An adaptive relay protection method and system for a power distribution network containing distributed power sources

Abstract

This invention discloses an adaptive relay protection method and system for distribution networks with distributed power sources, belonging to the field of distribution network relay protection technology. The method involves: acquiring three-phase voltage and current signals at both ends of the protected line in the distribution network; extracting the positive-sequence voltage and current components to calculate the positive-sequence unit admittance; interacting with the positive-sequence unit admittance at both ends to calculate the change in positive-sequence unit admittance, the difference in positive-sequence unit admittance change, and the phase angle difference of the positive-sequence unit admittance change, thus obtaining the fault identification result; constructing a distribution network topology map based on the fault identification result and the distribution network operating status; using a graph attention network to calculate the coupling relationship weights and weighted aggregation to obtain comprehensive state characteristics; and outputting the operating current setting value and operating time setting value through a mapping function; and outputting a trip command when preset conditions are met based on the above parameters and real-time measured current. This invention improves the fault identification accuracy of the protected line in the distribution network, adapts to dynamic changes in the distribution network, and ensures the safe and stable operation of the distribution network.

Description

Technical Field

[0001] This invention relates to the field of power distribution network relay protection technology, and in particular to an adaptive relay protection method and system for power distribution networks containing distributed generation sources. Background Technology

[0002] With the large-scale integration of distributed generation (DG) into distribution networks, traditional distribution systems dominated by single-source radial structures are gradually evolving into complex network structures with multiple sources and multiple power flow directions. Distributed power sources such as photovoltaic (PV), wind power, and energy storage devices are characterized by dispersed access locations and strong output fluctuations, resulting in highly uncertain and dynamically changing operation modes in distribution networks. Against this backdrop, existing relay protection methods based on fixed topology and single power flow direction assumptions face severe challenges, such as complex fault current distribution, significant short-circuit capacity variations, and difficulties in defining protection ranges, seriously affecting the safe and stable operation of distribution networks. Therefore, researching novel relay protection methods suitable for high-proportion distributed generation integration has become an important topic in the power system field.

[0003] Existing distribution network relay protection technologies mainly include overcurrent protection, directional protection, and differential protection. Overcurrent protection relies on the fault current amplitude for judgment, and is prone to decreased sensitivity, false tripping, or failure to trip under multi-source power supply conditions. While directional protection can identify the fault direction to some extent, its reliability is significantly reduced after the integration of distributed power sources due to frequent changes in power flow direction. Differential protection, although possessing good selectivity and sensitivity, typically relies on high-quality communication channels, and its large-scale deployment in distribution networks faces limitations such as high cost and complex implementation. Furthermore, existing protection settings mostly employ offline calculation methods, setting parameters based on typical operating modes, which is difficult to adapt to the actual operating environment of distributed power source output fluctuations and frequent network topology changes, leading to decreased protection coordination performance and even protection mismatch problems.

[0004] In recent years, with the development of communication and artificial intelligence technologies, the integration of multi-source information fusion and intelligent algorithms into the field of relay protection has become a new research direction. For example, improving fault identification accuracy through two- or multi-terminal information interaction, and using machine learning methods to achieve adaptive setting of protection parameters. However, existing research mostly focuses on improving single aspects, lacking systematic modeling of the overall distribution network topology and the coupling characteristics between various protection devices, making it difficult to achieve globally optimal protection coordination in complex network environments. Meanwhile, how to construct a protection model that reflects real-time topology changes and achieve fast and accurate fault identification and adaptive setting based on this remains a key technical problem to be solved. Therefore, it is necessary to propose an adaptive relay protection technology that integrates two-terminal information criteria and graph structure learning methods to improve the protection performance and operational reliability of distribution networks with a high proportion of distributed power sources. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes an adaptive relay protection method and system for distribution networks containing distributed power sources.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] An adaptive relay protection method for distribution networks containing distributed generation sources includes:

[0008] The three-phase voltage and three-phase current signals at both ends of the protected line in the distribution network are synchronously acquired using a unified time reference, and the positive sequence voltage component and positive sequence current component are extracted to calculate the positive sequence unit admittance.

[0009] After transmitting the positive sequence unit admittances at both ends of the protected line in the distribution network to each other and synchronizing and aligning them in time, the change in positive sequence unit admittance is calculated. Based on the change in positive sequence unit admittance, the difference in positive sequence unit admittance change and the phase angle difference in positive sequence unit admittance change are obtained, and the fault judgment result is obtained according to the preset fault judgment rules.

[0010] A distribution network topology map is constructed based on the fault identification results and the distribution network operation status. Based on the distribution network topology map, a graph attention network is used to calculate the coupling relationship weights. The weighted aggregation based on the coupling relationship weights is used to obtain the comprehensive state features. The action current setting value and action time setting value are output through a pre-trained mapping function.

[0011] Based on the fault diagnosis result, the operating current setting value, and the operating time setting value, a trip command is output when the real-time measured current meets the preset conditions.

[0012] As a preferred embodiment of the present invention, the step of synchronously acquiring the three-phase voltage and three-phase current signals at both ends of the protected line of the distribution network using a unified time reference, extracting the positive-sequence voltage component and the positive-sequence current component, and calculating the positive-sequence unit admittance includes:

[0013] The acquired three-phase voltage and three-phase current signals are converted into positive-sequence voltage and positive-sequence current components using the symmetrical component method; among them, the positive-sequence voltage component... The expression is:

[0014]

[0015] In the formula, for Phase voltage, for Phase voltage, for Phase voltage, It is the rotation factor;

[0016] Positive sequence current component The expression is:

[0017]

[0018] In the formula, for Phase current signal, for Phase current signal, for Phase current signal;

[0019] Based on the positive-sequence voltage component and the positive-sequence current component, the expression for calculating the positive-sequence unit admittance is as follows:

[0020]

[0021] In the formula, It is the positive-order unit admittance.

[0022] As a preferred embodiment of the present invention, the step of transmitting the positive-sequence unit admittances at both ends of the protected line of the distribution network to each other, performing time synchronization and alignment, calculating the change in positive-sequence unit admittance, and obtaining the difference in positive-sequence unit admittance change and the phase angle difference in positive-sequence unit admittance change based on the change in positive-sequence unit admittance includes:

[0023] The positive-sequence unit admittances at both ends of the protected line in the distribution network are transmitted to each other through a communication channel. Time synchronization and alignment processing is performed on the positive-sequence unit admittances at both ends of the protected line to obtain the corresponding first-end and last-end positive-sequence unit admittances at the same moment, and the change in positive-sequence unit admittance is calculated. The change in positive-sequence unit admittance includes the change in first-end and last-end positive-sequence unit admittance. Specifically, the change in first-end positive-sequence unit admittance is calculated... The expression is:

[0024]

[0025] In the formula, For a moment The first positive-order unit admittance below. The time interval between adjacent sampling times;

[0026] Calculate the change in positive-sequence unit admittance at the end. The expression is:

[0027]

[0028] In the formula, For a moment The terminal positive-sequence unit admittance is below;

[0029] Calculate the difference in positive-sequence unit admittance based on the changes in the first-order and last-order positive-sequence unit admittances. The expression is:

[0030]

[0031] Calculate the phase angle difference of the positive-sequence unit admittance change. The expression is:

[0032]

[0033] In the formula, The phase angle is the change in the first positive sequence unit admittance. The phase angle represents the change in positive-sequence unit admittance at the terminal.

[0034] As a preferred embodiment of the present invention, obtaining the fault discrimination result according to the preset fault discrimination rule includes:

[0035] The preset fault detection rule is as follows:

[0036]

[0037] In the formula, The threshold for positive-order unit admittance difference. The phase angle difference threshold, For logical AND relation;

[0038] The fault identification results include: faults within the zone or faults outside the zone;

[0039] If the preset fault discrimination rules are met, the fault is determined to be an internal fault; otherwise, it is an external fault.

[0040] As a preferred embodiment of the present invention, the power grid operating status includes the power grid switch status, the distributed power source access status, and the line connection relationship;

[0041] The power distribution network topology diagram for:

[0042]

[0043] In the formula, For a set of nodes, It is an edge set;

[0044] For each node in the node set, define a node feature vector, expressed as follows:

[0045]

[0046] In the formula, For nodes The node feature vectors, For nodes positive-order unit admittance, For nodes The positive-order unit admittance change. For nodes Fault discrimination state variables;

[0047] When a node is located in a faulty area within the zone, the fault discrimination status variable is set to 1; otherwise, it is set to 0.

[0048] As a preferred embodiment of the present invention, the step of calculating the coupling relationship weights using a graph attention network includes:

[0049] The correlation calculation function of graph attention network is used to calculate the coupling relationship weight between adjacent nodes in the distribution network topology graph. The expression is as follows:

[0050]

[0051]

[0052] In the formula, For nodes For adjacent nodes The normalized coupling weights, For nodes For adjacent nodes The unnormalized attention coefficient, For nodes The set of adjacent nodes, The relevance calculation function for graph attention networks. For nodes Node feature vectors;

[0053] Based on the coupling relationship weights, the nodes are... All neighboring nodes The corresponding node feature vectors are weighted and aggregated to obtain the node. The comprehensive state characteristics are expressed as follows:

[0054]

[0055] In the formula, For nodes The comprehensive state characteristics.

[0056] As a preferred embodiment of the present invention, the step of outputting the action current setting value and the action time setting value through the pre-trained mapping function includes:

[0057] The pre-trained mapping functions are mapping functions from comprehensive state features to action current settings and mapping functions from comprehensive state features to action time settings, which are trained from historical operating data or simulation data.

[0058] The expression for calculating the operating current setting value is as follows:

[0059]

[0060] In the formula, For the first The operating current setting value of each protection device. This is a mapping function from the comprehensive state characteristics to the operating current setting value;

[0061] The expression for calculating the motion time setpoint is:

[0062]

[0063] In the formula, For the first The setpoint for the operating time of each protection device. It is a mapping function from the comprehensive state characteristics to the action time tuning value.

[0064] As a preferred embodiment of the present invention, the step of outputting a trip command when the real-time measured current meets preset conditions based on the fault diagnosis result, the operating current setting value, and the operating time setting value includes:

[0065] When the real-time measured current detected by the protection device is greater than the operating current setting value, the fault identification result is an intra-zone fault, and the duration of the real-time measured current being greater than the operating current setting value exceeds the operating time setting value, a trip command is output, and the protected line is isolated by disconnecting it through the circuit breaker.

[0066] As a preferred embodiment of the present invention, the time difference coordination constraint is applied to the operating time setting value of each protection device according to the operating time setting value of adjacent protection devices. Based on the distribution network topology and the distribution characteristics of fault electrical quantities, the electrical distance between each protection device and the fault point is calculated. The protection device with the smallest electrical distance is used as the main protection with priority operation, and the adjacent protection devices are used as backup protection with delayed operation.

[0067] The protection action information and related electrical quantities are uploaded to the distribution network dispatch system or distribution network monitoring system for distribution network operation analysis and iterative optimization of graph attention network; the related electrical quantities include three-phase voltage, three-phase current signal, positive sequence voltage component, positive sequence current component, positive sequence unit admittance, positive sequence unit admittance change, positive sequence unit admittance change difference, and positive sequence unit admittance change phase angle difference.

[0068] An adaptive relay protection system for a distribution network containing distributed power sources, characterized in that,

[0069] The acquisition and extraction module synchronously acquires the three-phase voltage and three-phase current signals at both ends of the protected line of the distribution network using a unified time reference, and extracts the positive sequence voltage component and positive sequence current component to calculate the positive sequence unit admittance.

[0070] The fault identification module transmits the positive sequence unit admittance at both ends of the protected line of the distribution network to each other, performs time synchronization and alignment, calculates the change in positive sequence unit admittance, obtains the difference in positive sequence unit admittance change and the phase angle difference in positive sequence unit admittance change based on the change in positive sequence unit admittance, and obtains the fault identification result according to the preset fault identification rules.

[0071] The state feature module constructs a distribution network topology map based on the fault identification results and the distribution network operation status. Based on the distribution network topology map, a graph attention network is used to calculate the coupling relationship weights. The weighted aggregation based on the coupling relationship weights is used to obtain the comprehensive state features. The action current setting value and action time setting value are output through a pre-trained mapping function.

[0072] The output command module outputs a trip command when the real-time measured current meets the preset conditions, based on the fault diagnosis result, the operating current setting value, and the operating time setting value.

[0073] The beneficial effects of this invention are as follows: This invention effectively solves the technical problems of poor adaptability of relay protection in distribution networks containing distributed power sources, low fault discrimination accuracy, and uncoordinated action coordination, ensuring the safe and stable operation of the distribution network. By synchronously collecting the three-phase voltage and three-phase current signals at both ends of the protected line in the distribution network, extracting the positive sequence voltage component and the positive sequence current component, and calculating the positive sequence unit admittance, the invention combines the change in the positive sequence unit admittance at both ends, the difference in the change in the positive sequence unit admittance, and the phase angle difference of the change in the positive sequence unit admittance to identify faults, significantly improving the accuracy of fault identification inside and outside the protected line in the distribution network, and avoiding false fault identification and missed fault identification. Based on the fault discrimination results and the operating status of the distribution network, a distribution network topology map is constructed. A graph attention network is used to calculate the coupling relationship weights and weighted aggregation to obtain comprehensive state features. The operating current setting value and operating time setting value are output through a pre-trained mapping function, realizing adaptive setting of protection parameters to adapt to the dynamic changes in the distribution network topology and operating status. By combining the fault identification results, the operating current setting value, the operating time setting value, and the preset conditions of real-time measured current, the trip command is output to ensure accurate disconnection when the protected line of the distribution network is faulty, reduce the power outage range, improve the selectivity and reliability of relay protection, and reduce the risk of protection maloperation and failure to operate. Attached Figure Description

[0074] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart of the method in an embodiment of the present invention; Figure 2 This is a system structure diagram in an embodiment of the present invention. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0076] like Figure 1 As shown, this is an embodiment of the present invention, which provides an adaptive relay protection method for a distribution network containing distributed generation sources, comprising:

[0077] S1. Using a unified time reference, the three-phase voltage and three-phase current signals at both ends of the protected line in the distribution network are synchronously acquired, and the positive sequence voltage component and positive sequence current component are extracted to calculate the positive sequence unit admittance.

[0078] In this embodiment, the collected three-phase voltage and three-phase current signals are converted into positive-sequence voltage and positive-sequence current components using the symmetrical component method.

[0079] Specifically, an independent measurement unit is configured at both the beginning (M end) and the end (N end) of the protected line in the distribution network. Each measurement unit includes one three-phase voltage transformer and one three-phase current transformer. The three-phase voltage transformer is used to collect the three-phase voltage at the corresponding protected line end, and the three-phase current transformer is used to collect the three-phase current signal at the corresponding protected line end. All voltage and current transformers have an accuracy class of 0.2, meeting the measurement accuracy requirements of distribution network relay protection. A unified time reference is provided for the measurement units at both ends of the protected line using the Global Positioning System (GPS) or the IEEE 1588 precise time protocol, with synchronization accuracy controlled within ±1μs. The sampling clocks of the measurement units at both ends of the protected line are strictly synchronized with this unified time reference, ensuring that the sampling times at the beginning and end are completely consistent, providing a unified time basis for subsequent comparative analysis of electrical quantities at both ends. The acquired three-phase voltage and three-phase current signals are first filtered by a low-pass filter with a cutoff frequency of 1kHz to remove electromagnetic interference and high-order harmonic components, and suppress the influence of noise on the subsequent positive sequence component calculation. The filtered signals are then transmitted to the central processing unit of the protection device for further calculation.

[0080] Among them, the positive sequence voltage component The expression is:

[0081]

[0082]

[0083] In the formula, for Phase voltage, for Phase voltage, for Phase voltage, For rotation factor, It is the standard imaginary unit for power systems.

[0084] Positive sequence current component The expression is:

[0085]

[0086] In the formula, for Phase current signal, for Phase current signal, for Phase current signal.

[0087] Specifically, the symmetrical component method is used to convert the acquired three-phase voltage and three-phase current signals into positive-sequence current and positive-sequence voltage components. This conversion can extract the positive-sequence voltage and positive-sequence current components reflecting the normal operating characteristics of the system from the asymmetrical three-phase voltage and three-phase current signals, thereby reducing the influence of fault type and imbalance factors on subsequent criterion calculations. The calculation of the above positive-sequence voltage and positive-sequence current components needs to be performed independently at the beginning and end of the protected line, respectively, to obtain: the positive-sequence voltage component at the beginning, the positive-sequence current component at the beginning, the positive-sequence voltage component at the end, and the positive-sequence current component at the end.

[0088] Based on the positive-sequence voltage and current components, the expression for calculating the positive-sequence unit admittance for any end of the protected line is as follows:

[0089]

[0090] In the formula, It is a positive-sequence unit admittance, reflecting the equivalent electrical characteristics of the line under fault conditions.

[0091] Specifically, when calculating the first end of the protected line, the positive-sequence unit admittance at the first end is obtained; when calculating the last end of the protected line, the positive-sequence unit admittance at the last end is obtained. When a fault occurs within the protected line's coverage area, both ends of the protected line are directly affected by the fault, and the changing trend of the positive-sequence unit admittance is basically the same. When a fault occurs outside the coverage area, the changes in electrical quantities at both ends of the protected line are significantly different, and the amplitude and phase angle of the positive-sequence unit admittance will exhibit different changing characteristics. Therefore, this positive-sequence unit admittance provides a basic criterion for subsequent comparison of information at both ends and fault identification.

[0092] S2. The positive sequence unit admittance at both ends of the protected line of the distribution network is transmitted to each other and synchronized and aligned in time. The change in positive sequence unit admittance is calculated. Based on the change in positive sequence unit admittance, the difference in positive sequence unit admittance change and the phase angle difference in positive sequence unit admittance change are obtained. The fault judgment result is obtained according to the preset fault judgment rules.

[0093] In this embodiment, the positive-sequence unit admittances at both ends of the protected line of the distribution network are transmitted to each other through a communication channel, and the positive-sequence unit admittances at both ends of the protected line of the distribution network are time-synchronized and aligned to obtain the corresponding first-end positive-sequence unit admittance and last-end positive-sequence unit admittance at the same time, and the change in positive-sequence unit admittance is calculated; the change in positive-sequence unit admittance includes the change in first-end positive-sequence unit admittance and the change in last-end positive-sequence unit admittance.

[0094] Specifically, a dedicated communication channel for dual-end data interaction is adopted using fiber optic Ethernet or a 5G industrial private network, with a communication bandwidth of no less than 10Mbps, an end-to-end one-way transmission latency of no more than 10ms, and a packet loss rate of no more than 0.01%, meeting the real-time and reliability requirements of distribution network relay protection. The head-end protection device and the end-end protection device of the protected line in the distribution network establish a point-to-point encrypted connection through this communication channel to achieve bidirectional real-time data transmission. The head-end protection device of the protected line encapsulates the head-end positive sequence unit admittance calculated in the current sampling period, along with the corresponding microsecond-level timestamp of the sampling time, device number, and data check code, into a standard data frame and sends it to the end-end protection device of the protected line. Simultaneously, the end-end protection device encapsulates the end-end positive sequence unit admittance calculated in the current sampling period, along with the corresponding timestamp, device number, and data check code, into a data frame of the same format and sends it to the head-end protection device of the protected line. The data frame adopts a CRC32 check mechanism; if the receiving end fails the check, it immediately requests a retransmission of the previous frame.

[0095] To eliminate the impact of communication transmission delay on the synchronization of data at both ends of the protected lines in the distribution network, a bidirectional delay measurement method is adopted for real-time measurement and dynamic compensation of communication delay. The protection device at the head end of the protected line sends a delay measurement frame to the protection device at the tail end. Upon receiving the delay measurement frame, the tail end protection device immediately sends a delay response frame to the head end protection device. The delay response frame contains the times at which the tail end protection device received the delay measurement frame and sent the delay response frame. After receiving the delay response frame, the head end protection device calculates the one-way communication delay and compensates for the delay of the received positive-sequence unit admittance data based on this one-way communication delay, ensuring that the compensated data corresponds to its actual acquisition time. Delay measurement and updates are performed at fixed intervals. When the measured one-way communication delay shows significant fluctuations, an additional delay measurement is immediately triggered to ensure the accuracy of delay compensation.

[0096] Based on the unified time reference established in step S1, the protection devices at the beginning and end of the protected lines in the distribution network align the positive sequence unit admittance after time delay compensation to the same sampling time, ensuring the physical consistency and accuracy of subsequent characteristic quantity calculations.

[0097] A linear interpolation synchronization method is used to achieve accurate alignment of data at both ends of the protected lines in the distribution network. Using the sampling time of the local protection device as a reference, the target time to be aligned is determined. In the positive-sequence unit admittance sequence of the opposite end after time delay compensation, the two closest sampling times before and after the target time, and the corresponding positive-sequence unit admittances for these two sampling times, are found. The positive-sequence unit admittance of the opposite end corresponding to the target time is calculated using linear interpolation. This process is repeated to obtain the aligned positive-sequence unit admittances at both ends for all target times. After this step, the first-end and last-end positive-sequence unit admittances at the same time are obtained.

[0098] Among them, the change in the first-order positive-sequence unit admittance is calculated. The expression is:

[0099]

[0100] In the formula, For a moment The first positive-order unit admittance below. This represents the time interval between adjacent sampling times.

[0101] Calculate the change in positive-sequence unit admittance at the end. The expression is:

[0102]

[0103] In the formula, For a moment The terminal positive sequence unit admittance.

[0104] Specifically, the change in positive sequence unit admittance at the beginning or end is used to reflect the difference in the response of electrical quantities at both ends of the protected line when a fault occurs.

[0105] Calculate the difference in positive-sequence unit admittance based on the changes in the first-order and last-order positive-sequence unit admittances. The expression is:

[0106]

[0107] Specifically, the difference in positive-sequence unit admittance reflects the consistency of the response to fault disturbances at both ends of the protected line: the smaller the difference, the more consistent the changes at both ends of the protected line.

[0108] Calculate the phase angle difference of the positive-sequence unit admittance change. The expression is:

[0109]

[0110] In the formula, The phase angle is the change in the first positive sequence unit admittance. The phase angle represents the change in positive-sequence unit admittance at the terminal.

[0111] Specifically, the phase angle difference of the positive sequence unit admittance change reflects the consistency of the response direction of the fault disturbance at both ends of the protected line.

[0112] In this embodiment, the preset fault discrimination rule is:

[0113]

[0114] In the formula, The positive-sequence unit admittance difference threshold is set based on line parameters and operating experience, and is typically between 0.03 seconds and 0.08 seconds. The phase angle difference threshold is usually set to a small angle, such as 10 degrees or 15 degrees. This is a logical AND relation, meaning that both conditions must be met simultaneously.

[0115] The fault identification results include: faults within the zone or faults outside the zone;

[0116] If the preset fault discrimination rules are met, the fault is determined to be an internal fault; otherwise, it is an external fault.

[0117] Specifically, when a fault occurs within the protected area, both ends of the protected line are directly affected by the fault, and the positive-sequence unit admittance changes in a basically consistent manner, resulting in small differences in the magnitude of the changes and phase angle differences. When a fault occurs outside the protected area, the fault is located outside the protected area, and there are significant differences in the changes in electrical quantities at both ends of the protected line, causing the magnitude of the changes or phase angle differences to exceed the threshold.

[0118] To suppress misjudgments caused by transient electromagnetic disturbances, measurement noise, and communication errors, a 10ms anti-jitter delay is set for the fault identification results. Specifically, the final fault identification result is only output when the fault identification results at all sampling times within a consecutive 10ms are all faults within the zone; if the fault identification result is a fault outside the zone at any sampling time within the anti-jitter delay, the anti-jitter timer is immediately reset to zero and restarted.

[0119] S3. Construct a distribution network topology map based on the fault identification results and the distribution network operation status; based on the distribution network topology map, use a graph attention network to calculate the coupling relationship weights, perform weighted aggregation based on the coupling relationship weights to obtain comprehensive state features, and output the action current setting value and action time setting value through a pre-trained mapping function.

[0120] In this embodiment, the power grid operating status includes the power grid switch status, the distributed power source access status, and the line connection relationship.

[0121] Specifically, three types of operating status data are synchronously collected from the Supervisory Control and Data Acquisition (SCADA) system of the power distribution network, the distributed power source monitoring system, and the measurement units of each protection device, with the data acquisition frequency consistent with the sampling frequency of the protection devices. The sampling frequency of the protection devices is set to 2.56kHz, meaning 512 sampling points are collected per power frequency cycle (20ms). The three types of operating status data are as follows:

[0122] Power grid switch status: including the opening and closing status and operation timestamps of each circuit breaker and disconnector, used to determine the continuity of the line;

[0123] Distributed power source access status: including the grid-connected or off-grid status of each distributed power source, the current active or reactive power output level, and the access node number;

[0124] Line connection relationship: Based on the static topology database of the distribution network and combined with the real-time switch status, the electrical connection topology of the lines at the current moment is generated.

[0125] The distribution network is abstracted as an undirected graph structure, specifically a distribution network topology graph; the distribution network topology graph for:

[0126]

[0127] In the formula, For a set of nodes, It is an edge set.

[0128] In the node set, each node corresponds to the installation location of a protection device or a bus node in the distribution network. The node number corresponds one-to-one with the unique identifier of the protection device in the distribution network. In the edge set, each edge corresponds to a transmission line in the distribution network. An edge is added between two nodes if and only if there is a direct electrical connection between them and the corresponding line is in a conducting state.

[0129] For each node in the node set, a three-dimensional node feature vector is defined, expressed as:

[0130]

[0131] In the formula, For nodes The node feature vectors, For nodes positive-order unit admittance, For nodes The positive-order unit admittance change. For nodes Fault discrimination state variables.

[0132] When a node is located in a faulty area within the zone, the fault discrimination status variable is set to 1; otherwise, it is set to 0.

[0133] In this embodiment, the correlation calculation function of the graph attention network is used to calculate the coupling relationship weight between adjacent nodes in the distribution network topology graph. The expression is as follows:

[0134]

[0135]

[0136] In the formula, For nodes For adjacent nodes The normalized coupling weights range from [0,1]. Larger values ​​indicate larger adjacent nodes. The state of the node The greater the impact of the protection setting parameters, For nodes For adjacent nodes The unnormalized attention coefficient, For nodes The set of adjacent nodes, The relevance calculation function for the graph attention network is implemented using a single-layer feedforward neural network. For nodes The node feature vectors.

[0137] Specifically, the constructed distribution network topology and the node feature vectors of all nodes are used as input to the graph attention network. The input layer transforms the low-dimensional node feature vectors into high-dimensional feature representations suitable for attention computation through a shared linear transformation layer. For any two adjacent nodes in the distribution network topology... and adjacent nodes compute nodes For adjacent nodes The unnormalized attention coefficients are used to ensure that they reflect the relative importance of different neighboring nodes to the current node. This is done by assigning a value to each node. All neighboring nodes The unnormalized attention coefficients are normalized to obtain the normalized coupling weights.

[0138] Based on the coupling relationship weights, the nodes are... All neighboring nodes The corresponding node feature vectors are weighted and aggregated to obtain the node. The comprehensive state characteristics are expressed as follows:

[0139]

[0140] In the formula, For nodes The comprehensive state characteristics.

[0141] Specifically, to further enhance the expressive power and stability of the graph attention network, a multi-head attention mechanism is adopted. This involves executing K independent attention calculations and feature aggregation processes in parallel, where K is set to 4. The results from these K groups are then concatenated or averaged to obtain the final comprehensive state feature. When concatenating, the final comprehensive state feature has a dimension of K×d; when averaging, the dimension remains d.

[0142] In this implementation, the pre-trained mapping functions are mapping functions from comprehensive state characteristics to operating current settings and mapping functions from comprehensive state characteristics to operating time settings, which are trained from historical operating data or simulation data.

[0143] Specifically, the network structure of the mapping function adopts a 3-layer fully connected neural network, as follows:

[0144] Input layer: The number of neurons equals the dimension of the integrated state features;

[0145] Hidden layers: Contains 2 hidden layers, with 64 and 32 neurons in each hidden layer, respectively, and ReLU activation function is used;

[0146] Output layer: Contains 1 neuron, no activation function, outputs the corresponding tuning parameter value.

[0147] The two mapping functions were pre-trained offline using historical operating data and simulation data. The training process is as follows:

[0148] Collect comprehensive state characteristic data of the distribution network under different operating modes, i.e. different distributed power outputs, different topologies, different fault types, i.e. single-phase ground faults, two-phase short-circuit faults, three-phase short-circuit faults, etc., and different fault locations, as well as the corresponding protection setting parameters, which are determined by traditional setting methods or expert experience, to construct training datasets and validation datasets, with the ratio of training datasets to validation datasets being 8:2.

[0149] The input comprehensive state characteristic data is standardized to make the mean 0 and the variance 1; the output protection setting parameters are normalized to make the value range [0,1].

[0150] The Adam optimizer was used for training, the mean squared error (MSE) was used as the loss function, the initial learning rate was set to 0.001, and the number of training epochs was set to 1000. Training was terminated early when the loss of the validation dataset no longer decreased for 50 consecutive epochs.

[0151] Validation was performed using a validation dataset to ensure that the prediction error met the engineering requirements, specifically: the relative error of the operating current setting value did not exceed 5%, and the absolute error of the operating time setting value did not exceed 10ms.

[0152] After training, the two mapping functions are deployed in the edge computing unit of the protection device or the regional control center of the distribution network for real-time reasoning of protection setting parameters.

[0153] After a fault occurs and fault identification is completed, the comprehensive state characteristics of each node are calculated and input into the mapping function to obtain the operating current setting value and operating time setting value of each protection device.

[0154] The expression for calculating the operating current setting value is as follows:

[0155]

[0156] In the formula, For the first The operating current setting value of each protection device. This is a mapping function from the comprehensive state characteristics to the operating current setting value.

[0157] The expression for calculating the motion time setpoint is:

[0158]

[0159] In the formula, For the first The setpoint for the operating time of each protection device. It is a mapping function from the comprehensive state characteristics to the action time tuning value.

[0160] Specifically, A unique identifier for the protection device, associated with the node. One-to-one correspondence. To ensure the selectivity and coordination of protection actions, before outputting protection setting parameters, time difference coordination constraints are applied to the action time setting values ​​of adjacent protection devices to ensure that the protection device closest to the fault point acts as the main protection with priority, and the adjacent protection device acts as the backup protection with a delay. The time difference between the main protection with priority and the backup protection with delay is 0.3 to 0.5 seconds.

[0161] S4. Based on the fault diagnosis result, the operating current setting value, and the operating time setting value, when the real-time measured current meets the preset conditions, output a trip command.

[0162] In this embodiment, the step of outputting a trip command when the real-time measured current meets preset conditions, based on the fault diagnosis result, the operating current setting value, and the operating time setting value, includes:

[0163] When the real-time measured current detected by the protection device is greater than the operating current setting value, the fault identification result is an intra-zone fault, and the duration of the real-time measured current being greater than the operating current setting value exceeds the operating time setting value, a trip command is output, and the protected line is isolated by disconnecting it through the circuit breaker.

[0164] Specifically, the protection device verifies the protection action conditions in real time at a frequency consistent with the sampling period. The protection action conditions must simultaneously meet three independent conditions, which are strictly logically ANDed. If any condition is not met, a trip command should not be output. Specifically:

[0165] Condition a: The real-time measured current detected by the protection device is greater than the corresponding operating current setting value of the protection device. The real-time measured current is the effective value of the positive sequence current component of the three-phase current signal at the installation location of the protection device, calculated from the extracted positive sequence current component, with the calculation frequency consistent with the sampling frequency of the protection device.

[0166] Condition b: The fault identification result is an internal fault, that is, an internal fault has occurred in the protected line, and the protection device is located in the area directly affected by the fault.

[0167] Condition c: The duration for which the real-time measured current is greater than the operating current setting value exceeds the corresponding protection device's operating time setting value.

[0168] The timing logic for the duration is as follows: when the real-time measured current exceeds the operating current setting value for the first time, the high-precision operating timing unit inside the protection device starts and accumulates the time from zero; during the timing process, if the real-time measured current falls back below the operating current setting value at any time, the operating timing unit is immediately cleared and the timing stops; when the accumulated value of the operating timing unit is greater than the operating time setting value, the duration condition is met.

[0169] When the above three protection action conditions are met simultaneously, the protection device immediately executes the trip command output and fault isolation operation. The specific process is as follows:

[0170] The central processing unit of the protection device sends a level-triggered trip signal to the dedicated output relay. After the output relay is activated, it connects the DC power supply circuit of the corresponding circuit breaker trip coil.

[0171] When the circuit breaker trip coil is energized, it drives the circuit breaker operating mechanism to act quickly, causing the three-phase contacts of the circuit breaker to open simultaneously and disconnect the protected faulty line.

[0172] After the circuit breaker has completed its tripping operation, its auxiliary normally closed contact will feed back the tripping status signal to the input acquisition unit of the protection device. If the protection device does not receive the tripping status signal within the set feedback time limit, usually 100ms, it will determine that the circuit breaker has failed to operate and immediately send a circuit breaker failure to operate signal to the adjacent backup protection device to start the backup protection acceleration action process.

[0173] In this embodiment, the time difference of the operating time setting of each protection device is constrained according to the operating time setting value of adjacent protection devices. Based on the distribution network topology and the distribution characteristics of fault electrical quantities, the electrical distance between each protection device and the fault point is calculated. The protection device with the smallest electrical distance is selected as the main protection and takes priority action, while the adjacent protection devices are selected as backup protection and take delayed action.

[0174] The protection action information and related electrical quantities are uploaded to the distribution network dispatch system or distribution network monitoring system for distribution network operation analysis and iterative optimization of graph attention network; the related electrical quantities include three-phase voltage, three-phase current signal, positive sequence voltage component, positive sequence current component, positive sequence unit admittance, positive sequence unit admittance change, positive sequence unit admittance change difference, and positive sequence unit admittance change phase angle difference.

[0175] Specifically, to ensure the selectivity of protection actions and prevent non-faulty areas from being mistakenly isolated, a strict time difference coordination constraint is applied to the action time settings of adjacent protection devices after the action time setting value is set but before it is sent to the protection device. This is specifically implemented as follows:

[0176] Based on the fault identification results, the complete fault area within the zone is determined, and all protection devices within the fault area and those directly adjacent to the fault area within the zone are extracted.

[0177] Based on the distribution network topology and the distribution characteristics of fault electrical quantities, the electrical distance between each candidate protection device and the fault point is calculated. The protection device with the smallest electrical distance is determined as the main protection for this fault, and the remaining protection devices that have a direct electrical connection with the main protection are determined as backup protection.

[0178] Set the time difference between the main protection and the backup protection, with a value ranging from 0.3 to 0.5 seconds. The specific value is determined based on the voltage level of the distribution network, the inherent operating time of the circuit breaker, and the operating time of the relay protection device.

[0179] The operating time settings of all backup protections are corrected. The corrected operating time settings of backup protections are equal to the sum of the operating time settings of the main protections and the time difference, ensuring that the main protections take priority in clearing faults. Backup protections only operate with a delay when the main protections fail to operate or the corresponding circuit breakers fail to operate.

[0180] After the protection device completes fault isolation or determines that the fault has disappeared, it immediately uploads the protection action information and related electrical quantities of this fault to the distribution network dispatching system or distribution network monitoring system for distribution network operation analysis and iterative optimization of graph attention networks. The specific implementation is as follows:

[0181] All uploaded data carries a unified timestamp accurate to the millisecond level, specifically:

[0182] Protection action information includes: unique identification number of the protection device, action time, tripping circuit breaker number, fault diagnosis result, action current setting value, action time setting value, actual action current amplitude, actual action duration, and circuit breaker tripping feedback time.

[0183] Relevant electrical quantities include three-phase voltage signals, three-phase current signals, positive-sequence voltage components, positive-sequence current components, positive-sequence unit admittance, change in positive-sequence unit admittance, difference in change in positive-sequence unit admittance, and phase angle difference in change in positive-sequence unit admittance for the five cycles before the fault occurs to the five cycles after the fault is cleared.

[0184] After receiving and uploading data, the distribution network dispatching system or distribution network monitoring system stores it in a dedicated historical operation database, with a data retention period of no less than one year. Operators can use this database to query historical fault records, analyze protection action behavior, assess the overall performance of the protection system, and investigate the causes of protection malfunctions, failures to operate, or abnormal actions.

[0185] Using a fixed monthly cycle, newly added fault data and protection action information are extracted from a dedicated historical operation database to construct an incremental training dataset. This dataset is then used to incrementally train the graph attention network and two mapping functions. The incremental training process is consistent with the offline pre-training process, employing mean squared error as the loss function and the Adam optimizer for parameter updates. After training, performance is validated. Once validation is successful, the original deployed model is replaced, ensuring continuous adaptation to long-term changes in distribution network operation modes and continuously improving the accuracy and reliability of protection setting parameters.

[0186] like Figure 2 As shown, this is one embodiment of the present invention, which provides an adaptive relay protection system for a distribution network with distributed power sources, comprising:

[0187] The acquisition and extraction module synchronously acquires the three-phase voltage and three-phase current signals at both ends of the protected line of the distribution network using a unified time reference, and extracts the positive sequence voltage component and positive sequence current component to calculate the positive sequence unit admittance.

[0188] The fault identification module transmits the positive sequence unit admittance at both ends of the protected line of the distribution network to each other, performs time synchronization and alignment, calculates the change in positive sequence unit admittance, obtains the difference in positive sequence unit admittance change and the phase angle difference in positive sequence unit admittance change based on the change in positive sequence unit admittance, and obtains the fault identification result according to the preset fault identification rules.

[0189] The state feature module constructs a distribution network topology map based on the fault identification results and the distribution network operation status. Based on the distribution network topology map, a graph attention network is used to calculate the coupling relationship weights. The weighted aggregation based on the coupling relationship weights is used to obtain the comprehensive state features. The action current setting value and action time setting value are output through a pre-trained mapping function.

[0190] The output command module outputs a trip command when the real-time measured current meets the preset conditions, based on the fault diagnosis result, the operating current setting value, and the operating time setting value.

[0191] In summary, this invention addresses the problems of relay protection criterion failure and setting difficulties in distribution networks under conditions of high-proportion distributed power source access. It constructs positive-sequence unit admittance related features based on electrical quantity information at both ends of the protected line in the distribution network, and achieves information exchange between the two ends of the protected line through a communication channel. This enables collaborative analysis of positive-sequence unit admittance change features, thereby achieving accurate identification of internal and external faults in the protected line. Simultaneously, during the protection setting process, the real-time topology of the distribution network is introduced, abstracting the protection devices and their electrical connections corresponding to the protected line into a distribution network topology. The power grid topology diagram is used, and graph attention networks are integrated to adaptively model the coupling relationship between adjacent protection devices, enabling dynamic characterization of the influence between multiple protection devices. Based on this, by constructing a mapping relationship from comprehensive state characteristics to operating current setting values ​​and operating time setting values, the traditional setting method that relies on experience or offline calculation is transformed into an adaptive setting process that combines data-driven and mechanism-constrained approaches. This forms a complete protection process covering "signal acquisition - dual-end discrimination - topology modeling - adaptive setting - action execution", effectively improving the response capability and coordination performance of relay protection in complex operating environments.

[0192] Compared with existing technologies, this invention overcomes the problems of sensitivity degradation and directional failure of traditional overcurrent protection and directional protection under high-proportion distributed power source access conditions by introducing a dual-end collaborative criterion based on positive-sequence unit admittance, thus improving the accuracy and reliability of fault identification of protected lines in distribution networks. Simultaneously, by fusing graph attention networks, it achieves adaptive learning of the coupling relationship between protection devices, overcoming the limitations of traditional fixed setting methods in adapting to changes in distribution network topology and operational adjustments, significantly improving the dynamic adaptability of protection settings. Furthermore, this invention combines data-driven methods with distribution network topology and electrical characteristics, effectively avoiding errors caused by relying solely on empirical setting or static analysis, enhancing the engineering feasibility and generalization ability of the protection strategy. Moreover, this invention can adapt to complex operating conditions such as distributed power source output fluctuations, frequent changes in distribution network structure, and multiple fault scenarios, possessing good real-time performance and scalability. It can be widely applied to relay protection and operation control of new distribution networks with high proportions of distributed power sources, demonstrating high engineering application value and promising prospects for promotion.

[0193] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any other combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product, which includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another.

[0194] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. This storage medium can be a read-only memory, a disk, or an optical disk, etc.

[0195] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An adaptive relay protection method for distribution networks containing distributed generation sources, characterized in that, include: The three-phase voltage and three-phase current signals at both ends of the protected line of the distribution network are synchronously acquired using a unified time reference, and the positive sequence voltage component and positive sequence current component are extracted to calculate the positive sequence unit admittance. After transmitting the positive sequence unit admittances at both ends of the protected line in the distribution network to each other and synchronizing and aligning them in time, the change in positive sequence unit admittance is calculated. Based on the change in positive sequence unit admittance, the difference in positive sequence unit admittance change and the phase angle difference in positive sequence unit admittance change are obtained, and the fault judgment result is obtained according to the preset fault judgment rules. Construct a distribution network topology diagram based on the fault identification results and the distribution network operating status; Based on the distribution network topology, a graph attention network is used to calculate the coupling relationship weights. The weighted aggregation based on the coupling relationship weights is used to obtain the comprehensive state features. The action current setting value and action time setting value are output through a pre-trained mapping function. Based on the fault diagnosis result, the operating current setting value, and the operating time setting value, a trip command is output when the real-time measured current meets the preset conditions.

2. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 1, characterized in that, The method involves synchronously acquiring the three-phase voltage and three-phase current signals at both ends of the protected line in the distribution network using a unified time reference, extracting the positive-sequence voltage and current components, and calculating the positive-sequence unit admittance, including: The three-phase voltage and three-phase current signals are converted into positive sequence voltage and positive sequence current components by means of the symmetrical component method. Among them, the positive sequence voltage component The expression is: ； In the formula, for Phase voltage, for Phase voltage, for Phase voltage, The rotation factor; Positive sequence current component The expression is: ； In the formula, for Phase current signal, for Phase current signal, for Phase current signal; Based on the positive-sequence voltage component and the positive-sequence current component, the expression for calculating the positive-sequence unit admittance is as follows: ； In the formula, It is the positive-order unit admittance.

3. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 1, characterized in that, The process involves transmitting the positive-sequence unit admittances at both ends of the protected line in the distribution network to each other, performing time synchronization and alignment, calculating the change in positive-sequence unit admittance, and obtaining the difference in positive-sequence unit admittance change and the phase angle difference in positive-sequence unit admittance change based on the change in positive-sequence unit admittance. This includes: The positive-sequence unit admittances at both ends of the protected line in the distribution network are transmitted to each other through a communication channel. Time synchronization and alignment processing is performed on the positive-sequence unit admittances at both ends of the protected line to obtain the corresponding first-end and last-end positive-sequence unit admittances at the same moment, and the change in positive-sequence unit admittance is calculated. The change in positive-sequence unit admittance includes the change in first-end and last-end positive-sequence unit admittance. Specifically, the change in first-end positive-sequence unit admittance is calculated... The expression is: ； In the formula, For a moment The first positive-order unit admittance below. The time interval between adjacent sampling times; Calculate the change in positive-sequence unit admittance at the end. The expression is: ； In the formula, For a moment The terminal positive-sequence unit admittance is below; Calculate the difference in positive-sequence unit admittance based on the changes in the first-order and last-order positive-sequence unit admittances. The expression is: ； Calculate the phase angle difference of the positive-sequence unit admittance change. The expression is: ； In the formula, The phase angle is the change in the first positive sequence unit admittance. The phase angle represents the change in positive-sequence unit admittance at the terminal.

4. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 3, characterized in that, The step of obtaining the fault identification result according to the preset fault identification rules includes: The preset fault detection rule is as follows: ； In the formula, The threshold for positive-order unit admittance difference. The phase angle difference threshold, For logical AND relation; The fault identification results include: faults within the zone or faults outside the zone; If the preset fault discrimination rules are met, the fault is determined to be an internal fault; otherwise, it is an external fault.

5. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 4, characterized in that, The power grid operating status includes the power grid switch status, distributed power source access status, and line connection relationships. The power distribution network topology diagram for: ； In the formula, For a set of nodes, It is an edge set; For each node in the node set, define a node feature vector, expressed as follows: ； In the formula, For nodes The node feature vectors, For nodes positive-order unit admittance, For nodes The positive-order unit admittance change. For nodes Fault discrimination state variables; When a node is located in a faulty area within the zone, the fault identification status variable is set to 1; otherwise, it is set to 0.

6. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 5, characterized in that, The calculation of coupling weights using a graph attention network includes: The correlation calculation function of graph attention network is used to calculate the coupling relationship weight between adjacent nodes in the distribution network topology graph. The expression is as follows: ； ； In the formula, For nodes For adjacent nodes The normalized coupling weights, For nodes For adjacent nodes The unnormalized attention coefficient, For nodes The set of adjacent nodes, The relevance calculation function for graph attention networks. For nodes Node feature vectors; Based on the weighted coupling relationship, the nodes are... All neighboring nodes The corresponding node feature vectors are weighted and aggregated to obtain the node. The comprehensive state characteristics are expressed as follows: ； In the formula, For nodes The comprehensive state characteristics.

7. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 6, characterized in that, The pre-trained mapping function outputs the action current setting value and the action time setting value, including: The pre-trained mapping functions are mapping functions from comprehensive state features to action current settings and mapping functions from comprehensive state features to action time settings, which are trained from historical operating data or simulation data. The expression for calculating the operating current setting value is as follows: ； In the formula, For the first The operating current setting value of each protection device. This is a mapping function from the comprehensive state characteristics to the operating current setting value; The expression for calculating the motion time setpoint is: ； In the formula, For the first The setpoint for the operating time of each protection device. It is a mapping function from the comprehensive state characteristics to the action time tuning value.

8. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 4, characterized in that, Based on the fault diagnosis result, the operating current setting value, and the operating time setting value, when the real-time measured current meets the preset conditions, a trip command is output, including: When the real-time measured current detected by the protection device is greater than the operating current setting value, the fault identification result is an intra-zone fault, and the duration of the real-time measured current being greater than the operating current setting value exceeds the operating time setting value, a trip command is output, and the protected line is isolated by disconnecting it through the circuit breaker.

9. The adaptive relay protection method for a distribution network containing distributed power sources according to claim 8, characterized in that, Based on the operating time settings of adjacent protection devices, time difference coordination constraints are applied to the operating time settings of each protection device. Based on the distribution network topology and the distribution characteristics of fault electrical quantities, the electrical distance between each protection device and the fault point is calculated. The protection device with the smallest electrical distance is selected as the main protection with priority operation, and the adjacent protection devices are selected as backup protection with delayed operation. The protection action information and related electrical quantities are uploaded to the distribution network dispatch system or distribution network monitoring system for distribution network operation analysis and iterative optimization of graph attention network; the related electrical quantities include three-phase voltage, three-phase current signal, positive sequence voltage component, positive sequence current component, positive sequence unit admittance, positive sequence unit admittance change, positive sequence unit admittance change difference, and positive sequence unit admittance change phase angle difference.

10. An adaptive relay protection system for a distribution network containing distributed power sources, and an adaptive relay protection method for a distribution network containing distributed power sources as described in claims 1 to 9, characterized in that, The acquisition and extraction module synchronously acquires the three-phase voltage and three-phase current signals at both ends of the protected line of the distribution network using a unified time reference, and extracts the positive sequence voltage component and positive sequence current component to calculate the positive sequence unit admittance. The fault identification module transmits the positive sequence unit admittance at both ends of the protected line of the distribution network to each other, performs time synchronization and alignment, calculates the change in positive sequence unit admittance, obtains the difference in positive sequence unit admittance change and the phase angle difference in positive sequence unit admittance change based on the change in positive sequence unit admittance, and obtains the fault identification result according to the preset fault identification rules. The status feature module constructs a distribution network topology map based on fault identification results and distribution network operating status. Based on the distribution network topology, a graph attention network is used to calculate the coupling relationship weights. The weighted aggregation based on the coupling relationship weights is used to obtain the comprehensive state features. The action current setting value and action time setting value are output through a pre-trained mapping function. The output command module outputs a trip command when the real-time measured current meets the preset conditions, based on the fault diagnosis result, the operating current setting value, and the operating time setting value.

Images