power distribution area, storage medium and program product

Abstract

The application relates to a power dispatching method, device and equipment for a power distribution area, a storage medium and a program product. The method comprises the following steps: collecting electric signal data of the power distribution area, and performing data preprocessing to obtain a data feature set; a substation area topology graph is constructed according to the node relationship of the power distribution area; according to the data feature set, the substation area topology graph and a multi-node cooperation strategy, an initial fault section, a fault path edge coordinate and a node weight vector of the power distribution area are obtained; according to the initial fault section, the fault path edge coordinate and the node weight vector, a target fault section of the power distribution area is determined; according to the hardware constraint condition and the target fault section of the power distribution area, a fault area of the power distribution area is isolated and cut off, and power dispatching is performed on the power distribution area after isolation and cutting off. The method effectively reduces the number of power-off users and the power-off capacity, and provides key technical support for the rapid self-healing and power supply reliability improvement of the power distribution network.

Description

Technical Field

[0001] This application relates to the field of power grid technology, and in particular to a power dispatching method, apparatus, equipment, storage medium, and program product for a distribution substation. Background Technology

[0002] As distribution networks accelerate their digitalization and intelligentization, distribution substations, as the fundamental units for intelligent sensing and active control in the power system, are playing an increasingly important role in supporting precise grid operation and ensuring power supply reliability. Currently, distribution substation automation systems typically consist of distribution terminals, communication links, and a dispatch master station, enabling basic functions such as fault detection, isolation, and power restoration. Fault location is a crucial prerequisite for quickly identifying fault points to reduce fault propagation and plays a key role in improving fault handling efficiency. Post-fault power dispatch is the core link in ensuring continuous power supply to non-faulty areas and optimizing grid resource allocation. Compared to fault location, post-fault power dispatch requires more dynamic and collaborative decision-making by combining real-time load data, distributed power generation output, and the operating status of energy storage devices.

[0003] However, existing distribution substation automation systems mostly rely on fixed rules or preset scenario tables for power dispatch after a fault, which makes it difficult to meet the high requirements of digital smart distribution networks for power supply continuity and operational economy after fault handling.

[0004] Therefore, how to efficiently and accurately dispatch power after a distribution substation fault based on the real-time operating status of the distribution network has become an urgent problem to be solved. Summary of the Invention

[0005] Therefore, it is necessary to provide a power dispatching method, apparatus, equipment, storage medium, and program product for a distribution substation to address the aforementioned technical problems.

[0006] Firstly, this application provides a power dispatching method for a distribution substation, including:

[0007] Electrical signal data from the distribution transformer substation is collected and preprocessed to obtain a data feature set. The data feature set is used to characterize the electrical response characteristics of each monitoring node in the distribution transformer substation to fault events.

[0008] Based on the node relationships of the distribution transformer areas, a transformer area topology map is constructed;

[0009] Based on the data feature set, the distribution area topology map, and the multi-node collaborative strategy, the initial fault section, fault path edge coordinates, and weight vectors of each node in the distribution area are obtained. The multi-node collaborative strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model. The graph-constrained linear model is used to characterize the linear mapping relationship between the time-series observations of each monitoring node and the distribution area topology map. The waveform similarity network model is used to characterize the correlation between the transient waveforms of the faults of each monitoring node. The node credibility prior model is used to quantify the credibility level of each monitoring node.

[0010] The target fault section of the distribution transformer area is determined based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node.

[0011] Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cleared, and the isolated distribution transformer area is used for power dispatch.

[0012] In one embodiment, based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy, the initial fault segment, fault path edge coordinates, and weight vectors of each node of the distribution transformer area are obtained, including:

[0013] The data feature set is input into the single-node initial value model to estimate the distance to each node in the transformer area topology map, thereby obtaining the initial position of each node from the initial fault point. The single-node initial value model includes impedance mutation unit and transient energy center unit. The impedance mutation unit and transient energy center unit operate in parallel.

[0014] Based on the data feature set, multi-node collaboration strategy, and initial location, the fault point is calculated to obtain the initial fault segment, fault path edge coordinates, and weight vector of each node in the transformer area topology map.

[0015] In one embodiment, the fault point is calculated based on the data feature set, multi-node collaboration strategy, and initial location to obtain the initial fault segment, fault path edge coordinates, and weight vectors of each node in the transformer substation topology map, including:

[0016] The coordinates of the fault path edges are obtained by solving the data feature set using a graph-constrained linear model.

[0017] Fault waveform analysis is performed based on a waveform similarity network model to obtain the initial fault segment;

[0018] The credibility of each node in the transformer substation topology is weighted according to the prior model of node credibility, and the weight vector of each node is obtained.

[0019] In one embodiment, the target fault section of the distribution transformer area is determined based on the initial fault section, the coordinates of the fault path edges, and the weight vectors of each node, including:

[0020] The initial location and fault path edge coordinates are scaled to obtain the initial fault segment set;

[0021] The target fault section is obtained by calculating based on the initial set of fault sections and the weight vector of each node.

[0022] In one embodiment, based on the hardware constraints of the distribution transformer area and the target fault section, the faulty area of ​​the distribution transformer area is isolated and cleared, and power dispatch is performed using the isolated distribution transformer area, including:

[0023] The scheduling scheme is boundary-defined based on hardware constraints, and a scheduling model is constructed.

[0024] The target faulty section is input into the scheduling model, and the faulty area of ​​the distribution radio area is isolated and removed to obtain the scheduling scheme.

[0025] In one embodiment, the method further includes:

[0026] The target scheduling scheme is determined based on the communication operation status of the distribution radio area after isolation and disconnection; the communication operation status includes any one of the following: normal state, degraded state, and autonomous state;

[0027] When the communication operation status of the distribution radio area is normal or degraded, power dispatch is carried out in accordance with the first dispatching scheme using the distribution radio area after isolation and disconnection.

[0028] When the communication operation status of the distribution radio area is autonomous, power dispatch is carried out according to the second dispatch scheme using the distribution radio area after isolation and disconnection.

[0029] Secondly, this application also provides a fault dispatch control device for a distribution radio area, comprising:

[0030] The preprocessing module is used to collect electrical signal data from the distribution transformer area and perform data preprocessing to obtain a set of data features.

[0031] The topology module is used to construct a topology map of the distribution area based on the node relationships of the distribution area.

[0032] The collaboration module is used to obtain the initial fault section, fault path edge coordinates, and weight vector of each node of the distribution transformer area based on the data feature set, transformer area topology map, and multi-node collaboration strategy.

[0033] The fusion module is used to determine the target fault section of the distribution radio area based on the initial fault section, the edge coordinates of the fault path, and the weight vector of each node.

[0034] The scheduling module is used to isolate and remove faulty areas of the distribution radio station based on the hardware constraints and target faulty sections, and to use the isolated distribution radio station for power dispatch.

[0035] 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:

[0036] Collect electrical signal data from the distribution radio station area and perform data preprocessing to obtain a set of data features;

[0037] Based on the node relationships of the distribution transformer areas, a transformer area topology map is constructed;

[0038] Based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy, the initial fault section, fault path edge coordinates, and weight vectors of each node in the distribution transformer area are obtained. The multi-node collaboration strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model. The graph-constrained linear model, the waveform similarity network model, and the node credibility prior model are run in parallel.

[0039] The target fault section of the distribution transformer area is determined based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node.

[0040] Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cleared, and the isolated distribution transformer area is used for power dispatch.

[0041] 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:

[0042] Collect electrical signal data from the distribution radio station area and perform data preprocessing to obtain a set of data features;

[0043] Based on the node relationships of the distribution transformer areas, a transformer area topology map is constructed;

[0044] Based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy, the initial fault section, fault path edge coordinates, and weight vectors of each node in the distribution transformer area are obtained. The multi-node collaboration strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model. The graph-constrained linear model, the waveform similarity network model, and the node credibility prior model are run in parallel.

[0045] The target fault section of the distribution transformer area is determined based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node.

[0046] Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cleared, and the isolated distribution transformer area is used for power dispatch.

[0047] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0048] Collect electrical signal data from the distribution radio station area and perform data preprocessing to obtain a set of data features;

[0049] Based on the node relationships of the distribution transformer areas, a transformer area topology map is constructed;

[0050] Based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy, the initial fault section, fault path edge coordinates, and weight vectors of each node in the distribution transformer area are obtained. The multi-node collaboration strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model. The graph-constrained linear model, the waveform similarity network model, and the node credibility prior model are run in parallel.

[0051] The target fault section of the distribution transformer area is determined based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node.

[0052] Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cleared, and the isolated distribution transformer area is used for power dispatch.

[0053] The aforementioned power dispatching methods, devices, equipment, storage media, and program products for distribution transformer substations acquire electrical signal data from the substations and perform data preprocessing to obtain a data feature set; construct a transformer substation topology map based on the node relationships within the substations; and obtain the initial fault segment, fault path edge coordinates, and node weight vectors of the distribution transformer substations based on the data feature set, the substation topology map, and a multi-node collaborative strategy. The multi-node collaborative strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model. The graph-constrained linear model characterizes the linear mapping relationship between the time-series observations of each monitoring node and the substation topology map; the waveform similarity network model characterizes the correlation between the transient fault waveforms of each monitoring node; and the node credibility prior model quantifies the credibility level of each monitoring node. Based on the initial fault segment, fault path edge coordinates, and node weight vectors, the distribution transformer topology map is determined. The target fault section of the distribution transformer area is identified. Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cut off, and power dispatch is carried out using the isolated distribution transformer area. The above method achieves rapid, accurate and robust fault location of the distribution transformer area by integrating a graph-constrained linear model, a waveform similarity network and node credibility priors through a multi-model collaborative strategy, which significantly reduces the risk of misjudgment and missed detection in traditional single-point detection. Furthermore, based on the node credibility quantification mechanism, it can adapt to changes in the operating status of the transformer area and signal interference, ensuring the reliability of location under complex operating conditions. At the same time, relying on topological constraints and real-time fault path analysis, the optimal isolation scheme can be quickly generated, accurately controlling the impact of power outages within the smallest electrical range, effectively reducing the number of power outage users and power outage capacity, and providing key technical support for the rapid self-healing of the distribution network and the improvement of power supply reliability. Attached Figure Description

[0054] 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.

[0055] Figure 1 This is an application environment diagram of the power dispatching method for a distribution radio station in one embodiment;

[0056] Figure 2 This is one of the flowcharts illustrating a power dispatching method for a distribution radio station in one embodiment;

[0057] Figure 3 This is a second flowchart illustrating the power dispatching method for a distribution radio station in one embodiment;

[0058] Figure 4This is the third flowchart illustrating the power dispatching method for a distribution radio area in one embodiment;

[0059] Figure 5 This is the fourth flowchart illustrating the power dispatching method for a distribution radio station in one embodiment;

[0060] Figure 6 This is the fifth flowchart illustrating the power dispatching method for a distribution radio station in one embodiment;

[0061] Figure 7 This is a flowchart of the power dispatching method for a distribution radio area in one embodiment, number six.

[0062] Figure 8 This is a communication state switching diagram of a power dispatching method for a distribution radio area in one embodiment;

[0063] Figure 9 This is the seventh flowchart illustrating the power dispatching method for a distribution radio station in one embodiment;

[0064] Figure 10 This is a structural block diagram of a power dispatching device for a distribution radio station in one embodiment. Detailed Implementation

[0065] 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.

[0066] 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.

[0067] As distribution networks accelerate their digitalization and intelligentization, distribution substations, as the fundamental units for intelligent sensing and active control in the power system, are playing an increasingly important role in supporting precise grid operation and ensuring power supply reliability. Currently, distribution substation automation systems typically consist of distribution terminals, communication links, and a dispatch master station, enabling basic functions such as fault detection, isolation, and power restoration. Fault location is a crucial prerequisite for quickly identifying fault points to reduce fault propagation and plays a key role in improving fault handling efficiency. Post-fault power dispatch is the core link in ensuring continuous power supply to non-faulty areas and optimizing grid resource allocation. Compared to fault location, post-fault power dispatch requires more dynamic and collaborative decision-making by combining real-time load data, distributed power generation output, and the operating status of energy storage devices.

[0068] However, existing distribution substation automation systems mostly rely on fixed rules or preset scenario tables for power dispatch after a fault, which makes it difficult to meet the high requirements of digital smart distribution networks for power supply continuity and operational economy after fault handling.

[0069] Therefore, how to efficiently and accurately dispatch power after a distribution substation fault based on the real-time operating status of the distribution network has become an urgent problem to be solved.

[0070] In view of the above-mentioned technical problems, this application provides a power dispatching method for a distribution substation. The following embodiments will specifically illustrate the power dispatching method for the distribution substation.

[0071] The power dispatching method for distribution substations provided in this application can be applied to, for example... Figure 1The computer device shown includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces 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 interfaces. 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 interfaces are 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 the computer program is executed by the processor, it implements power dispatching for a distribution substation. The display unit is used to create a visually visible image and can be a display screen, a projection device, or a 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.

[0072] Those skilled in the art will understand that Figure 1 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.

[0073] In one exemplary embodiment, such as Figure 2 As shown, a power dispatching method for a distribution substation is provided, which can be applied to... Figure 1 Taking computer devices as an example, the explanation includes:

[0074] S201: Collect electrical signal data from the distribution station area and perform data preprocessing to obtain a set of data features.

[0075] The electrical signal data includes line voltage, current, phase angle, zero sequence, transient waveform, equipment status quantity, and time synchronization signal; the data feature set includes the electrical signal data and the preprocessed waveform, signal-to-noise ratio, packet loss rate, timing jitter, and long-term drift obtained after preprocessing of the electrical signal data collected by each terminal; the data feature set is used to characterize the electrical response characteristics of each monitoring node in the distribution substation to fault events.

[0076] In this embodiment, the computer equipment first comprehensively collects multi-dimensional electrical signal data, including line voltage, current, phase angle, zero-sequence component, transient waveform, equipment status variables, and time synchronization signals, through multiple measurement and control terminals deployed at key locations such as line nodes and power distribution equipment within the distribution area. Then, the collected electrical signals are transmitted to the computer equipment in real time using a dedicated power communication network or a wireless Internet of Things (IoT) network. The computer equipment then preprocesses the electrical signal data. The preprocessing process mainly consists of two aspects: the first aspect focuses on processing the time information in the electrical signals, including multi-terminal time synchronization calibration based on time synchronization signals, standardized format conversion of timestamps, and time sequence alignment of multi-source collected data, thereby ensuring consistency and accuracy of data from different terminals or different collection points in the time dimension; the second aspect focuses on processing the waveform data and collection noise of the electrical signals, filtering out high-frequency interference and random noise mixed in during the collection process.

[0077] Regarding the first aspect, it mainly consists of five steps. The first step involves the computer equipment establishing two parallel time synchronization channels using a dual-channel time synchronization architecture to ensure high-precision synchronization of the acquired electrical signal data. The dual-channel time synchronization architecture employs a hybrid architecture of master-slave redundancy (channel A) and relative time synchronization (channel B). Channel A uses the IEEE 1588 Precision Time Protocol (PTP) as its core, with the edge controller acting as the boundary clock and the terminal acting as the ordinary clock. The synchronization period is 1-2 seconds, supporting automatic switching between master and backup links. Channel B performs relative time synchronization based on a reference terminal, selecting a terminal with high topology centrality and strong clock stability (such as a TCXO / OCXO clock) as the reference node R. Other terminals achieve relative time difference correction through cross-correlation of event waveforms. When PTP or the Global Navigation Satellite System (GNSS) fails, the computer equipment automatically switches to Channel B, maintaining millisecond-level time synchronization accuracy through relative timing and residual correction. When PTP recovers, the computer equipment employs a gradual fusion strategy to avoid time phase jumps and ensure synchronization continuity and stability. In the second step, the computer equipment uses a data-driven time difference estimation method to provide data input to Channel B from the first step. This method works by analyzing waveforms acquired from different terminals under the same fault event. and Cross-correlation analysis is performed, and the time difference of event arrival is obtained by solving for the peak position of the cross-correlation function. Regarding the time difference of event arrival The solution can be obtained using relation (1), which is shown below:

[0078] (1);

[0079] in, The parameter that makes the function reach its maximum value. ; Indicates terminal The acquired waveform signal (time-shifted).

[0080] The specific process includes: waveform preprocessing (first, removing the DC component from the waveform signal, then normalizing it to unify the amplitude scale of waveforms from different terminals, thereby eliminating cross-correlation peak deviation caused by gain differences; finally, applying a Hanning window function within the truncated event time window to suppress spectral leakage caused by data truncation and ensure more concentrated frequency domain features), fast cross-correlation (i.e., using Fourier convolution to obtain cross-correlation peak points), subsample interpolation (i.e., parabolic or cubic spline interpolation of cross-correlation peak points), and multi-band enhancement (i.e., performing cross-correlation calculation and weighted averaging in frequency bands within the transient main energy band (e.g., 1-8 kHz), with the weights adaptively adjusted according to the signal-to-noise ratio (e.g., increasing the weight of high frequencies in arc events)). After the computer equipment completes the above five steps, the waveforms acquired from different terminals are then processed. and Input the relation (1) into the equation and perform the calculation. This data is directly used as input for the relative time synchronization and residual correction calculations in Channel B, ensuring that the computer equipment can maintain millisecond-level synchronization accuracy even when PTP / GNSS fails. The third step primarily relies on the data-driven time difference estimation output from the second step. Residual time difference is calculated across the entire network (i.e., relative time synchronization calculation) to achieve accurate global relative time synchronization correction. The principle behind this is to construct a network-wide time difference matrix. With the reference terminal R having high topological centrality as the time origin ( =0), construct the objective function to jointly solve for the relative offsets of each terminal. Its objective function is shown in relation (2):

[0081] (2);

[0082] in, This represents the set of participating paired terminals (filtered by link quality). This indicates the prior offset from PTP / GNSS; This represents the prior confidence level (0.1–0.3 when PTP is normal, and 0 when it fails).

[0083] After the computer device constructs the objective function, robust computation is then performed, that is, the computer device uses the Random Sample Consensus (RANSAC) algorithm to perform robust computation. Abnormal time differences are filtered out, and the sparse positive definite equation (i.e., relation (2)) is solved using the conjugate gradient method. Finally, the computer device processes the solution. First-order hysteresis smoothing is used to avoid sudden jumps in time phase. The solution process is shown in relation (3):

[0084] (3);

[0085] in, Indicates the smoothing factor; This represents the cumulative offset since the last application; This indicates the new offset to be applied after this update; This indicates the new offset applied in this update.

[0086] The fourth step involves the computer equipment quantifying and evaluating the time-driven time difference estimate after completing time synchronization between channels A and B through a synchronization monitoring and threshold mechanism. The core of this step is establishing a three-level threshold system and alarm triggering logic. First, the computer equipment defines the target time accuracy: an event-level relative error ≤ 0.2 milliseconds (when PTP is normal), and a normal error ≤ 0.5 milliseconds. When PTP or GNSS fails, an autonomous synchronization accuracy of ≤ 1 millisecond is maintained through channel B (reference terminal and relative time synchronization). Subsequently, the computer equipment establishes threshold logic, which involves real-time detection of the relative time offset of each terminal. (From the result of step 3), if >1 millisecond or sampling timing jitter If triggered three times consecutively within 0.5 milliseconds, the computer equipment will automatically reduce the weight of that node in the crowdsourced ranging (down to a minimum of 0.1) or temporarily remove it. Meanwhile, when PTP / GNSS completely fails, the computer equipment can maintain millisecond-level relative synchronization for several hours using the event-level residuals and linear drift model of data-driven time difference estimation, ensuring that crowdsourced ranging can still function normally in scenarios with a broken link.

[0087] The fifth step primarily involves the computer equipment quantifying the linear drift characteristics of node clocks to provide a correction benchmark for data-driven time difference estimation. The core of this step is constructing a linear prediction model for clock drift. For this linear prediction model, the computer equipment uses the relative time offset output from the residual time difference full network solution. Least squares fitting is performed on the offset sequence of the most recent M events to obtain the nodes. drift rate The construction of a linear prediction model by computer equipment can be represented by relation (4), which is shown below:

[0088] (4);

[0089] in, Indicates the initial offset; For extended runtime, if the terminal has a temperature sensor, a temperature coupling term can be introduced. ( , (To improve prediction accuracy by using temperature coefficients).

[0090] Regarding the second aspect, it mainly consists of five steps. The first step involves the computer equipment using an event window selection mechanism to dynamically capture the temporal distribution characteristics of the fault waveform. Its core is an adaptive window length adjustment mechanism based on line topology and transient characteristics. In the event window selection mechanism, the computer equipment first configures the window length, adopting a two-segment structure of a front window and a back window, with the front window as the default. =20 milliseconds (covering the steady-state reference waveform before the fault), back window =80 milliseconds (covering the transient main energy of the fault and the initial reflected wave). Subsequently, the computer equipment selects an adaptive adjustment strategy based on the length of the main trunk line in the transformer area. With the speed of propagation ( , (where the speed is light), the back window is dynamically expanded to cover multiple reflections. The computer device can dynamically expand the back window according to the relation (5), which is shown below:

[0091] (5);

[0092] in, To cover multiple reflections. For example, with long feeders (L>500 m), the back window can be extended to 120~160 ms; for arc-type faults (long transient duration), it can be further extended to 200 ms.

[0093] Finally, the computer equipment is equipped with a multi-window overlay mechanism. In complex branch or multi-reflection scenarios, 2 to 3 consecutive time windows (20 to 160 ms / window) can be extracted in a rolling manner. The waveform consistency is checked by cross-correlation between windows to ensure complete coverage of transient features.

[0094] The second step mainly involves denoising the fault waveform. First, the computer equipment uses a dual second-order notch filter to suppress the 50 / 60 Hz power frequency and 150 / 250 / 350 Hz odd harmonics in the fault waveform, setting the Q value to 20–40 to ensure a balance between notch depth and bandwidth. Then, the computer equipment uses a 1–8 kHz bandpass filter (the parameters can be adjusted according to the conductor and event type, such as extending to 10–12 kHz for arc fault scenarios) to focus the frequency band of the transient main energy. Finally, the computer equipment uses wavelet soft thresholding for denoising, selecting db6 / db8 wavelets or sym8 wavelets to decompose 4–6 layers, and then applying a general threshold to the high-frequency detail coefficients (d1–d3). , The noise estimate is processed with a soft threshold so that the reconstructed fault waveform retains the sharp features of the transient edges.

[0095] The third step involves the computer equipment processing and calculating the data integrity of each terminal in the distribution area. First, the computer equipment processes and calculates the data integrity of each terminal. The signal-to-noise ratio (SNR), packet loss rate, sampling timing jitter, and long-term drift are calculated. The SNR can be expressed by the following formula (6):

[0096] (6);

[0097] Where S represents the event valid region and N represents the pre-event silent region. A threshold of ≥30dB is recommended.

[0098] The packet loss rate is represented by the difference between the expected number of samples and the actual number of samples received within an event window, divided by the expected value. The threshold is ≤2%.

[0099] Sampling timing jitter This represents the main lobe half-width at half-maximum / sampling rate estimation near the cross-correlation peak, with a threshold ≤ 0.5ms;

[0100] Long-term drift refers to the drift over the most recent M events. The linear fitting slope (ms / h) has a threshold of ≤1ms / h;

[0101] Finally, the computer equipment normalizes the above indicators, which can be represented using relation (7):

[0102] (7);

[0103] After normalizing the above indicators, the computer equipment can input the results into the subsequent node weight calculation model for calculation.

[0104] The fourth step involves the computer equipment performing anomaly detection on the acquired electrical signals. For abnormal signals, the computer equipment uses abrupt change detection (4–6 times the standard deviation threshold) to pinpoint fault edges. For missing electrical signals (<5 ms), the computer equipment uses cubic splines or neighbor node time-shift consistency interpolation for repair. For nodes that fail to meet the quality threshold twice consecutively, the computer equipment automatically reduces the weight of that node to 0.1 or temporarily removes it.

[0105] The fifth step involves the computer equipment performing a triple consistency check on the electrical signals from the first to the fourth steps: time window overlap ≥ 95%, waveform correlation coefficient between adjacent nodes ≥ 0.8, and wavelet energy centroid deviation < 10%. This ensures strict alignment of the electrical signal data of multiple nodes in the time domain.

[0106] In summary, the key parameters for data preprocessing are shown in Table 1.

[0107] Table 1

[0108]

[0109] S202. Based on the node relationships of the distribution radio areas, a topology map of the distribution area is constructed.

[0110] In this embodiment, the computer device abstracts the distribution radio area into a radio area topology map based on the node relationships of the distribution radio area. Each node Indicates the terminal of the measurement and control terminal, switch or transformer, edge Represents the path / branch, and the length of the associated set for each edge. and electrical parameters (such as impedance per unit length) speed of transmission This allows for the formation of a complete electrical topology. Simultaneously, the fault data collected by each terminal includes pre-processed waveforms. And quality indicators (signal-to-noise ratio) Packet loss rate Timing jitter and long-term drift Furthermore, the preprocessed waveform and quality indicators have passed [the relevant standards / tests]. Figure 2 The example shown completes residual correction.

[0111] S203, based on the data feature set, the transformer area topology map and the multi-node collaboration strategy, obtain the initial fault section, fault path edge coordinates and the weight vector of each node of the distribution transformer area.

[0112] The multi-node collaborative strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model. The graph-constrained linear model is used to characterize the linear mapping relationship between the time-series observations of each monitoring node and the topology of the transformer area. The waveform similarity network model is used to characterize the correlation between the transient waveforms of faults among each monitoring node. The node credibility prior model is used to quantify the credibility level of each monitoring node.

[0113] In this embodiment, the computer device primarily calculates and analyzes the data feature set using a multi-node collaborative strategy and a distribution network topology map to obtain the initial fault section, fault path edge coordinates, and weight vectors of each node in the distribution network. First, based on the actual distribution network conditions, a distribution network topology map is established, and the initial fault section is obtained based on the distribution network topology map and the data feature set. Subsequently, the computer device inputs the initial fault section into the model within the multi-node collaborative strategy to obtain the fault path edge coordinates and weight vectors of each node.

[0114] S204. Based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node, determine the target fault section of the distribution radio area.

[0115] In this embodiment, the computer device uses the initial fault segment as a constraint space to filter edges that are located within or directly connected to the initial fault segment in the fault path edge coordinates, thus obtaining a candidate path set. Then, the computer device performs a weighted evaluation of path reliability based on node weight vectors, eliminating abnormal edges with weights below a threshold, forming a candidate path subgraph. Finally, the computer device fuses and analyzes the candidate path subgraphs, selecting the edge with the highest fault probability as the target fault segment.

[0116] S205, based on the hardware constraints of the distribution radio area and the target fault section, isolates and cuts off the fault area of ​​the distribution radio area, and uses the isolated and cut-off distribution radio area for power dispatch.

[0117] In this embodiment, the computer equipment performs isolation operations on the switches on both sides of the target fault section of the distribution substation based on hardware constraints such as switch breaking capacity, communication reachability, and protection interlocking relationships. After the isolation operation, the computer equipment synchronously verifies the switch position and line electrical quantities (current drop, zero sequence disappearance), and only after both are successfully confirmed is the isolation deemed effective. Subsequently, the computer equipment constructs a feasible operating domain that satisfies radial constraints, equipment capacity, voltage and current limits, and tie-line interlocking relationships, taking the isolated distribution substation as the object. Then, with the comprehensive objectives of minimizing untransferred loads, minimizing the number of operations, minimizing network losses, and optimizing voltage quality, it uses a heuristic approach to quickly generate feasible solutions and combines lightweight mixed integer programming (MILP) to perform power dispatching on the isolated distribution substation.

[0118] In the aforementioned power dispatching method for distribution substations, electrical signal data of the distribution substations is collected and preprocessed to obtain a data feature set; a substation topology map is constructed based on the node relationships of the distribution substations; based on the data feature set, the substation topology map, and a multi-node collaborative strategy, the initial fault segment, fault path edge coordinates, and weight vectors of each node of the distribution substation are obtained; the multi-node collaborative strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model; the graph-constrained linear model, the waveform similarity network model, and the node credibility prior model are run in parallel; based on the initial fault segment, fault path edge coordinates, and weight vectors of each node, the target fault segment of the distribution substation is determined; based on the hardware constraints of the distribution substation and the target fault segment, the distribution substation is dispatched... The faulty area is isolated and cleared, and power dispatch is carried out using the isolated distribution transformer area. The above method achieves rapid, accurate, and robust fault location in distribution transformer areas through a multi-model collaborative strategy that integrates graph-constrained linear models, waveform similarity networks, and node credibility priors, significantly reducing the risk of misjudgment and missed detection in traditional single-point detection. Furthermore, based on the node credibility quantification mechanism, it can adapt to changes in the operating status of the transformer area and signal interference, ensuring the reliability of location under complex operating conditions. At the same time, relying on topological constraints and real-time fault path analysis, the optimal isolation scheme can be quickly generated, accurately controlling the impact of power outages within the smallest electrical range, effectively reducing the number of users affected and the outage capacity, and providing key technical support for the rapid self-healing of the distribution network and the improvement of power supply reliability.

[0119] In one exemplary embodiment, such as Figure 3 As shown, the "obtaining the initial fault section, fault path edge coordinates, and weight vectors of each node of the distribution transformer area based on the data feature set, transformer area topology map, and multi-node collaborative strategy" in S203 above includes:

[0120] S301, input the data feature set into the single-node initial value model, perform distance estimation on each node of the transformer area topology map, and obtain the initial position of each node from the initial fault point.

[0121] The single-node initial value model includes impedance mutation units and transient energy center units; the impedance mutation units and transient energy center units operate in parallel.

[0122] In this embodiment, after obtaining the data feature set, the computer device inputs the data feature set into the single-node initial value model for calculation. The single-node initial value model mainly uses two complementary algorithms to quickly generate distance estimates from each node to the fault point, i.e., the initial location of the initial fault point. The single-node initial value model consists of two units: an impedance mutation unit and a transient energy center unit, and the two units run in parallel. For the impedance mutation unit, the computer device mainly uses the mutation characteristics of voltage and current phasors before and after the fault event to estimate the fault distance, and analyzes the phasor changes before and after the event collected by the node. and Calculate the equivalent impedance change Combined with the impedance per unit length of the line Alternatively, if there is a segment label library, the distance can be calculated along the topological path using the cumulative matching method, and can be represented by the relation (8):

[0123] (8);

[0124] in, This represents the cumulative length of the edge e;

[0125] Meanwhile, to address the impact of branch loads, computer equipment can introduce empirical correction coefficients or small-scale power flow sensitivity linear compensation to reduce the interference of topology complexity on estimation.

[0126] For transient energy centroid units, the computer equipment locates faults by the energy distribution characteristics of the transient waveform. Wavelet decomposition is performed on the preprocessed transient waveform to extract the time-domain energy distribution of the main energy frequency band from 1–8 kHz (10–12 kHz for arc faults). The energy centroid of each frequency band is calculated using equation (9), which is shown below:

[0127] (9);

[0128] Multi-band weighting center , and reference time Compare and estimate the propagation time .

[0129] Finally, computer devices at average propagation speed Get distance Optionally, the computer device can adaptively assign weights based on the signal-to-noise ratio (SNR) within the frequency band, prioritizing the energy centroid of high SNR frequency bands.

[0130] S302, based on the data feature set, multi-node collaboration strategy and initial position, calculate the fault point to obtain the initial fault section, fault path edge coordinates and weight vector of each node in the transformer area topology map.

[0131] In this embodiment, the computer device uses initial location information (obtained from impedance mutation unit and transient energy center unit) as a reference point, and combines it with a set of data features (including transient energy amplitude, waveform distortion, phase jump angle, etc.) to estimate the initial fault segment under a multi-node collaborative strategy framework. Then, the computer device defines the segment where the initial location is located as the center and expands along the transformer substation topology map in both directions with a finite step size to construct the initial fault segment. Based on the timestamp sequence of each node's features and the connection relationships in the transformer substation topology map, the time sequence of fault disturbance propagation is reconstructed and extracted. Simultaneously, each edge in the path is uniquely encoded to obtain the fault path edge coordinates. Finally, the computer device combines feature reliability weights, collaborative contribution weights, and topological criticality weights to obtain the weight vector for each node.

[0132] In one exemplary embodiment, such as Figure 4 As shown, the "calculation of fault points based on data feature sets, multi-node collaboration strategies, and initial locations to obtain the initial fault segment, fault path edge coordinates, and weight vectors of each node in the transformer substation topology map" in S302 above includes:

[0133] S401 uses a graph-constrained linear model to solve the data feature set and obtain the edge coordinates of the fault path.

[0134] In the graph-constrained linear model, the coordinates of the fault path edges are obtained by constructing a set of time-difference equations based on topology, which represents the projected length of the fault to each edge. .

[0135] In this embodiment, the computer device inputs a set of data features into a graph-constrained linear model and sets fault points. To the node The shortest path propagation time difference is equal to the arrival time difference of the observed waveforms at the two nodes. It can be expressed using relation (10):

[0136] (10);

[0137] in, This represents the weighted shortest path distance on the graph (unit: meters or equivalent length). This indicates the propagation speed of edge dependencies.

[0138] The computer equipment introduces a vector d to represent the projection length of the fault point on each edge (non-negative, only the fault path is non-zero); simultaneously, a coefficient matrix A is constructed, with each row corresponding to a pair of nodes ( Its coefficients are derived from two lines from the unknown point to... The shortest path is composed of the sign path difference on each edge (+1 for the same direction, -1 for the opposite direction, 0 for not on the path), and a constraint function is constructed, which is represented by relation (11):

[0139] (11);

[0140] The non-zero support of d is limited by the two shortest paths from the source edge to the two nodes.

[0141] The computer device sets the objective function to solve for d, and the objective function is expressed as in relation (12):

[0142] (12);

[0143] Simultaneously, the computer device adds an L1-norm penalty term, forcing solutions to concentrate on a single continuous path; and enforces path connectivity through dynamic programming or projection methods, avoiding non-physically dispersed localization. In practical applications, the computer device can first narrow down the candidate subgraph (by single-node priors and waveform similarity networks, occupying the entire subgraph) to the boundary. Figure 10 The algorithm operates within the -30% edge range, and simultaneously uses either the iterative weighted least squares algorithm or the alternating direction multiplier algorithm to iteratively solve the objective function.

[0144] S402, based on the waveform similarity network model, the fault waveform is analyzed to obtain the initial fault section.

[0145] In this embodiment, the computer device utilizes a waveform similarity network model to obtain the initial fault segment through similarity analysis of waveform features between nodes. First, the computer device constructs a node similarity matrix S and integrates normalized cross-correlation peak values. Inverse distance of Dynamic Time Warping (DTW) and energy center of gravity difference The weight distribution relationship of these three types of indicators is shown in equation (13), which is as follows:

[0146] (13);

[0147] in, The initial weights are 0.5, 0.4 and 0.1, respectively.

[0148] Subsequently, the computer equipment performs semi-supervised label propagation on the transformer area topology map, that is, using a small number of near-fault and far-fault nodes in the initial value of a single node as seeds, and generating a scalar field close to the fault level by optimizing the relation (14). The relation (14) is shown below:

[0149] (14);

[0150] Finally, the computer equipment segments the topology of the distribution area by determining the location of the maximum gradient or the maximum inconsistency in the scalar field, thereby identifying the initial fault section. This is then used as a candidate path subgraph for subsequent ranging modules.

[0151] S403, based on the prior model of node credibility, weight the credibility of each node in the transformer area topology map to obtain the weight vector of each node.

[0152] In this embodiment of the application, the computer device will Figure 2 The signal-to-noise ratio of each node obtained from the embodiments involved Packet loss rate Timing jitter Together with the centrality parameter, it forms a set of credibility. Subsequently, the computer device performs a confidence weighting based on relation (15), which is shown below:

[0153] (15);

[0154] The centrality parameter can be normalized using betweenness / closeness centrality, prioritizing the trust of topological key points.

[0155] In one exemplary embodiment, such as Figure 5 As shown, the phrase "determine the target fault section of the distribution transformer area based on the initial fault section, fault path edge coordinates, and weight vectors of each node" in S204 above includes:

[0156] S501, scale alignment is performed on the initial position and fault path edge coordinates to obtain the initial fault segment set.

[0157] In this embodiment of the application, the computer device uses the initial position obtained from the single-node initial value model for each node. , And the fault path edge coordinates obtained using a graph-constrained linear model Scale alignment is performed. This involves mapping the prior node distances to edge coordinates, distributing the distances from nodes to unknown points along the shortest path across the edges. This allows the computer to determine the initial position. , and fault path edge coordinates It is transformed into a set of initial fault segments on the same scale.

[0158] S502, the target fault section is obtained by calculating based on the initial set of fault sections and the weight vector of each node.

[0159] In this embodiment, the computer device calculates the initial set of fault segments and the weight vector of each node through M-estimation fusion, assigns dynamic weights to the ranging results of different nodes, and obtains the target fault segment by minimizing the weighted residual. The computer device constructs the objective function of M-estimation fusion with the fault point location as the variable, and the objective function is shown in relation (16):

[0160] (16);

[0161] in, Represents the path from candidate point to node The shortest path distance on the graph; express and Weighted synthesis (can be further weighted by frequency band / quality); This represents the loss function for Huber or pseudo-Huber.

[0162] Finally, the computer equipment solves the objective function using iterative weighted least squares (IRLS) or alternating direction multiplier method (ADMM) to obtain the optimal fault location and its confidence level.

[0163] Optionally, the computer device can be equipped with a confidence and consistency scoring unit, which can generate interpretable decision-making criteria by fusing multi-dimensional indicators. First, this unit calculates model consistency, i.e., it calculates... First, it can reflect the fitting accuracy of the graph-constrained linear model; second, the unit performs statistical analysis on node consistency, namely the proportion of effective inliers after M-estimation (the proportion of nodes that participate in fusion and whose residuals are within a reasonable range), which is used to measure the synergy of multi-node data; finally, combined with the overall quality metric, the unit normalizes the mean weighted combination of model consistency, node consistency and participating node weights using the Sigmoid function, and its normalization can be expressed by relation (17):

[0164] (17);

[0165] in, This indicates Sigmoid normalization. This represents the average weight of the participating nodes.

[0166] Finally, the confidence and consistency scoring unit outputs a comprehensive confidence score (conf). The decision threshold is set as follows: when conf ≥ 0.8, the scheduling instruction is executed directly; when conf ∈ [0.6, 0.8), secondary verification is triggered (such as adding data collection from adjacent nodes or waiting for subsequent event verification); when conf < 0.6, only alarm information is output, and self-healing is not initiated. This scoring mechanism achieves hierarchical decision-making for ranging results, ensuring rapid response in high-confidence scenarios while reducing the risk of erroneous actions in low-confidence scenarios through secondary verification, thus improving the overall reliability of the method.

[0167] When the computer equipment is solving for the optimal fault location, if a few nodes are distorted or offline, it can automatically reduce the weight of abnormal nodes using node confidence weights and remove outliers using the M-estimation algorithm. If fewer than three nodes are effectively involved, a low-confidence result is output and only an alarm is triggered to avoid invalid decisions. If topology inconsistency or model error causes abnormally large residuals in the graph-constrained linear model, the computer equipment will trigger a topology self-check to reconfirm the status of near-end switches. If necessary, it will switch to a conservative strategy of near-end segmentation and priority isolation to prioritize power supply safety. If there is an uncertainty in the propagation speed, the computer equipment will search the propagation speed within a ±10% range, select the speed value corresponding to the minimum residual, and perform self-learning optimization using measured fault backreference data. If multiple points or crosstalk events occur, the waveform similarity network model will exhibit multi-peak characteristics, causing the candidate segment set to widen. In this case, the computer equipment will use the minimum cut algorithm to generate Top-2 candidate results and selectively isolate the nearest two ends to gradually narrow the fault range. For scenarios with strong reflections or multiple echoes, although the energy centroid method may introduce biases, the constraints of the graph-constrained linear model and the M-estimation can still suppress interference. If necessary, the ranging accuracy can be improved by extending the post-event sampling window (t_post) and solving it a second time.

[0168] In one exemplary embodiment, such as Figure 6 As shown, the above-mentioned S205, "based on the hardware constraints of the distribution transformer area and the target fault section, isolate and disconnect the faulty area of ​​the distribution transformer area, and use the isolated distribution transformer area for power dispatching," includes:

[0169] S601 defines the boundaries of the scheduling scheme based on hardware constraints and constructs a scheduling model.

[0170] In this embodiment, after determining the target fault section, the computer device defines the boundaries of the scheduling scheme based on hardware constraints and constructs a scheduling model. First, the decision variables for the scheduling model are set around three categories of variables: topology / switch state, node voltage / line current, and power supply / transfer allocation. For topology / switch state, the computer device uses binary variables to represent edges (switches). The closing (1) or opening (0) of the , i.e. The switches in the faulty section are forcibly set to 0. For node voltage / line current, the computer equipment sets node voltage / line current as a continuous variable, including the node per-unit voltage. and the upper limit of line current amplitude or phase current For power supply / distribution, the computer equipment settings for power supply / distribution are also continuous variables, including load points. Actual power supply, i.e. and power supply / upstream feeder upper limit of output In practical applications, for low-voltage / distribution area scenarios, computer equipment can choose linear power flow approximation or current-voltage linear envelope model to compress the computational scale and ensure that edge devices can efficiently solve constrained optimization problems.

[0171] Subsequently, the computer equipment needs to define the boundaries of the scheduling scheme by considering six factors. The first factor is topological constraint, meaning that when formulating the scheduling scheme, the computer equipment needs to maintain the radial structure of the distribution area, i.e. Alternatively, a controlled ring network mode can be used, which allows single-ring closures but requires interlocking protection: , This indicates a permitted controlled loop. The second factor condition is equipment limits, including mandatory constraints that the line current does not exceed the conductor current carrying capacity limit, i.e. The transformer load rate is within the rated capacity; before switching operation, it is necessary to estimate whether the inrush current exceeds the limit (real-time verification during the execution phase). The third factor is to evaluate the quality of voltage or current, including setting the node voltage deviation threshold (e.g., per-unit value ±5%), the safe upper limit of line current, and the total harmonic distortion (THD), and using these three points as verification thresholds for power restoration operation. The fourth factor is power supply feasibility. The computer equipment uses a linearized power flow model (e.g., LinDistFlow approximation or current-voltage envelope method) to constrain the power balance. In this process, the computer equipment uses the principle that the injected power of each node is equal to the sum of the downstream loads and the line current is limited by path impedance and load distribution as the core constraint. The fifth factor is protection interlocking, which requires that the tie switch can only be operated after the switches at both ends of the fault section are forcibly disconnected, and an interlock set is set (prohibiting the simultaneous operation of specific switch combinations). Before closing the loop, the computer equipment needs to detect the voltage phase difference (e.g., ≤10°) and voltage difference. The sixth factor is a communication constraint, which fixes the status of unreachable or unhealthy devices to their current value, excluding them from the feasible region, and requires each action to be confirmed within a timeout window. Based on the above six factors and the constraints of the decision variables, the computer equipment constructs the objective function of the scheduling model. The objective function of the scheduling model can be represented by relation (18), which is shown below:

[0172] (18);

[0173] in, This indicates the capacity not being transferred; Indicates the number of times an action is performed; This indicates an approximate network loss. Indicates voltage deviation; (You can start with 1.0:0.1:0.05:0.01 and then fine-tune it according to the station area strategy); if phase imbalance constraints are needed, you can further increase the number of constraints. .

[0174] S602, input the target faulty section into the scheduling model, isolate and remove the faulty area of ​​the distribution radio area, and obtain the scheduling scheme.

[0175] In this embodiment, the computer device first inputs the target fault segment into the scheduling model and solves for the target scheduling action sequence. Then, the computer device generates a corresponding scheduling scheme based on the target scheduling action sequence. The computer device solves the scheduling model in three main stages.

[0176] The first phase primarily uses a greedy strategy to quickly generate an initial sequence of scheduling actions that satisfy safety constraints. First, the computer performs a fault isolation operation, setting the nearest switches at both ends of the faulty section to the open state (skipping this step if the switches are already open), and forcibly locking the switches within the faulty section to the open state. Then, for each candidate tie line, the computer estimates the available capacity upstream. and path bottlenecks (such as line current limits). Finally, based on load priority (important users > ordinary users > end users) and the shortest path principle, the power loads are attempted to be connected to the candidate tie line one by one. If any constraint is violated during the connection process (such as current exceeding the limit or voltage deviation exceeding the standard), the current path is skipped or a different path is chosen, thus forming the initial scheduling action sequence. .

[0177] The second phase mainly involves the initial scheduling action sequence generated in the first phase. Building upon this foundation, the objective function of the scheduling model (including unused capacity, number of actions, network loss, and voltage deviation) can be further optimized through neighborhood search or lightweight mixed-integer linear programming (MILP). Specifically, there are two approaches: The first is neighborhood search, where the computer equipment defines a neighborhood of operations (e.g., single tie-line replacement, two tie-line swapping, power supply area merging / splitting), evaluates the incremental cost at each step, and only accepts feasible and cost-reducing neighborhood solutions, iterating until convergence or reaching the time limit (typically 50–150 ms). The second approach is lightweight mixed-integer linear programming (MILP), which optimizes the binary variables of the candidate edges obtained in the first stage. To preserve the current flow, linearized power flow constraints (such as LinDistFlow or current envelope and voltage L1 norm) and parent-child flow constraints are employed, utilizing the hot-start function of the CBC / Gurobi solver (with... Using initial values, run on an edge computing device equipped with a 4- to 8-core CPU, and within a time window of hundreds of milliseconds (e.g., 200 milliseconds), obtain an intermediate scheduling action sequence that is significantly improved compared to the initial scheduling action sequence. .

[0178] The third phase is the intermediate scheduling action sequence in the second phase. Based on this, the computer equipment uses security reinforcement learning technology to deploy an intelligent decision-making strategy model on edge computing nodes to solve for the target scheduling action sequence. Its implementation comprises four core components: state space, action space, reward mechanism, and safety filtering. The state space consists of the current topology embedding (adjacency / switch state), real-time measurement data (voltage / current over-limit margin), feasible domain flags (interlocking / communication reachability), location results (fault segment and confidence level), and a summary feature of the intermediate scheduling action sequence from the second stage. The action space is a set of feasible actions filtered under a safety shield mask (such as closing tie lines or opening sectionalizing switches). All candidate actions must pass a triple safety check at the rule layer (radiality, interlocking, equipment rating, voltage deviation, and phase difference threshold checks), the prediction layer (one-step linear power flow assessment and loop inrush current estimation), and the execution layer (post-action readback verification). Actions that fail the check are directly masked. The reward mechanism aims to minimize unused capacity, number of actions, and network losses, while introducing positive incentives based on safety margins. When training the intelligent decision-making strategy model on computer equipment, the main methods employed are offline simulation (using historical fault sets in the main station environment and training the near-end policy optimization algorithm / deep Q-network algorithm under random operating conditions, while embedding an action masking network to learn rules for actions that cannot be performed) and online inference separation mode. The edge device only performs policy inference and does not perform online training, ultimately outputting the target scheduling action sequence. .

[0179] Finally, the computer equipment transforms the target scheduling action sequence into an executable action sequence and ensures safe implementation. First, the computer equipment compares the current topology with the target scheduling action sequence to generate a set of switches to be changed. Then, it sorts them according to topology dependency, following the principle of opening before closing. It first isolates the switches at both ends of the faulty section, and then closes them into the tie line in batches, with each batch changing about 2-3 actions. Before each action is executed, the voltage difference, phase difference, and predicted loop inrush current need to be checked. Only after the threshold is met can the command be issued. After execution, it can be verified by local or remote status readback (i.e., double confirmation) and measurements such as voltage, current, zero sequence, and harmonics. If it passes, it proceeds to the next step. If any step fails, it immediately reverts to the most recent safe state and records the cause and measurement snapshot, switching to a conservative strategy of only isolating and not transferring power across transformer areas.

[0180] When a link interruption occurs, the computer equipment can maintain functionality through a communication degradation strategy (i.e., switching to the backup link when the main link fails, and entering autonomous mode to perform only isolation and conservative power transfer when the link is completely lost). For situations with insufficient ranging confidence (conf∈[0.6,0.8)), the computer equipment can initiate a secondary verification process or limit the scheduling strategy to small-step probing. For equipment refusing to operate or abnormal readback, the equipment is immediately marked as unreachable and the scheme is recalculated (if it fails after ≤1 retries, it conservatively exits). For circulating current risks, the computer equipment uses pre-loop closing calculations and phase difference or voltage difference thresholds to suppress risks, and activates soft loop closing steps when necessary. Simultaneously, the computer equipment verifies protection selectivity during the scheduling scheme generation phase, and avoids overstepping actions by temporarily adjusting the reclosing / pressure plate strategy, ensuring that all operations are performed within safety boundaries.

[0181] The default parameters and thresholds are shown in Table 2.

[0182] Table 2

[0183]

[0184] In one exemplary embodiment, such as Figure 7 As shown, the fault dispatch control method for the above-mentioned distribution radio area also includes:

[0185] S701 determines the target scheduling scheme based on the communication operation status of the distribution radio area after isolation and disconnection.

[0186] The communication operation state includes any one of the following: normal state, degraded state, and autonomous state.

[0187] In this embodiment, the computer device defines the communication operation status of the isolated distribution radio area into three operating modes: normal state (S0), degraded state (S1), and autonomous state (S2). Furthermore, the computer device determines the target scheduling scheme based on these three communication operation states. The normal state (S0) indicates good master station communication; the degraded state (S1) indicates intermittent master station communication; and the autonomous state (S2) indicates master station communication interruption. When the master link heartbeat times out three consecutive times (3 seconds each time), the degraded state is triggered, and the communication link automatically switches to the backup link for communication. Functionality is not degraded, but throughput and latency are limited.

[0188] S702, when the communication operation status of the distribution radio area is normal or degraded, power dispatch is carried out according to the first dispatching scheme using the distribution radio area after isolation and disconnection.

[0189] In this embodiment, the normal state relies on full-duplex communication on the main link and performs power dispatching according to the first dispatching scheme, that is, transforming the executable sequence of actions into the first dispatching scheme. The first dispatching scheme consists of six steps. The first step involves the computer device setting constraints based on input elements and preconditions. The input elements include the aforementioned... Figure 4 The edge where the fault is located as described in the implementation Distance along the edge Execution and quality indicators Real-time snapshots of system operation (voltage, current, power, zero sequence, frequency, harmonic THD, device health status, communication reachability) are required. Prerequisites include: Direct execution mode is entered when the location result confidence level (conf) is ≥ 0.8; secondary verification and constraint strategies (such as adding adjacent waveform acquisition or increasing the verification threshold) are enabled when conf is ≤ 0.6 and conf is < 0.8; and alarms are only triggered for conf < 0.6 without triggering self-healing operations. Simultaneously, the computer equipment needs to load preset safety constraints and scheduling strategies (including topology radiality / controlled loop requirements, switch interlock sets, device rated limits, phase difference / voltage difference thresholds, load priority, and protection coordination rules) to ensure subsequent actions are executed within the safe and feasible domain.

[0190] The second step involves the computer equipment projecting the fault edge ID and distance along the edge from the location results onto the transformer substation topology map to determine the operable switches at both ends of the faulty segment. Next, a tiered execution strategy is triggered based on the ranging confidence level (conf): if conf ≥ 0.8, the process directly proceeds; if 0.6 ≤ conf < 0.8, adjacent waveform acquisition is added or a secondary verification restriction strategy is enabled; if conf < 0.6, only an alarm is triggered without self-healing. The computer equipment, combining communication reachability, switch interlocking rules, equipment rated capacity, and pre-calibrated phase difference or voltage difference thresholds, selects a set of safe and feasible candidate tie lines (CAND) from the topology, providing precise path information for subsequent isolation and power transfer.

[0191] The third step involves the computer equipment verifying the waveform consistency between adjacent nodes, requiring a cross-correlation coefficient ≥ 0.8 and an energy centroid deviation ≤ 10%, ensuring consistent feature extraction for the same event across different nodes. Subsequently, the time synchronization residual across the entire network is checked, requiring a relative time difference between nodes ≤ 0.5 ms. If this is exceeded, the relevant nodes are automatically downweighted or their fusion is delayed. Simultaneously, the computer equipment verifies topology consistency by comparing the near-end switch status with a pre-stored topology map to confirm that the physical location of the switches at both ends of the faulty segment matches the logical topology. Finally, the computer equipment, in conjunction with protection criteria, correlates and verifies real-time measurements such as zero-sequence current and short-circuit current with fault location results, ensuring logical consistency between fault characteristics and location conclusions. If any verification item fails, a secondary verification process is automatically initiated (adding adjacent waveform acquisition or increasing the verification threshold), only triggering an alarm without initiating self-healing. If all verifications pass, the process directly enters the isolation phase, with typical processing time controlled within 50-100 ms.

[0192] Fourth, the computer equipment issues a blocking command to the switches at both ends of the faulty section and exits the automatic reclosing logic to prevent false reclosing or reverse power supply. Then, it performs the disconnection operation in the order of remote side priority and power flow small end priority. After disconnection, the computer equipment synchronously triggers a dual confirmation mechanism, namely, execution confirmation (ACK-E) is completed by reading back the contact status or remote signaling quantity, and measurement confirmation (ACK-M) is completed by the line current dropping below the threshold and the zero-sequence component decreasing by ≥90%. If the disconnection timeout (0.8-1.0 seconds) or the dual confirmation fails, it automatically switches to the backup section switch for retry. If it still fails, the equipment is marked as having a communication or mechanical abnormality and a local minimum isolation mechanism is executed (only the nearby visible switch is disconnected), stopping the subsequent power transfer process to avoid risk propagation. The typical operation time is controlled within 300-700 milliseconds, depending on the mechanical action response speed of the switch mechanism.

[0193] The fifth step, primarily after safely isolating the faulty section, involves closing a backup tie switch to restore power to the non-faulty area, minimizing the outage scope. First, the computer equipment performs batch planning based on the action sequence output by the adaptive scheduling engine, executing 1-3 actions per batch to control risk. Each candidate tie switch undergoes three pre-closing checks, specifically: voltage difference ≤5%, phase difference ≤10° (tightened to ≤3-5% in autonomous mode), and inrush current not exceeding limits. Then, the operation is executed according to the process of unlocking the pressure plate, closing the switch, and double confirmation. The double confirmation includes execution confirmation (ACK-E) based on equipment status readback, measurement confirmation of line current or voltage stability (ACK-M), and real-time verification of node voltage deviation, line current exceeding limits, and zero-sequence or harmonic anomalies. If any indicator exceeds the limit, the system immediately reverts to the nearest safe state or switches to a backup tie switch. The total execution time is strictly controlled within 3-5 seconds. If the timeout is exceeded, subsequent batches will be automatically stopped to avoid the spread of risk. The typical single batch takes 300-800 milliseconds, depending on the mechanical response of the contact switch and the measurement and verification speed.

[0194] The sixth step, primarily following the completion of the fifth step, involves multi-dimensional measurement and verification to confirm the power supply quality and safety boundaries of non-faulty areas. First, computer equipment verifies the steady-state voltage of key nodes, requiring all node voltage deviations to be ≤5% (tightened to ≤3% in autonomous state). Then, line current and load verification is performed, specifically ensuring the line current does not exceed the rated upper limit and the transformer load rate is controlled within the nameplate limits. Simultaneously, transient and harmonic verification is conducted, confirming no abnormal rise through zero-sequence or differential current monitoring, with the total harmonic distortion (THD) at key points ≤5%-8% (specific thresholds adapted to equipment type). Finally, short-window stability observation is performed, requiring the above indicators to remain stable without fluctuations within a 1-2 second observation window. If any indicator exceeds the limit, a batch-by-batch rollback to the nearest safe state or switching to an alternative connection path is immediately triggered. If the process still fails, subsequent power transfer procedures are stopped and marked as "power transfer exceeding limits." Typical verification time is controlled within 500–1500 milliseconds, depending on the sampling frequency and data transmission speed of the measurement equipment.

[0195] S703: When the communication operation status of the distribution radio area is autonomous, power dispatch is carried out according to the second dispatch scheme using the distribution radio area after isolation and disconnection.

[0196] In this embodiment, when the round-trip delay of the backup communication link exceeds 800ms for five consecutive times or the packet loss rate exceeds 20%, the communication operation state is autonomous, and minimal coordination can only be maintained through short message direct network connection. The state switching employs a jitter suppression mechanism; the primary link needs to stabilize for 30 seconds before exiting the degraded state, and exiting the autonomous state requires a 10-30 second delay after link recovery and handshake verification to ensure a disturbance-free switching process. Figure 8As shown. The core mechanism of the short message direct connection network is that, during link degradation, computer equipment maintains minimal coordination through a simplified communication protocol. Specifically, this includes: computer equipment using RS-485, CAN, or Sub-GHz / LoRa short message direct connection networks (Local Skinny Bus) to construct low-bandwidth communication links within the transformer area; the transmission unit adopts a 256-byte TLV frame structure, containing key frame types such as event fingerprint (EVENT_SIG), location digest (LOC_SUM), command request or receipt (CMD_REQ / CMD_ACK), heartbeat (HB), and audit chain fragment (AUDIT); simultaneously, a sequence number, session ID, and a rolling window anti-replay mechanism are uniformly set in the frame header to ensure message uniqueness and security.

[0197] The second scheduling scheme consists of two steps. The first step mainly involves the computer equipment dividing the scheduling boundary. Specifically, firstly, the computer equipment defines the degraded operation rules under the condition of complete communication disconnection (autonomous state), that is, freezing the model parameters (as mentioned above). Figure 4 The node weights mentioned in the implementation are fixed as the most recent qualified assessment values ​​issued by the main station, while the propagation speed, ranging threshold, and scheduling strategy hyperparameters are kept to be the most recent valid versions; in the autonomous state, only time-sharing with the aforementioned events through event-level residuals is allowed. Figure 4The objective function (such as relation (16)) described in the implementation is locally adjusted (such as using the cross-correlation residual of event waveforms to correct clock drift), and updating the global model or weight allocation rules is prohibited. If the propagation speed is uncertain or other boundary scenarios are encountered, the computer equipment will optimize the parameters based on historical fault data. However, all learning results are only updated offline by the main station after communication is restored to ensure parameter stability and decision security during the autonomous period. Subsequently, the computer equipment only performs disconnection operations (isolation priority) on both ends of the fault section. Under strict conditions, local transfer of power supply is allowed in a small range. That is, when the load point and the power supply are in the same area, the electrical distance is short (<300 meters), the power supplies on both sides are in phase, and the phase difference before closing is ≤5° and the voltage difference is ≤3%, a single-step closing operation can be performed (multi-step or batch operations are prohibited). At the same time, cross-area transfer of power supply, long-distance power supply, multi-hop reconfiguration and the construction of any form of controlled ring network are clearly prohibited to ensure that the operation under autonomous state is strictly limited to the minimum safety range and to prevent the risk spread caused by communication interruption. Then, the computer equipment observes the operation, fixing the batch size to 1 (i.e., only one action is executed at a time), with an action interval of at least 1 second. Furthermore, after each action is completed, it must pass double confirmation (including execution confirmation ACK-E and measurement confirmation ACK-M) and real-time measurement verification (covering parameters such as voltage V, current I, zero sequence, and harmonic THD). ACK-E requires the equipment to read back the status (such as contact position, GOOSE signal, remote signaling status, and no error code after motor stroke completion) to be consistent with expectations. ACK-M requires on-site secondary sampling to verify changes in status quantities (such as breaking current dropping below the threshold, closing voltage being qualified, and phase difference / harmonic within limits). Only after both pass simultaneously and the measurement verification is within limits (e.g., voltage deviation ≤ 5%, harmonic THD ≤ threshold) will the computer equipment allow the next action to be executed. If any confirmation fails or verification is abnormal, a rollback mechanism is immediately triggered (such as trying an alternative path or terminating the transfer), ensuring that the execution of actions is strictly controlled and safe and reliable in autonomous mode.

[0198] The second step primarily involves the computer equipment executing distributed decision-making within the framework defined in the first step, generating legitimate scheduling instructions. Specifically, firstly, under a complete communication disconnection (autonomous state), the computer equipment defines the node set and operating rules. The set of nodes participating in consensus arbitration must be of an odd number (typically 3 / 5 / 7 nodes), including three types of entities: edge controllers, key monitoring and control terminals (such as FTUs, TTUs, or ring network cabinet control units), and backup gateways. Within this node set, three dynamic roles are defined: a temporary leader, acting as the sole issuer of instructions and responsible for generating commands; followers, used to execute voting verification and actions; and candidates, who initiate elections after the election timeout. All decisions follow the majority rule. The election mechanism involves a candidate node broadcasting a voting request to the consensus node set containing its term number, latest log index, and health status when a temporary leader is continuously missing from the direct heartbeat signal for more than a preset random timeout period of 1.2-1.8 seconds. The participating nodes are elected as the temporary leader based on the majority principle (receiving more than half of the votes). Nodes with low health (such as insufficient SNR or excessive clock drift) or those blacklisted are strictly excluded from the election or voting. To prevent frequent election jitter, the active temporary leader needs to broadcast a heartbeat signal periodically (every 300-500 milliseconds) to maintain its authority. If a follower receives a valid heartbeat before the timeout, it automatically forfeits its right to initiate a new election.

[0199] Finally, when the temporary leader needs to execute an operation command, it first broadcasts the command request to all consensus nodes, requiring each node to perform local security verification (including interlocking logic, device reachability, phase difference or voltage difference thresholds, loop inrush current estimation, etc.). If the verification passes, it returns to the ready state; otherwise, it returns to the reject state. If the temporary leader receives a ready state from more than half of the nodes, it issues a COMMIT instruction to all nodes for execution; otherwise, it forcibly terminates the process. At the same time, each node must form a chain audit log in sequence with the current term number, operation type, and hash value of the previous operation to ensure the integrity and immutability of the operation sequence. This process is supplemented by timeout control (500 milliseconds for the preparation phase and 800 milliseconds for the COMMIT readback). Timeout or failure immediately triggers the termination of the process and marks the relevant devices or links as downgraded, while automatically switching to alternative operations or conservative isolation strategies. The parameters are set as shown in Table 3.

[0200] Table 3

[0201]

[0202] In summary, based on all the above embodiments, a fault scheduling and control method for a distribution radio area is also provided, such as... Figure 9 As shown, the method includes:

[0203] S801 collects electrical signal data from the distribution radio station area and performs data preprocessing to obtain a set of data features;

[0204] S802, based on the node relationships of the distribution radio areas, constructs the radio area topology map;

[0205] S803 inputs the data feature set into the single-node initial value model, performs distance estimation on each node of the transformer area topology map, and obtains the initial position of each node from the initial fault point.

[0206] S804 uses a graph-constrained linear model to solve the data feature set and obtain the edge coordinates of the fault path;

[0207] S805, based on the waveform similarity network model, performs fault waveform analysis to obtain the initial fault section;

[0208] S806, based on the prior model of node credibility, the credibility of each node in the transformer area topology is weighted to obtain the weight vector of each node;

[0209] S807, scale alignment is performed on the initial position and fault path edge coordinates to obtain the initial fault segment set;

[0210] S808, the target fault section is obtained by calculating based on the initial set of fault sections and the weight vector of each node;

[0211] S809 defines the boundaries of the scheduling scheme based on hardware constraints and constructs a scheduling model.

[0212] S810: Determine the target scheduling scheme based on the communication operation status of the distribution radio area after isolation and disconnection. When the communication operation status of the distribution radio area is normal or degraded, execute S811. When the communication operation status of the distribution radio area is autonomous, execute S812.

[0213] S811, when the communication operation status of the distribution radio area is normal or degraded, power dispatch is carried out according to the first dispatching scheme using the distribution radio area after isolation and disconnection.

[0214] S812, when the communication operation status of the distribution radio area is autonomous, power dispatch is carried out according to the third dispatch scheme using the distribution radio area after isolation and disconnection.

[0215] The methods described in each of the above steps have been described in the foregoing embodiments. For details, please refer to the foregoing descriptions. They will not be repeated here.

[0216] 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.

[0217] Based on the same inventive concept, this application also provides a fault scheduling and control device for a distribution radio station to implement the fault scheduling and control method for the distribution radio station involved above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of the fault scheduling and control device embodiments for one or more distribution radio stations provided below can be found in the limitations of the fault scheduling and control method for distribution radio stations above, and will not be repeated here.

[0218] In one exemplary embodiment, such as Figure 10 As shown, a fault dispatch control device for a distribution radio area is provided, comprising:

[0219] The preprocessing module 11 is used to collect electrical signal data from the distribution transformer area and perform data preprocessing to obtain a set of data features.

[0220] Topology module 12 is used to construct a topology map of the distribution area based on the node relationships of the distribution area;

[0221] The collaboration module 13 is used to obtain the initial fault section, fault path edge coordinates and weight vectors of each node of the distribution transformer area based on the data feature set, transformer area topology map and multi-node collaboration strategy.

[0222] The fusion module 14 is used to determine the target fault section of the distribution radio area based on the initial fault section, the edge coordinates of the fault path, and the weight vector of each node.

[0223] The scheduling module 15 is used to isolate and remove the faulty area of ​​the distribution radio station according to the hardware constraints of the distribution radio station and the target faulty section, and to use the isolated distribution radio station for power dispatch.

[0224] Each module in the fault dispatch control device of the aforementioned distribution radio area can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0225] 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:

[0226] Collect electrical signal data from the distribution radio station area and perform data preprocessing to obtain a set of data features;

[0227] Based on the node relationships of the distribution transformer areas, a transformer area topology map is constructed;

[0228] Based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy, the initial fault section, fault path edge coordinates, and weight vectors of each node of the distribution transformer area are obtained.

[0229] The target fault section of the distribution transformer area is determined based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node.

[0230] Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cleared, and the isolated distribution transformer area is used for power dispatch.

[0231] 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:

[0232] Collect electrical signal data from the distribution radio station area and perform data preprocessing to obtain a set of data features;

[0233] Based on the node relationships of the distribution transformer areas, a transformer area topology map is constructed;

[0234] Based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy, the initial fault section, fault path edge coordinates, and weight vectors of each node of the distribution transformer area are obtained.

[0235] The target fault section of the distribution transformer area is determined based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node.

[0236] Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cleared, and the isolated distribution transformer area is used for power dispatch.

[0237] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0238] Collect electrical signal data from the distribution radio station area and perform data preprocessing to obtain a set of data features;

[0239] Based on the node relationships of the distribution transformer areas, a transformer area topology map is constructed;

[0240] Based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy, the initial fault section, fault path edge coordinates, and weight vectors of each node of the distribution transformer area are obtained.

[0241] The target fault section of the distribution transformer area is determined based on the initial fault section, the edge coordinates of the fault path, and the weight vectors of each node.

[0242] Based on the hardware constraints of the distribution transformer area and the target fault section, the fault area of ​​the distribution transformer area is isolated and cleared, and the isolated distribution transformer area is used for power dispatch.

[0243] 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.

[0244] 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.

[0245] 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 power dispatching method for a distribution substation, characterized in that, The method includes: Electrical signal data of the distribution transformer area is collected and preprocessed to obtain a data feature set; the data feature set is used to characterize the electrical response characteristics of each monitoring node of the distribution transformer area to fault events. Based on the node relationships of the distribution radio areas, a topology map of the radio areas is constructed; Based on the data feature set, the distribution area topology map, and the multi-node collaborative strategy, the initial fault segment, fault path edge coordinates, and weight vectors of each node in the distribution area are obtained. The multi-node collaborative strategy includes a graph-constrained linear model, a waveform similarity network model, and a node credibility prior model. The graph-constrained linear model is used to characterize the linear mapping relationship between the time-series observations of each monitoring node and the distribution area topology map. The waveform similarity network model is used to characterize the correlation between the transient waveforms of faults among each monitoring node. The node credibility prior model is used to quantify the credibility level of each monitoring node. The target fault section of the distribution transformer area is determined based on the initial fault section, the fault path edge coordinates, and the weight vectors of each node. Based on the hardware constraints of the distribution radio station and the target fault section, the fault area of ​​the distribution radio station is isolated and cut off, and power dispatch is carried out using the isolated distribution radio station.

2. The method according to claim 1, characterized in that, The step of obtaining the initial fault segment, fault path edge coordinates, and weight vectors of each node of the distribution transformer area based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy includes: The data feature set is input into the single-node initial value model to estimate the distance to each node in the transformer area topology map, thereby obtaining the initial position of each node from the initial fault point; the single-node initial value model includes an impedance mutation unit and a transient energy center unit; the impedance mutation unit and the transient energy center unit operate in parallel; Based on the data feature set, the multi-node collaboration strategy, and the initial position, the fault point is calculated to obtain the initial fault segment, the fault path edge coordinates, and the weight vector of each node in the transformer area topology map.

3. The method according to claim 2, characterized in that, The step of calculating the fault point based on the data feature set, the multi-node collaboration strategy, and the initial position to obtain the initial fault segment, the fault path edge coordinates, and the weight vector of each node in the transformer area topology map includes: Based on the data feature set and the graph-constrained linear model, the initial position is solved to obtain the edge coordinates of the fault path; The initial fault segment is obtained by analyzing the fault waveform based on the waveform similarity network model. The credibility of each node in the transformer area topology is weighted according to the node credibility prior model to obtain the node weight vector.

4. The method according to claim 1, characterized in that, Determining the target fault section of the distribution transformer area based on the initial fault section, the fault path edge coordinates, and the weight vectors of each node includes: The initial fault segment and the fault path edge coordinates are scale-aligned to obtain an initial fault segment set. The target fault segment is obtained by calculating based on the initial set of fault segments and the weight vector of each node.

5. The method according to claim 1, characterized in that, The step of isolating and disconnecting the faulty area of ​​the distribution transformer area based on the hardware constraints of the distribution transformer area and the target faulty section, and then using the isolated distribution transformer area for power dispatching, includes: Based on the hardware constraints, the scheduling scheme is boundary-defined, and a scheduling model is constructed. Based on the scheduling model, the faulty area of ​​the distribution radio station is isolated and removed, and a scheduling scheme is obtained.

6. The method according to any one of claims 1-5, characterized in that, The method further includes: The target scheduling scheme is determined based on the communication operation status of the distribution radio area after isolation and removal; the communication operation status includes any one of normal state, degraded state, and autonomous state; The method of power dispatching using the isolated distribution transformer area includes: When the communication operation status of the distribution radio area is normal or degraded, power dispatch is carried out according to the first dispatching scheme using the isolated distribution radio area. When the communication operation state of the distribution radio area is the autonomous state, power dispatch is carried out according to the second dispatch scheme using the distribution radio area after isolation and disconnection.

7. A fault dispatch control device for a distribution radio area, characterized in that, The device includes: The preprocessing module is used to collect electrical signal data from the distribution radio station area and perform data preprocessing to obtain a set of data features. The topology module is used to construct a topology map of the distribution area based on the node relationships of the distribution area. The collaboration module is used to obtain the initial fault section, fault path edge coordinates, and weight vector of each node of the distribution transformer area based on the data feature set, the transformer area topology map, and the multi-node collaboration strategy. The fusion module is used to determine the target fault section of the distribution radio area based on the initial fault section, the fault path edge coordinates, and the weight vectors of each node; The scheduling module is used to isolate and remove the faulty area of ​​the distribution radio station according to the hardware constraints of the distribution radio station and the target faulty section, and to use the isolated distribution radio station for power dispatch.

8. 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 6.

9. 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 6.

10. A computer program product, comprising a computer program, 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 6.

Images