Fault location method and system for ring-type submarine direct current power supply system

Abstract

This invention discloses a fault location method and system for a ring-shaped submarine DC power supply system. The method includes: deconstructing the ring-shaped submarine DC power supply system into a simplified topology network; setting measurement points and a local fault location model based on the branch units in the simplified topology network; performing feature analysis on the local fault current data collected at the branch units; inputting the fault features into the local fault location model in time step order for fault prediction; and sending the fault line and fault location to the shore base station based on the predicted fault location and the fault distance between the branch unit to complete the fault location. This invention simplifies the complex ring-shaped submarine DC power supply system by simplifying the topology, thereby reducing the number of lines and measurement points required. By selecting fault current features that are not affected by changes in the number of LLC branches, the regularity of the features is improved, thus improving the fault location accuracy. This method is effectively adapted to scenarios with few measurement points in the ring-shaped submarine DC power supply system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power systems, in particular to a ring-type seabed DC power supply system fault positioning method and system. BACKGROUND

[0002] The seabed observation network is an important infrastructure for protecting territorial security, developing marine resources and disaster warning. The DC power supply system is the "energy heart" of the seabed observation network, providing power support for deep-sea observation, detection equipment, etc., generally adopting a single-pole 10kV power supply mode. Most seabed observation projects reach a depth of 1000m under the seabed, with a constant temperature of 2-4℃ and a pressure of more than 10MPa, in a high-pressure and low-temperature environment. The seabed cable laying environment is special, with a high failure rate, the seabed observation network is complex, the seabed cable is long, and the damage to the system is great when a fault occurs, so it is urgent to study the fault positioning means of the seabed DC power supply system to locate the specific fault position of the fault cable to maintain the seabed system. For a ring-type seabed DC power supply system, on the one hand, electrical faults of seabed cables are often accompanied by interruption of optical fiber communication links, causing the reliability of fault diagnosis methods that rely on wired communication to decrease significantly. On the other hand, the multi-terminal ring-type seabed DC power supply system has limited measurement nodes and data, and the traditional fault positioning method based on local current data requires measurement points to be arranged on each line, so the traditional method is difficult to cover the entire network in the ring-type seabed DC power supply system.

[0003] At present, the fault positioning research of the ring-type seabed DC system has developed various methods such as traveling wave detection, Kirchhoff's law analysis, impedance spectrum analysis, leakage current vector method and local current data method, but all have some defects: the traveling wave method has difficulty in identifying the wave front due to the severe high-frequency attenuation in the seabed environment, which seriously restricts the positioning accuracy; the Kirchhoff's law method requires strict synchronization of multi-node measurement data, which has high computational complexity and strict real-time communication requirements; the impedance spectrum analysis method relies on wideband excitation signal injection and high-reliability communication link support, which is difficult to implement in a complex ring network with sparse measurement; the leakage current vector method is easily disturbed by system load fluctuations and power disturbances, has large calculation redundancy and requires continuous communication support; the local current data method avoids communication dependence, but requires dense measurement points in a multi-branch ring network. These methods cannot simultaneously adapt to the problems of unreliable communication resources and sparse measurement nodes in the ring-type seabed DC power supply system. SUMMARY

[0004] The main purpose of the present application is to provide a ring-type seabed DC power supply system fault positioning method and system, which aims to solve the technical problem of low fault positioning accuracy in the prior art in the ring-type seabed DC system due to unreliable communication resources and sparse measurement nodes.

[0005] To achieve the above object, the application provides a ring-type seabed direct-current power supply system fault positioning method, which comprises the following steps:

[0006] The ring-type seabed direct-current power supply system is decomposed into a simplified topology network, and a measuring point and a local fault positioning model corresponding to the measuring point are set based on a branch unit in the simplified topology network;

[0007] The local fault current data collected at the branch unit are subjected to feature analysis to obtain fault features;

[0008] The fault features are input into the local fault positioning model in time step order for fault prediction to obtain a fault distance between the fault position and the branch unit, and the local fault positioning model is constructed based on a pre-trained LSTM network;

[0009] The fault line and the fault position are determined based on the fault distance, and the fault line and the fault position are sent to a shore base station to complete fault positioning.

[0010] Optionally, the ring-type seabed direct-current power supply system is decomposed into a simplified topology network, comprising:

[0011] The ring-type seabed direct-current power supply system is decomposed to obtain a radial network;

[0012] The lines at both ends of the nodes in the radial network that meet the line merging condition are merged to obtain a simplified topology network, so as to realize topology simplification of the network, and the line merging condition is that the number of connections of the node is 2.

[0013] Optionally, the ring-type seabed direct-current power supply system is decomposed to obtain a radial network, comprising:

[0014] The longest connection path between two shore base stations of the ring-type seabed direct-current power supply system is determined, and the ring-type seabed direct-current power supply system is reconstructed based on the longest connection path to obtain a reconstructed network, and the reconstructed network comprises a multi-connected ring network and a single ring network;

[0015] The multi-connected ring network is decomposed by taking the intersection points on the non-longest connection paths as decomposition points;

[0016] The node position of the longest connection path of the single ring network is determined, the shortest path between the node position of the longest connection path and the two shore base stations is calculated, the decomposition points are determined based on the nodes in the shortest path, and the single ring network is decomposed based on the decomposition points;

[0017] The radial network is obtained based on the decomposition results of the multi-connected ring network and the decomposition results of the single ring network.

[0018] Optionally, the fault characteristics include instantaneous slope and instantaneous curvature; the step of performing feature analysis on the local fault current data collected at the branch unit to obtain fault characteristics includes:

[0019] The start time for characteristic calculation is determined based on the fault initiation time, referring to the following formula:

[0020]

[0021]

[0022] in, This represents the inductance per unit length of the submarine cable. Capacitance per unit length of submarine cable. The time when the fault begins. This is the empirical offset. The distance to the fault. The speed at which current propagates in a submarine cable. Calculate the start time for the feature;

[0023] Based on the aforementioned feature calculation start time, feature analysis is performed on the local fault current data collected at the branch unit to obtain the instantaneous slope and instantaneous curvature of the fault current, referring to the following formula:

[0024]

[0025]

[0026]

[0027]

[0028]

[0029] in, The instantaneous slope of the fault current. Indicates the start time of feature calculation Next fault current data, express The next fault current data, Indicates the sampling time interval. and Indicates the fault current difference. The instantaneous curvature of the fault current. Indicates the sampling frequency.

[0030] Optionally, the calculation formula for each unit in the local fault location model is as follows:

[0031]

[0032] in, , and These represent the output value, weight coefficient, and bias of the forget gate, respectively. Indicates the first Input volume at any given time Indicates the first The state value of the hidden layer at any given time. , and These represent the output value, weighting coefficient, and bias of the input gate, respectively. , and These represent the output value, weighting coefficient, and bias of the output gate, respectively. , and These represent the output value, weight coefficient, and bias, respectively, during the feature extraction process. and These represent the activation function and the hyperbolic tangent function, respectively. Indicates the input gate and the first The weighting coefficients associated with the input values ​​at different times. Indicates the output gate and the first The weighting coefficients associated with the input values ​​at different times. Indicates the feature extraction output and the first The weighting coefficients associated with the input values ​​at different times. Indicates the output of the forget gate and the first... Weighting coefficients associated with input values ​​at different times;

[0033] The activation function and the hyperbolic tangent function are respectively:

[0034]

[0035] in, Input data;

[0036] The local fault location model is configured to determine the cell state value at each time step based on the output value of each unit, referring to the following formula:

[0037]

[0038] in, and Representing time respectively and time Cell state values, This indicates element-wise multiplication;

[0039] The local fault location model is further configured to determine the hidden layer state value at each time step based on the cell state value and the output value during the feature extraction process, and to convert the hidden layer state value into the fault location at each time step through the output mapping module. The hidden layer state value is calculated with reference to the following formula:

[0040]

[0041] in, Indicates time The hidden layer state values.

[0042] Optionally, before inputting the fault features into the local fault location model in time step order for fault prediction, the method further includes:

[0043] Send a data sharing request to other branch units in the simplified topology network so that the other branch units can send local model parameters based on the data sharing request. The local model parameters are obtained by each branch unit after training a local fault location model based on local fault characteristics and historical fault sample data.

[0044] The local fault location model is pre-trained based on historical fault sample data to obtain initial model parameters.

[0045] Based on the historical fault location reliability and request response speed of other branch units, the initial model parameters are aggregated with the local model parameters sent by the other branch units to obtain global model parameters;

[0046] The local fault location model is updated based on the global model parameters to obtain the final local fault location model, and the step of inputting the fault features into the local fault location model in time step order for fault prediction is executed.

[0047] Furthermore, to achieve the above objectives, the present invention also proposes a fault location system for a ring-shaped submarine DC power supply system, the ring-shaped submarine DC power supply system fault location system comprising:

[0048] The system loop-breaking module is used to break the ring-shaped submarine DC power supply system into a simplified topology network, and set measurement points and local fault location models corresponding to the measurement points based on the branch units in the simplified topology network.

[0049] The feature analysis module is used to perform feature analysis on the local fault current data collected at the branch unit to obtain fault characteristics;

[0050] The fault analysis module is used to input the fault features into the local fault location model in the order of time steps to predict the fault and obtain the fault distance between the fault location and the branch unit. The local fault location model is built based on a pre-trained LSTM network.

[0051] The fault location module is used to determine the faulty line and the fault location based on the fault distance, and send the faulty line and the fault location to the shore base station to complete the fault location.

[0052] Optionally, the system loop-breaking module is further used to break the loop in the ring-shaped submarine DC power supply system to obtain a radial network; and to merge the lines at both ends of the nodes in the radial network that meet the line merging condition to obtain a simplified topology network, thereby realizing topology simplification of the network. The line merging condition is that the number of connected nodes is 2.

[0053] Optionally, the system loop-breaking module is further configured to determine the longest connection path between two shore base stations of the ring-shaped submarine DC power supply system, and reconstruct the ring-shaped submarine DC power supply system based on the longest connection path to obtain a reconstructed network, the reconstructed network including a multi-ring network and a single-ring network; decompose the multi-ring network by using the intersection points on the non-longest connected paths in the multi-ring network as decomposition points; determine the node positions of the longest connected path in the single-ring network, calculate the shortest path between the node positions of the longest connected path and the two shore base stations, determine the decomposition points based on the nodes in the shortest path, and decompose the single-ring network based on the decomposition points; and obtain a radial network based on the loop-breaking results of the multi-ring network and the single-ring network.

[0054] Optionally, the fault features include instantaneous slope and instantaneous curvature; the feature analysis module is further configured to determine the feature calculation start time based on the fault initiation time, referring to the following formula:

[0055]

[0056]

[0057] in, This represents the inductance per unit length of the submarine cable. Capacitance per unit length of submarine cable. The time when the fault begins. This is the empirical offset. The distance to the fault. The speed at which current propagates in a submarine cable. Calculate the start time for the feature;

[0058] Based on the aforementioned feature calculation start time, feature analysis is performed on the local fault current data collected at the branch unit to obtain the instantaneous slope and instantaneous curvature of the fault current, referring to the following formula:

[0059]

[0060]

[0061]

[0062]

[0063]

[0064] in, The instantaneous slope of the fault current. Indicates the start time of feature calculation Next fault current data, express The next fault current data, Indicates the sampling time interval. and Indicates the fault current difference. The instantaneous curvature of the fault current. Indicates the sampling frequency.

[0065] This invention deconstructs a ring-shaped submarine DC power supply system into a simplified topology network. Based on the branch units within this simplified topology network, measurement points and corresponding local fault location models are set up. Feature analysis is performed on the local fault current data collected at the branch units to obtain fault features. These fault features are then input into the local fault location model in time step order for fault prediction, obtaining the fault distance between the fault location and the branch unit. The local fault location model is constructed based on a pre-trained LSTM network. Based on the fault distance, it determines the faulty line and fault location, and sends the faulty line and fault location to a shore base station to complete fault location. Because this invention simplifies the complex ring-shaped submarine DC power supply system topology, it reduces the number of lines and measurement points required. By selecting fault current features unaffected by changes in the number of LLC branches, the regularity of the features is improved, thereby enhancing fault location accuracy. This invention effectively adapts to scenarios with few measurement points in ring-shaped submarine DC power supply systems. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 This is a schematic diagram of the structure of the fault location device for the ring-shaped submarine DC power supply system in the hardware operating environment involved in the embodiments of the present invention;

[0068] Figure 2 This is a flowchart illustrating the first embodiment of the fault location method for a ring-shaped submarine DC power supply system according to the present invention.

[0069] Figure 3 This is a schematic diagram of the structure of a ring-shaped DC power supply system in one embodiment of the fault location method for a ring-shaped submarine DC power supply system according to the present invention;

[0070] Figure 4 This is a schematic diagram of the reconstructed network structure in one embodiment of the fault location method for a ring-shaped submarine DC power supply system of the present invention;

[0071] Figure 5 This is a simplified topology diagram of an embodiment of the fault location method for a ring-shaped submarine DC power supply system of the present invention;

[0072] Figure 6(a) is a schematic diagram of the merged line topology of line B1 in one embodiment of the fault location method of the ring-shaped submarine DC power supply system of the present invention;

[0073] Figure 6(b) is a schematic diagram of the combined line topology of line B7 in one embodiment of the fault location method of the ring-shaped submarine DC power supply system of the present invention.

[0074] Figure 6(c) is a schematic diagram of the merged line topology of line B9 in one embodiment of the fault location method of the ring-shaped submarine DC power supply system of the present invention;

[0075] Figure 7 This is a flowchart illustrating the second embodiment of the fault location method for a ring-shaped submarine DC power supply system of the present invention.

[0076] Figure 8 This is a structural block diagram of the first embodiment of the fault location system for the ring-shaped submarine DC power supply system of the present invention.

[0077] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0078] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0079] Reference Figure 1 , Figure 1 This is a schematic diagram of the fault location device structure of the ring-shaped submarine DC power supply system in the hardware operating environment involved in the embodiments of the present invention.

[0080] like Figure 1 As shown, the fault location device for the ring-shaped submarine DC power supply system may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be high-speed random access memory (RAM) or stable non-volatile memory (NVM), such as a disk storage device. Optionally, the memory 1005 may also be a storage system independent of the aforementioned processor 1001.

[0081] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the fault location device for the ring-shaped submarine DC power supply system. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0082] like Figure 1 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a fault location program for a ring-shaped submarine DC power supply system.

[0083] exist Figure 1 In the fault location device for the ring-shaped submarine DC power supply system shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the fault location device for the ring-shaped submarine DC power supply system of the present invention can be set in the fault location device for the ring-shaped submarine DC power supply system. The fault location device for the ring-shaped submarine DC power supply system calls the fault location program for the ring-shaped submarine DC power supply system stored in the memory 1005 through the processor 1001 and executes the fault location method for the ring-shaped submarine DC power supply system provided in the embodiment of the present invention.

[0084] This invention provides a fault location method for a ring-shaped submarine DC power supply system, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the fault location method for the ring-shaped submarine DC power supply system of the present invention.

[0085] In this embodiment, the fault location method for the ring-shaped submarine DC power supply system includes the following steps:

[0086] Step S10: De-loop the ring-shaped submarine DC power supply system into a simplified topology network, and set measurement points and local fault location models corresponding to the measurement points based on the branch units in the simplified topology network.

[0087] It should be understood that the executing entity of this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or a terminal electronic device capable of performing the above functions. The following description uses a fault location device for a ring-shaped submarine DC power supply system (hereinafter referred to as the location device) as an example to illustrate this embodiment and the following embodiments.

[0088] It should be noted that the fault location model uses local current data for fault location. Measurement points are used to collect fault current data.

[0089] It is understood that this embodiment obtains a radial network by de-looping the ring-shaped submarine DC power supply system, simplifies the topology of the radial network, and arranges measurement points and fault location models according to the simplified topology network, thereby reducing the number of measurement points and fault location models.

[0090] It should be understood that the fault location model is deployed locally, and each line has its own local fault location model. By detecting sudden changes in current, it is possible to directly determine which line has a fault. The fault location model is then used to obtain the fault distance by inputting features into it, thus completing the fault location.

[0091] In practical implementation, loop decomposition involves merging the circuits, simplifying the circuitry and thus reducing the number of required measurement points. This is because local model-based fault location methods require installing measurement points on each circuit. My method merges the circuits, essentially combining several circuits into a single merged circuit, requiring only one measurement unit for location, thereby reducing the number of measurement units needed. The merged circuit has several LLCs (Limited Lines), and LLC branching is common. Selecting fault characteristics unaffected by LLC branches can improve the regularity of these characteristics, thereby increasing fault location accuracy.

[0092] Furthermore, in order to effectively simplify the ring-shaped submarine DC power supply system and thus accurately arrange measurement points and fault location models, step S10 above may include:

[0093] Step S101: De-loop the ring-shaped submarine DC power supply system to obtain a radial network;

[0094] Step S102: Merge the lines at both ends of the nodes that meet the line merging condition in the radial network to obtain a simplified topology network, thereby simplifying the network topology. The line merging condition is that the number of connected nodes is 2.

[0095] Understandably, this embodiment obtains a radial network by de-looping the ring-shaped submarine DC power supply system, merges the lines at both ends of the two-node connected components to obtain a simplified topology, and arranges measurement points and fault location models based on the simplified topology, thus reducing the number of measurement points and fault location models. With the merged lines becoming longer and containing several LLCs, the adaptability of the fault location model is improved by selecting fault current characteristics unaffected by changes in the number of LLC branches: the instantaneous slope and curvature of the fault current. Different fault characteristics are selected for calculating the start time of the fault location model for lines of different lengths. To ensure the salience of fault characteristics.

[0096] It should be noted that the measurement points and fault location models are deployed locally in the branch unit. A branch unit can connect up to three lines. Therefore, each line can have one measurement point and fault location model deployed accordingly, thus fault location does not require communication. The merged line obtained by merging lines has LLC branches and the line length increases. By setting the start time of fault feature calculation, the significance of features is improved, and fault features less affected by LLC branches are selected to improve fault location accuracy.

[0097] For example, one measurement unit is required for one line. The original topology had 26 lines, but after merging, there are only 13 lines, so only 13 measurement units are needed. A maximum of 3 measurement units can be arranged in a branch unit. According to the method described in this embodiment, only 13 measurement elements need to be arranged in the shore base and 4 branch units to complete the fault location of the entire line, which can adapt to the environment of sparse measurement.

[0098] In its implementation, the positioning device de-loops the ring network to obtain a radial network; it then merges the lines at both ends of nodes with a connectivity of 2 in the radial network to obtain a simplified topology; measurement points and a fault location model are then arranged within the simplified topology. Local data is collected, features are calculated and input into the fault location model to obtain the fault distance, thus completing the fault location.

[0099] Furthermore, in order to effectively deloop-free different types of reconstructed networks, step S101 above may include:

[0100] Step S1011: Determine the longest connection path between two shore base stations of the ring-shaped submarine DC power supply system, and reconstruct the ring-shaped submarine DC power supply system based on the longest connection path to obtain a reconstructed network, which includes a multi-ring network and a single-ring network;

[0101] Step S1012: De-loop the multi-ring network by using the intersection points on the non-longest connected paths in the multi-ring network as decomposition points;

[0102] Step S1013: Determine the node positions of the longest connected path in the single-ring network, calculate the shortest path between the node positions of the longest connected path and the two shore base stations, determine the decomposition points based on the nodes in the shortest path, and de-loop the single-ring network based on the decomposition points.

[0103] Step S1014: Obtain the radial network based on the de-looping results of the multi-ring network and the single-ring network.

[0104] In one embodiment, a schematic diagram of the ring-shaped DC power supply system is shown below. Figure 3 All main cables are 10km long, with 22 branch units, 2 shore base stations, and 26 submarine cables. Figure 3 To de-loop the submarine ring system shown, first determine the longest connection path between the two shore base stations, then reconstruct the network as follows: Figure 4 As shown, Figure 4 This diagram illustrates the reconstructed network structure. Ring network de-ringing is divided into multi-ring network decomposition and single-ring network decomposition. For multi-ring network decomposition, the intersection points of multiple ring networks on non-longest connected paths are used as decomposition points, ensuring that the multiple ring networks are completely decomposed and all branches connect to the longest connected path. For single-ring network decomposition, the node positions P1 and P2 of the ring network connected to the longest connected path need to be determined, and the shortest distances L1min and L2min between these nodes and the two shore base stations are calculated. The largest maximum value, max{L1min, L2min}, is selected from the shortest distances, and its corresponding node is used as the decomposition point. The decomposition points of the reconstructed network are shown below. Figure 5 As shown, Figure 5 To simplify the topology diagram, in Figure 5 In the text, S5 and S6 represent shore base stations, , , M1, M2, M3, and M4 represent measurement nodes, where... , and Measurement nodes are set up at the decomposition nodes of a multi-ring network.

[0105] After unblocking the loop, a two-ended radial network is obtained. In the radial network topology, if a node is connected to... The connectivity of a node is defined as the number of nodes connected together. For branch units, due to structural limitations, their maximum connectivity is 3. Nodes with a connectivity of 2 primarily function as connectors in the backbone network, merging the lines at both ends of a node with a connectivity of 2 to simplify the topology, resulting in a simplified topology as follows: Figure 5 As shown. After the line merging, some lines become longer, and are divided into four lengths: 10km, 20km, 40km, and 50km. The 10km lines are those that have not undergone simplification and merging. The 20km lines are: B1, B4, B5, B8, B12, and B13; the 40km line is B7; and the 50km line is B9. The topology diagrams of lines B1, B7, and B9 are shown in Figures 6(a), 6(b), and 6(c). Figure 6(a) is a schematic diagram of the merged line topology of line B1, Figure 6(b) is a schematic diagram of the merged line topology of line B7, and Figure 6(c) is a schematic diagram of the merged line topology of line B9. Taking line B9 as an example, the traditional method of fault location using local current data requires installing fault location models on all five lines, requiring more measurement points. However, after the line merging, only one fault location model needs to be installed at one end of the line, which greatly reduces the number of measurement points and fault models required.

[0106] After the lines are merged, the lines become longer, and the measured current data decreases with increasing fault distance. However, its characteristics compared to when no fault has occurred are still significant. The fault current data collected at the end side still contains rich information and can support fault location for the entire line. The general principle for selecting measurement points in topology simplification is to prioritize branches of non-longest connection paths. The specific selection steps are as follows: First, select the decomposition points of the multi-terminal ring network as measurement points to handle multiple branches. Second, select nodes with branches as measurement points. Third, handle other lines. Arranging measurement points after topology simplification can reduce the number of measurement points, adapting to the problem of fewer measurement points in ring-shaped submarine DC power supply systems.

[0107] Step S20: Perform feature analysis on the local fault current data collected at the branch unit to obtain fault characteristics.

[0108] It should be noted that the fault characteristics include instantaneous slope and instantaneous curvature. The instantaneous slope of the fault current directly reflects the steepness of the rising phase of the fault current, and its physical essence is the instantaneous rate of change of the current with respect to time; while its instantaneous curvature characterizes the bending characteristics of the current waveform, essentially reflecting the changing law of the second derivative of the current. When a short-circuit fault occurs in the system, the fault current increases sharply, and its waveform bending characteristics are significant. This indicates that the instantaneous slope and instantaneous curvature of the fault current are highly distinctive key features.

[0109] Understandably, in a ring-shaped submarine DC power supply system, the -10kV DC power is converted to 375V DC power in the main junction box (PJB) using an LLC converter. The LLC and its load are referred to as the LLC branch. However, considering that the resistance of the submarine fiber optic composite cable is as high as 1Ω per kilometer, which is tens or hundreds of times higher than that of a terrestrial DC cable, and that the equivalent capacitance of the MMC and LLC is much larger than the equivalent fault capacitance of a terrestrial DC power supply system, the fault current of a ring-shaped submarine DC power supply system can be equivalent to the non-oscillating current of an RLC series circuit, unlike terrestrial DC power supply systems. The transient process of the LLC during a system fault can be equivalent to the input capacitance. Variations in the number of LLC branches in submarine DC power supply systems are common, and these variations primarily affect the circuit's equivalent capacitance. The magnitude of the fault current in a ring-shaped submarine DC power supply system mainly depends on the square root of the system voltage, equivalent inductance, and equivalent capacitance, under overdamped conditions. When the value is much greater than 4L, changes in the number of LLC branches have a negligible effect on the circuit's equivalent capacitance, and the change in equivalent capacitance has a negligible effect on the fault current amplitude. Therefore, changes in the number of LLC branches have a small impact on the fault current amplitude. The instantaneous slope and instantaneous curvature of the fault current are related to the change in the fault current amplitude per unit time. Changes in the number of LLC branches have a small impact on the fault current amplitude; therefore, their impact on the instantaneous slope and instantaneous curvature of the fault current is also negligible. Furthermore, changes in the number of LLC branches only alter... The value of does not change the essential nature of the RLC topology of the system; therefore, the overall trend of the instantaneous slope and curvature of the fault current changing with the fault distance will not change due to the change in the number of LLC branches. From the above analysis, it can be concluded that the instantaneous slope and curvature of the fault current are fault characteristics that adapt to changes in the number of LLC branches. Finally, the nonlinear Pearson algorithm is used to evaluate the correlation between the instantaneous slope, curvature, and fault distance. The results show a strong correlation between the fault characteristics and the fault distance.

[0110] Furthermore, in order to accurately obtain fault current characteristics and improve fault location accuracy, step S20 above may include:

[0111] Step S201: Determine the feature calculation start time based on the fault start time.

[0112] It is understandable that this embodiment adapts to different fault location models for line layouts of varying lengths. The main difference in the fault location models lies in the starting time of fault feature calculation. Since the lines become longer after merging, the current transmission delay at the farthest point of the line must be considered to determine the starting time for calculating fault features. To ensure the significance of the fault features, the time for fault calculation is shifted backward. Refer to the following formula:

[0113]

[0114]

[0115] in, This represents the inductance per unit length of the submarine cable. Capacitance per unit length of submarine cable. The time when the fault begins. This is the empirical offset. The distance to the fault. The speed at which current propagates in a submarine cable. The start time is calculated for the feature.

[0116] Step S202: Based on the feature calculation start time, perform feature analysis on the local fault current data collected at the branch unit to obtain the instantaneous slope and instantaneous curvature of the fault current.

[0117] Understandably, when measuring local fault current data within 2ms at a branch unit of the submarine DC power supply system, the measured data is discrete. The instantaneous slope and instantaneous curvature of the fault current are calculated using the following formula:

[0118]

[0119]

[0120]

[0121]

[0122]

[0123] in, The instantaneous slope of the fault current. Indicates the start time of feature calculation Next fault current data, express The next fault current data, Indicates the sampling time interval. and Indicates the fault current difference. The instantaneous curvature of the fault current. Indicates the sampling frequency.

[0124] Step S30: Input the fault features into the local fault location model in the order of time steps to predict the fault and obtain the fault distance between the fault location and the branch unit. The local fault location model is built based on a pre-trained LSTM network.

[0125] It is understood that in this embodiment, the calculated fault features are sequentially arranged and input into the trained LSTM network to obtain the distance from the fault point to the measurement branch unit, and the fault line and location are sent to the shore base station to complete the fault location.

[0126] In some embodiments, the pre-trained LSTM network has 64 LSTM layers, 32 intermediate fully connected layers, an activation function of tanh, a maximum training epoch of 500, an initial learning rate of 0.001, a learning rate decay factor of 0.5, and a learning rate epoch of 100.

[0127] Furthermore, to improve the accuracy of fault location prediction, in one embodiment, the calculation formula for each unit in the local fault location model is as follows:

[0128]

[0129] in, , and These represent the output value, weight coefficient, and bias of the forget gate, respectively. Indicates the first Input volume at any given time Indicates the first The state value of the hidden layer at any given time. , and These represent the output value, weighting coefficient, and bias of the input gate, respectively. , and These represent the output value, weighting coefficient, and bias of the output gate, respectively. , and These represent the output value, weight coefficient, and bias, respectively, during the feature extraction process. and These represent the activation function and the hyperbolic tangent function, respectively. Indicates the input gate and the first The weighting coefficients associated with the input values ​​at different times. Indicates the output gate and the first The weighting coefficients associated with the input values ​​at different times. Indicates the feature extraction output and the first The weighting coefficients associated with the input values ​​at different times. Indicates the output of the forget gate and the first... Weighting coefficients associated with input values ​​at different times;

[0130] The activation function and the hyperbolic tangent function are respectively:

[0131]

[0132] in, Input data;

[0133] The local fault location model is configured to determine the cell state value at each time step based on the output value of each unit. According to the structural relationship of LSTM, the first... The cell state value at time t is given by the following formula:

[0134]

[0135] in, and Representing time respectively and time Cell state values, This indicates element-wise multiplication;

[0136] The local fault location model is further configured to determine the hidden layer state value at each time step based on the cell state value and the output value during the feature extraction process, and to convert the hidden layer state value into the fault location at each time step through the output mapping module. The hidden layer state value is calculated with reference to the following formula:

[0137]

[0138] in, Indicates time The hidden layer state values.

[0139] Step S40: Determine the faulty line and fault location based on the fault distance, and send the faulty line and fault location to the shore base station to complete the fault location.

[0140] It is understandable that this embodiment simplifies the complex ring network by ring network de-ringing and line merging. Based on the simplified topology, the number of measurement points and fault location models can be reduced. For merged lines, the fault location model uses local current data for fault location, which can adapt to changes in the number of LLC branches.

[0141] It should be understood that this embodiment addresses the problems of unreliable communication and sparse measurement points in a ring-shaped submarine DC power supply system. This embodiment achieves fault location by deploying a fault location model locally and collecting local data, eliminating the need for communication. The ring network is de-looped by merging the lines at both ends of nodes with a connectivity of 2, resulting in a simplified topology. Compared to the original topology, the simplified topology reduces the number of lines, thus reducing the required measurement components to adapt to sparse measurement environments. (The fault location method based on the local model requires the deployment of measurement points on each line.)

[0142] To verify the effectiveness of this invention, a model was established in MATLAB / Simulink as follows: Figure 3 The ring-shaped submarine DC power supply system model shown has shore base station 1 and shore base station 2 with a rated power of 0.5MW. The submarine cable is simulated using a lumped RL model, and the simulation time is 1.2s. The fault occurred within 1 second. Other parameters are shown in Table 1.

[0143] Table 1, Model Simulation Parameter Table

[0144]

[0145] exist Figure 5 Faults were set up on four merged lines of different lengths: B1, B7, B9, and B11. For each line, a fault point was set every 0.1 km, and each fault point had six different fault resistors ranging from 0.1Ω to 200Ω. The sample set for line B1 contained 1200 data sets, line B7 2400 data sets, line B9 3000 data sets, and line B11 600 data sets. Five samples of fault resistors were used as the training set, and one sample of a fault resistor was used as the test set. The sample data consisted of fault characteristics, fault resistance, and fault distance, arranged in the order of [slope, curvature, fault resistance, fault distance].

[0146] When the number of LLC branches remains unchanged, the fault location errors for different lines are shown in Table 2. As can be seen from the table, the root mean square error for each line is within 0.066 km, the mean absolute error is within 0.058 km, and the mean relative error is within 0.9%. In summary, the method proposed in this paper exhibits low prediction errors for different lines, high prediction accuracy of the fault location model, and strong model adaptability during normal LLC operation.

[0147] Table 2. Comparison of Line Fault Location Error Parameters

[0148]

[0149] Lines B1, B7, and B9 all have LLC branches, and the number of LLC branches often changes. This invention uses data with no change in the number of LLC branches as the training set, and collects fault current data when the number of LLC branches decreases. The instantaneous slope and curvature are calculated to form a test set to verify the adaptive capability of the fault location model when the number of LLC branches changes. The number of data sets collected when the number of LLC branches is unchanged is the same as when the number of LLC branches changes. Each time, data from one fault resistor is used as the test set. Two operating conditions are set on line B7: LLC1 exits and both LLC1 and LLC2 exit. The fault location errors are shown in Tables 3 and 4, respectively. Table 3 shows the first fault location error parameter table, and Table 4 shows the second fault location error parameter table. As can be seen from the tables, the root mean square error, mean absolute error, and mean relative error of fault location are relatively similar when one LLC branch is reduced and two LLC branches exit. Under different fault resistances, the root mean square error is below 0.09 km, the mean absolute error is below 0.09 km, and the mean relative error is below 0.25%, indicating high fault location accuracy of the model. The above experiments demonstrate that this invention can achieve high-precision fault location for a ring-shaped submarine DC power supply system using fewer measurement points and a fewer fault location model.

[0150] Table 3, First Fault Location Error Parameter Table

[0151]

[0152] Table 4, Second Fault Location Error Parameter Table

[0153]

[0154] This embodiment deconstructs the ring-shaped submarine DC power supply system into a simplified topology network. Based on the branch units in this simplified topology network, measurement points and corresponding local fault location models are set up. Feature analysis is performed on the local fault current data collected at the branch units to obtain fault features. These fault features are then input into the local fault location model in time step order for fault prediction, obtaining the fault distance between the fault location and the branch unit. The local fault location model is constructed based on a pre-trained LSTM network. Based on the fault distance, it determines the faulty line and fault location, and sends the faulty line and fault location to the shore base station to complete fault location. Because this embodiment simplifies the complex ring-shaped submarine DC power supply system topology, it reduces the number of lines and measurement points required. By selecting fault current features unaffected by changes in the number of LLC branches, the regularity of the features is improved, thereby enhancing fault location accuracy. This effectively adapts to scenarios with few measurement points in the ring-shaped submarine DC power supply system.

[0155] refer to Figure 7 , Figure 7This is a flowchart illustrating the second embodiment of the fault location method for the ring-shaped submarine DC power supply system of the present invention.

[0156] Based on the first embodiment described above, in this embodiment, before step S30, the method further includes:

[0157] Step S31: Send a data sharing request to other branch units in the simplified topology network so that the other branch units can send local model parameters based on the data sharing request.

[0158] It should be noted that the local model parameters are obtained by each branch unit after training a local fault location model based on local fault characteristics and historical fault sample data.

[0159] It should be noted that the simplified topology network refers to the dual-ended radial network formed after the de-looping of the submarine ring DC power supply system. It consists of nodes such as shore base stations, branch units, and communication submarine cables, and possesses clear node communication links and data interaction paths. Branch units can be power supply equipment units deployed on the seabed, possessing functions such as fault data acquisition (e.g., fault current, voltage), local model training, and parameter communication; they serve as edge computing nodes for fault location. Data sharing requests can be communication commands initiated by branch units to other nodes to obtain model parameters, including request identifiers, parameter format requirements, and security encryption agreements.

[0160] Understandably, in order to improve fault location accuracy and ensure consistent performance of local fault location models in each branch unit of the simplified topology network, and to avoid inconsistent fault location accuracy due to data differences or other influencing factors, this embodiment can achieve model parameter sharing among branch units through federated learning mechanisms or other communication network mechanisms. This allows the model parameters of each branch unit to be aggregated to update the local fault location model, thereby obtaining the optimal fault location model and improving fault location accuracy.

[0161] In some embodiments, the local branch unit sends a data sharing request to other branch units within the simplified topology network via a submarine communication cable using an encrypted communication protocol (such as TLS). Upon receiving the request, other branch units call the parameter interface of their local model training module to extract the parameters of the trained fault location model. After desensitizing and encrypting the parameters, they transmit them back to the requesting branch unit via the original communication link. Historical fault sample data can be a collection of historical fault records stored locally by the branch unit, containing characteristic data at the time of the fault occurrence, actual fault location, and other annotation information.

[0162] Step S32: Pre-train the local fault location model based on historical fault sample data to obtain initial model parameters.

[0163] In some embodiments, the local branch unit retrieves historical fault sample data from the storage module and divides it into training set / validation set according to a preset ratio; it organizes the fault features into a model input format according to time steps, loads the preset fault location model (LSTM network), performs pre-training on the local embedded computing power module using the gradient descent algorithm, iterates until the validation set loss stabilizes, and then saves the current model parameters as the initial model parameters.

[0164] Step S33: Based on the historical fault location reliability and request response speed of other branch units, aggregate the initial model parameters with the local model parameters sent by the other branch units to obtain global model parameters.

[0165] It should be noted that the reliability of historical fault location can be an indicator of the accuracy of local model location in a branch unit, which is usually the matching rate between the historical fault location results and the actual fault location of the unit.

[0166] In some embodiments, the local branch unit statistically analyzes the historical fault location reliability (such as the location matching rate in the past 3 months) and request response speed (the time taken from sending a request to receiving parameters) of other branch units; for example, it calculates the parameter weight of each unit according to the weight rule of "70% reliability and 30% response speed" (the shorter the time, the higher the weight), and performs a weighted average of its own initial model parameters and the local model parameters of other units to obtain the global model parameters.

[0167] Step S34: Update the parameters of the local fault location model based on the global model parameters to obtain the final local fault location model, and execute the step of inputting the fault features into the local fault location model in time step order for fault prediction.

[0168] In some embodiments, the local branch unit overwrites the global model parameters to the corresponding parameter positions of the local fault location model, and optionally uses the 10 most recently collected fault samples for 1-2 rounds of local fine-tuning; the fault features collected in real time are input into the updated model in time step order, the model outputs the fault location prediction result, and the fault location is completed, thereby improving the model's adaptability to fault features of different branch units and ensuring the real-time performance and accuracy of fault location.

[0169] This embodiment sends data sharing requests to other branch units in a simplified topology network, enabling them to send local model parameters based on these requests. These local model parameters are obtained by each branch unit after training a local fault location model using local fault characteristics and historical fault sample data. The local fault location model is pre-trained using historical fault sample data to obtain initial model parameters. These initial model parameters are then aggregated with the local model parameters sent by other branch units based on their historical fault location reliability and request response speed to obtain global model parameters. The local fault location model is then updated using these global model parameters to obtain the final local fault location model. Fault prediction is then performed based on this final local fault location model. This improves the efficiency of fault location in the submarine DC power supply system, reduces fault location time, effectively mines common fault characteristics among multiple branch units, enhances the generalization ability of the fault location model, reduces the computational cost of independent local model training, and optimizes model performance through joint learning.

[0170] Furthermore, this embodiment of the invention also proposes a computer-readable storage medium storing a fault location program for a ring-shaped submarine DC power supply system. When the ring-shaped submarine DC power supply system fault location program is executed by a processor, it implements the steps of the fault location method for a ring-shaped submarine DC power supply system as described above.

[0171] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0172] The aforementioned computer-readable storage medium may be included in the fault location device of the ring-shaped submarine DC power supply system; or it may exist independently and not be assembled into the fault location device of the ring-shaped submarine DC power supply system.

[0173] Furthermore, this invention also proposes a computer program product, including a fault location program for a ring-shaped submarine DC power supply system. When the ring-shaped submarine DC power supply system fault location program is executed by a processor, it implements the steps of the fault location method for a ring-shaped submarine DC power supply system as described above.

[0174] The specific implementation of the computer program product of the present invention is basically the same as the various embodiments of the above-described fault location method for the ring-shaped submarine DC power supply system, and will not be repeated here.

[0175] Reference Figure 8 , Figure 8 This is a structural block diagram of the first embodiment of the fault location system for the ring-shaped submarine DC power supply system of the present invention.

[0176] like Figure 8 As shown, the fault location system for the ring-shaped submarine DC power supply system proposed in this embodiment of the invention includes:

[0177] The system loop-breaking module 10 is used to break the ring-shaped submarine DC power supply system into a simplified topology network, and set measurement points and local fault location models corresponding to the measurement points based on the branch units in the simplified topology network.

[0178] Feature analysis module 20 is used to perform feature analysis on the local fault current data collected at the branch unit to obtain fault features;

[0179] The fault analysis module 30 is used to input the fault features into the local fault location model in the order of time steps to perform fault prediction and obtain the fault distance between the fault location and the branch unit. The local fault location model is built based on a pre-trained LSTM network.

[0180] The fault location module 40 is used to determine the faulty line and the fault location based on the fault distance, and send the faulty line and the fault location to the shore base station to complete the fault location.

[0181] Furthermore, the system loop-breaking module 10 is also used to break the loop in the ring-shaped submarine DC power supply system to obtain a radial network; and to merge the lines at both ends of the nodes in the radial network that meet the line merging condition to obtain a simplified topology network, thereby realizing topology simplification of the network. The line merging condition is that the number of connected nodes is 2.

[0182] Furthermore, the system loop-breaking module 10 is also used to determine the longest connection path between two shore base stations of the ring-shaped submarine DC power supply system, and to reconstruct the ring-shaped submarine DC power supply system based on the longest connection path to obtain a reconstructed network, which includes a multi-ring network and a single-ring network; to break the loop in the multi-ring network by using the intersection points on the non-longest connected paths in the multi-ring network as decomposition points; to determine the node positions of the longest connected path in the single-ring network, to calculate the shortest path between the node positions of the longest connected path and the two shore base stations, to determine the decomposition points based on the nodes in the shortest path, and to break the loop in the single-ring network based on the decomposition points; and to obtain a radial network based on the loop-breaking results of the multi-ring network and the single-ring network.

[0183] Furthermore, the fault features include instantaneous slope and instantaneous curvature; the feature analysis module 20 is also used to determine the feature calculation start time based on the fault initiation time, referring to the following formula:

[0184]

[0185]

[0186] in, This represents the inductance per unit length of the submarine cable. Capacitance per unit length of submarine cable. The time when the fault begins. This is the empirical offset. The distance to the fault. The speed at which current propagates in a submarine cable. Calculate the start time for the feature;

[0187] Based on the aforementioned feature calculation start time, feature analysis is performed on the local fault current data collected at the branch unit to obtain the instantaneous slope and instantaneous curvature of the fault current, referring to the following formula:

[0188]

[0189]

[0190]

[0191]

[0192]

[0193] in, The instantaneous slope of the fault current. Indicates the start time of feature calculation Next fault current data, express The next fault current data, Indicates the sampling time interval. and Indicates the fault current difference. The instantaneous curvature of the fault current. Indicates the sampling frequency.

[0194] This embodiment deconstructs the ring-shaped submarine DC power supply system into a simplified topology network. Based on the branch units in this simplified topology network, measurement points and corresponding local fault location models are set up. Feature analysis is performed on the local fault current data collected at the branch units to obtain fault features. These fault features are then input into the local fault location model in time step order for fault prediction, obtaining the fault distance between the fault location and the branch unit. The local fault location model is constructed based on a pre-trained LSTM network. Based on the fault distance, it determines the faulty line and fault location, and sends the faulty line and fault location to the shore base station to complete fault location. Because this embodiment simplifies the complex ring-shaped submarine DC power supply system topology, it reduces the number of lines and measurement points required. By selecting fault current features unaffected by changes in the number of LLC branches, the regularity of the features is improved, thereby enhancing fault location accuracy. This effectively adapts to scenarios with few measurement points in the ring-shaped submarine DC power supply system.

[0195] The fault location system for a ring-shaped subsea DC power supply system provided in this application employs the fault location method for a ring-shaped subsea DC power supply system described in the above embodiments, and can solve the technical problem of fault location in a ring-shaped subsea DC power supply system. Compared with the prior art, the beneficial effects of the fault location system for a ring-shaped subsea DC power supply system provided in this application are the same as those of the fault location method for a ring-shaped subsea DC power supply system provided in the above embodiments, and other technical features of the fault location system for a ring-shaped subsea DC power supply system are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0196] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solution of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0197] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0198] In addition, for technical details not described in detail in this embodiment, please refer to the fault location method of the annular submarine DC power supply system provided in any embodiment of the present invention, which will not be repeated here.

[0199] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0200] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0201] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0202] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A fault location method for a ring-shaped submarine DC power supply system, characterized in that, The fault location method for the ring-shaped submarine DC power supply system includes: The ring-shaped submarine DC power supply system is de-looped into a simplified topology network, and measurement points and local fault location models corresponding to the measurement points are set based on the branch units in the simplified topology network. Feature analysis is performed on the local fault current data collected at the branch unit to obtain fault characteristics; The fault features are input into the local fault location model in time step order to predict the fault location and obtain the fault distance between the fault location and the branch unit. The local fault location model is built based on a pre-trained LSTM network. The faulty line and fault location are determined based on the fault distance, and the faulty line and fault location are sent to the shore base station to complete the fault location. The process of de-looping the ring-shaped submarine DC power supply system into a simplified topology network includes: By de-looping the ring-shaped submarine DC power supply system, a radial network is obtained. In the radial network, the lines at both ends of the nodes that meet the line merging condition are merged to obtain a simplified topology network, thereby simplifying the network topology. The line merging condition is that the number of connected nodes is 2. The process of de-looping the ring-shaped submarine DC power supply system to obtain a radial network includes: The longest connection path between two shore base stations of a ring-shaped submarine DC power supply system is determined, and the ring-shaped submarine DC power supply system is reconstructed based on the longest connection path to obtain a reconstructed network, which includes a multi-ring network and a single-ring network. The multi-ring network is de-looped by using the intersection points on the non-longest connected paths in the multi-ring network as decomposition points; Determine the node positions of the longest connected path in a single-ring network, calculate the shortest path between the node positions of the longest connected path and the two shore base stations, determine decomposition points based on the nodes in the shortest path, and de-loop the single-ring network based on the decomposition points. A radial network is obtained based on the de-looping results of multi-ring networks and single-ring networks.

2. The fault location method for a ring-shaped submarine DC power supply system as described in claim 1, characterized in that, The fault characteristics include instantaneous slope and instantaneous curvature; the feature analysis of the local fault current data collected at the branch unit to obtain fault characteristics includes: The start time for characteristic calculation is determined based on the fault initiation time, referring to the following formula: in, This represents the inductance per unit length of the submarine cable. Capacitance per unit length of submarine cable. The time when the fault begins. This is the empirical offset. The distance to the fault. The speed at which current propagates in a submarine cable. Calculate the start time for the feature; Based on the aforementioned feature calculation start time, feature analysis is performed on the local fault current data collected at the branch unit to obtain the instantaneous slope and instantaneous curvature of the fault current, referring to the following formula: in, The instantaneous slope of the fault current. Indicates the start time of feature calculation Next fault current data, express The next fault current data, Indicates the sampling time interval. and Indicates the fault current difference. The instantaneous curvature of the fault current. Indicates the sampling frequency.

3. The fault location method for a ring-shaped submarine DC power supply system as described in claim 1, characterized in that, The calculation formulas for each unit in the local fault location model are as follows: in, , and These represent the output value, weight coefficient, and bias of the forget gate, respectively. Indicates the first Input volume at any given time Indicates the first The state value of the hidden layer at any given time. , and These represent the output value, weighting coefficient, and bias of the input gate, respectively. , and These represent the output value, weighting coefficient, and bias of the output gate, respectively. , and These represent the output value, weight coefficient, and bias, respectively, during the feature extraction process. and These represent the activation function and the hyperbolic tangent function, respectively. Indicates the input gate and the first The weighting coefficients associated with the input values ​​at different times. Indicates the output gate and the first The weighting coefficients associated with the input values ​​at different times. Indicates the feature extraction output and the first The weighting coefficients associated with the input values ​​at different times. Indicates the output of the forget gate and the first... Weighting coefficients associated with input values ​​at different times; The activation function and the hyperbolic tangent function are respectively: in, Input data; The local fault location model is configured to determine the cell state value at each time step based on the output value of each unit, referring to the following formula: in, and Representing time respectively and time Cell state values, This indicates element-wise multiplication; The local fault location model is further configured to determine the hidden layer state value at each time step based on the cell state value and the output value during the feature extraction process, and to convert the hidden layer state value into the fault location at each time step through the output mapping module. The hidden layer state value is calculated with reference to the following formula: in, Indicates time The hidden layer state values.

4. The fault location method for a ring-shaped submarine DC power supply system as described in claim 1, characterized in that, Before inputting the fault features into the local fault location model in time step order for fault prediction, the method further includes: Send a data sharing request to other branch units in the simplified topology network so that the other branch units can send local model parameters based on the data sharing request. The local model parameters are obtained by each branch unit after training a local fault location model based on local fault characteristics and historical fault sample data. The local fault location model is pre-trained based on historical fault sample data to obtain initial model parameters. Based on the historical fault location reliability and request response speed of other branch units, the initial model parameters are aggregated with the local model parameters sent by the other branch units to obtain global model parameters; The local fault location model is updated based on the global model parameters to obtain the final local fault location model, and the step of inputting the fault features into the local fault location model in time step order for fault prediction is executed.

5. A fault location system for a ring-shaped submarine DC power supply system, characterized in that, The fault location system for the ring-shaped submarine DC power supply system includes: The system loop-breaking module is used to break the ring-shaped submarine DC power supply system into a simplified topology network, and set measurement points and local fault location models corresponding to the measurement points based on the branch units in the simplified topology network. The feature analysis module is used to perform feature analysis on the local fault current data collected at the branch unit to obtain fault characteristics; The fault analysis module is used to input the fault features into the local fault location model in the order of time steps to predict the fault and obtain the fault distance between the fault location and the branch unit. The local fault location model is built based on a pre-trained LSTM network. The fault location module is used to determine the faulty line and the fault location based on the fault distance, and send the faulty line and the fault location to the shore base station to complete the fault location. The system loop-breaking module is also used to break the loop in the ring-shaped submarine DC power supply system to obtain a radial network; and to merge the lines at both ends of the nodes in the radial network that meet the line merging condition to obtain a simplified topology network, thereby realizing topology simplification of the network. The line merging condition is that the number of connected nodes is 2. The system loop-breaking module is further configured to determine the longest connection path between two shore base stations of the ring-shaped submarine DC power supply system, and reconstruct the ring-shaped submarine DC power supply system based on the longest connection path to obtain a reconstructed network, which includes a multi-ring network and a single-ring network; the multi-ring network is decomposed using the intersection points on the non-longest connected paths in the multi-ring network as decomposition points; the node positions of the longest connected path in the single-ring network are determined, the shortest path between the node positions of the longest connected path and the two shore base stations is calculated, decomposition points are determined based on the nodes in the shortest path, and the single-ring network is decomposed based on the decomposition points; a radial network is obtained based on the loop-breaking results of the multi-ring network and the single-ring network.

6. The fault location system for a ring-shaped submarine DC power supply system as described in claim 5, characterized in that, The fault features include instantaneous slope and instantaneous curvature; the feature analysis module is also used to determine the feature calculation start time based on the fault start time, referring to the following formula: in, This represents the inductance per unit length of the submarine cable. Capacitance per unit length of submarine cable. The time when the fault begins. This is the empirical offset. The distance to the fault. The speed at which current propagates in a submarine cable. Calculate the start time for the feature; Based on the aforementioned feature calculation start time, feature analysis is performed on the local fault current data collected at the branch unit to obtain the instantaneous slope and instantaneous curvature of the fault current, referring to the following formula: in, The instantaneous slope of the fault current. Indicates the start time of feature calculation Next fault current data, express The next fault current data, Indicates the sampling time interval. and Indicates the fault current difference. The instantaneous curvature of the fault current. Indicates the sampling frequency.

Images