Power distribution network override trip protection method based on distributed network

Abstract

The invention relates to the technical field of power distribution network protection, and provides a power distribution network override trip protection method based on a distributed network. S1, arranging multiple layers of intelligent distributed protection devices at different nodes of the power distribution network, wherein each layer of protection device has different fault detection and protection decision capabilities; and S2, realizing localized data processing and fault detection through the edge calculation unit of each protection device, wherein the fault detection carries out real-time fault identification based on a deep learning model. According to the arrangement scheme of the multi-layer intelligent distributed protection device, the protection device is divided into a high-voltage layer, a medium-voltage layer and a low-voltage layer according to the voltage grade and the load importance, the technical effects of step-by-step fault isolation and layered protection are achieved, and compared with the arrangement mode of a unified protection device in the prior art, the protection device has the advantages of being simple in structure and convenient to use. The problem of large-range override tripping is effectively solved, layered response of different levels of protection devices to faults is realized, the fault influence range is further controlled, and the stability of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network protection, and in particular to a distribution network over-tripping protection method based on a distributed network. Background Art

[0002] In modern distribution network systems, with the rapid growth of power demand and power supply load, the system faces more and more complex faults and load fluctuations during operation. The protection method of traditional distribution networks usually adopts centralized control mode and fixed protection parameters. Once a fault occurs, it often causes over-tripping, which further expands the scope of power outage and affects the stability of power supply. The existing technology has shortcomings in the following aspects and is difficult to cope with the increasingly complex distribution network protection needs.

[0003] Existing technologies usually adopt centralized or single protection device control mode. This mode relies on centralized data processing and decision-making. Therefore, when a sudden fault occurs, the protection device needs to transmit the fault information to the control center, and then execute the corresponding protection measures after processing. Due to the reliance on the calculation and command transmission of the central node, the delay is increased, resulting in a slow response speed of fault isolation and the inability to achieve real-time protection. In particular, it is difficult for the protection system in the wide area power grid to respond quickly, which is prone to cause more serious over-tripping accidents. The operating status of the distribution network is greatly affected by load fluctuations and environmental changes. Especially during peak hours or complex external environments, traditional protection devices usually use fixed protection parameters for fault detection. It is difficult to adjust the protection parameters according to real-time conditions, resulting in a decrease in protection accuracy, and even false tripping or missed tripping when the load fluctuates greatly. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a distribution network over-tripping protection method based on a distributed network, which solves the problems of slow response speed of centralized protection mode in the distribution network system, single protection level, insufficient fault detection adaptability, low decision consistency and lack of topology reconstruction and self-healing function after fault.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a distribution network over-tripping protection method based on a distributed network, comprising the following steps;

[0006] S1. Multiple layers of intelligent distributed protection devices are arranged at different nodes of the distribution network, and each layer of protection devices has different fault detection and protection decision-making capabilities;

[0007] S2. localized data processing and fault detection are realized through the edge computing unit of each protection device, and the fault detection is based on real-time fault identification based on a deep learning model;

[0008] S3. Use the federated learning mechanism to share the fault detection model among the protection devices and dynamically optimize the protection parameters of each protection device;

[0009] S4, multi-layer smart contracts and consistency protocols built on blockchain to establish distributed communication and decision-making coordination among various protection devices;

[0010] S5. Dynamically optimize protection parameters through spatiotemporal data analysis to adapt to changes in the environment and load and prevent over-tripping;

[0011] S6. After detecting a fault, the system automatically reconstructs the topology and self-heals, and switches to island mode if necessary to maintain system stability.

[0012] Preferably, the multi-layer intelligent distributed protection device in step S1 includes:

[0013] The high-voltage layer protection device uses a combined model of convolutional neural network (CNN) and long short-term memory network (LSTM) to extract the spatial and temporal characteristics of current and voltage signals to achieve full network fault detection;

[0014] The medium voltage layer protection device is based on the LSTM model and uses time series analysis methods to perform real-time monitoring and fault prediction of electrical parameters within the region;

[0015] The low-voltage layer protection device adopts a lightweight neural network model or a decision tree model to quickly detect terminal short circuit and overload faults.

[0016] Preferably, the fault detection in the edge computing unit in step S2 includes data preprocessing, feature extraction and real-time classification, wherein;

[0017] The feature extraction is based on a calculation formula of a convolutional neural network (CNN) model;

[0018] F(X)=W·X+b

[0019] Among them, F(X) is the feature matrix, X is the input signal data, W is the weight matrix, and b is the bias term;

[0020] The fault classification uses the softmax activation function to calculate the classification result.

[0021]

[0022] Among them, Z is the model output vector and y is the fault category.

[0023] Preferably, the step S3 specifically includes:

[0024] Each protection device trains its fault detection model locally and sends model parameter updates to the central server;

[0025] The central server aggregates and calculates the received model parameters to form a global model, and sends the global model to each protection device;

[0026] The model parameter aggregation calculation formula is:

[0027]

[0028] Among them, W k is the local model weight of the kth protection device, n k is the data volume, and N is the total number of devices.

[0029] Preferably, the multi-layer smart contract constructed based on blockchain in step S4 includes the following contents:

[0030] High-level smart contracts are used for data exchange and protection coordination between high-voltage layer protection devices;

[0031] The middle-level smart contract is used for local protection coordination between medium-voltage protection devices;

[0032] Low-level smart contracts are used for rapid fault isolation and terminal protection decision-making of low-voltage layer protection devices.

[0033] Preferably, in step S4, decision synchronization and consistency of each protection device are achieved through the RAFT consensus protocol, and the leader election of the RAFT protocol is calculated by the following formula:

[0034]

[0035] Among them, δ(i,j) represents the voting value of node i for node j, and N is the total number of nodes participating in the vote.

[0036] Preferably, the dynamic optimization of protection parameters performed by spatiotemporal data analysis in step S5 includes:

[0037] Identify load patterns and fault modes through time series analysis and clustering algorithms;

[0038] Use Bayesian optimization algorithm to determine the best combination of protection parameters to improve the system's adaptability to load changes;

[0039] The protection parameter adjustment target formula is:

[0040] S optimal = argmax S P(accurate detection|S)

[0041] Where S is the protection sensitivity parameter.

[0042] Preferably, in step S6, after the fault is isolated, the system automatically performs topology reconstruction, optimizes the power transmission path by using the shortest path algorithm, and the optimization goal of the topology reconstruction is;

[0043]

[0044] Among them, P loss (i) is the power loss on the i-th line.

[0045] Preferably, the topology reconstruction further optimizes the power supply path through a genetic algorithm, so that the system can still meet the load demand after the fault is isolated, thereby achieving dynamic balance of the load.

[0046] Preferably, in step S6, in the case of a serious fault, the system automatically switches to the island mode, and the judgment condition of the island detection is;

[0047]

[0048] When I island =0, the island mode is automatically triggered to ensure that the faulty area can operate in an independent mode and reduce the impact on other areas.

[0049] The present invention provides a distribution network over-tripping protection method based on a distributed network, which has the following beneficial effects:

[0050] 1. The present invention divides the protection devices into high voltage layer, medium voltage layer and low voltage layer according to the voltage level and load importance through the layout scheme of multi-layer intelligent distributed protection devices, thus achieving the technical effect of step-by-step fault isolation and hierarchical protection. Compared with the layout method of unified protection devices in the prior art, it effectively solves the problem of large-scale over-level tripping, realizes the hierarchical response of protection devices at different levels to faults, further controls the scope of fault impact, and improves the stability of the system.

[0051] 2. The present invention realizes the sharing and dynamic optimization of fault detection models among protection devices through the federated learning mechanism, achieving the technical effect of improving fault detection accuracy and protecting system privacy. Compared with the single model or data center model update solution in the prior art, federated learning solves the problems of insufficient data privacy protection and untimely model updates by distributing and optimizing model parameters, so that each protection device can adapt to the changes in the distribution network state in real time, further improving the accuracy and robustness of fault detection.

[0052] 2. The present invention realizes system recovery after fault isolation through topology reconstruction and self-healing function, achieving the technical effect of ensuring stable power supply of the distribution network after fault isolation. Compared with the prior art solution that is limited to fault isolation but does not have self-healing function, the present invention optimizes the power supply path through the shortest path algorithm and genetic algorithm, solves the problem of unbalanced power supply and untimely recovery after fault isolation, can automatically allocate loads, reconstruct network topology, and improve the self-healing ability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] Example:

[0056] Please see attached Figure 1 , an embodiment of the present invention provides a distribution network over-tripping protection method based on a distributed network, comprising the following steps;

[0057] S1. Multiple layers of intelligent distributed protection devices are arranged at different nodes of the distribution network, and each layer of protection devices has different fault detection and protection decision-making capabilities;

[0058] S2. localized data processing and fault detection are realized through the edge computing unit of each protection device, and the fault detection is based on real-time fault identification based on a deep learning model;

[0059] S3. Use the federated learning mechanism to share the fault detection model among the protection devices and dynamically optimize the protection parameters of each protection device;

[0060] S4, multi-layer smart contracts and consistency protocols built on blockchain to establish distributed communication and decision-making coordination among various protection devices;

[0061] S5. Dynamically optimize protection parameters through spatiotemporal data analysis to adapt to changes in the environment and load and prevent over-tripping;

[0062] S6. After detecting a fault, the system automatically reconstructs the topology and self-heals, and switches to island mode if necessary to maintain system stability.

[0063] In this embodiment, multi-layer intelligent distributed protection devices are arranged at different nodes of the distribution network. According to the voltage level and load importance of different nodes, the protection devices are divided into high voltage layer, medium voltage layer and low voltage layer to form a three-level hierarchical architecture to achieve step-by-step fault isolation and hierarchical protection strategy;

[0064] High-voltage layer protection device: Deploy protection devices at high-voltage nodes such as main substations to protect high-voltage lines and backbone networks. This layer uses a combination model of convolutional neural network (CNN) and long short-term memory network (LSTM). The convolutional layer extracts the spatial features of current and voltage signals, and uses LSTM to capture the temporal features of signals to achieve fault detection and classification of high-voltage lines. Specifically, CNN is used to extract the feature matrix:

[0065] F(X)=W·X+b

[0066] Among them, X is the input signal data, W is the weight matrix, b is the bias term, and the output feature matrix F(X) is used for subsequent fault judgment.

[0067] Medium voltage layer protection device: This layer of protection device is deployed in the medium voltage substation and is responsible for fault isolation and protection in the medium voltage area. The LSTM model is used for time series analysis, which can monitor the electrical parameters in the area in real time and predict the fault status. The LSTM model used here uses the following formula to update the implicit state:

[0068] h t =σ(W h ·[h t-1 ,x t ]+b h )

[0069] Among them, h t represents the hidden state, W h is the weight matrix, x t The current input ensures the fault identification capability of the medium voltage layer protection device within the area.

[0070] Low-voltage layer protection device: The low-voltage layer protection device is arranged at the load center or user end, and adopts a lightweight neural network or decision tree model, which is mainly used to quickly detect short circuits and overload protection at the end. The low-voltage layer has a high response speed and can make localized protection decisions in time;

[0071] Through the multi-level protection device arrangement, each level of protection device can independently detect and respond to faults according to its corresponding voltage level and regional characteristics, forming a step-by-step isolation mechanism. This design significantly reduces the probability of over-tripping and helps to minimize the impact when a fault occurs.

[0072] In this embodiment, localized data processing and fault detection are achieved through the edge computing unit integrated in each protection device. The edge computing unit includes a data acquisition module, a data preprocessing module, a feature extraction module and a fault classification module to ensure the real-time and reliability of fault detection.

[0073] Data preprocessing: Perform noise filtering, normalization and other preprocessing operations on the collected current and voltage data to enhance the validity of the data and the detection accuracy of the model;

[0074] Feature extraction: In the high-voltage layer protection device, the CNN model is used to perform convolution operations on the preprocessed signal data to extract the feature matrix F(X)=W·X+b, where X is the input signal data, W is the weight matrix, and b is the bias term. The medium-voltage and low-voltage layer protection devices use a more adaptable feature extraction method to ensure the detection efficiency of each level.

[0075] Fault classification: The fault classification module uses the softmax activation function to calculate the classification results;

[0076]

[0077] Where Z is the model output vector and y is the fault category. This module outputs the fault type in real time to guide the specific decision-making of fault handling;

[0078] Through localized data processing and fault detection by edge computing units, protection devices can identify fault types in the shortest possible time, reduce communication delays, improve the real-time nature of detection, and reduce dependence on central control systems;

[0079] In this embodiment, based on the federated learning mechanism, the fault detection model is shared and dynamically optimized among distributed protection devices. Each protection device trains its fault detection model locally, and the generated model parameter updates are synchronously shared through the federated learning mechanism;

[0080] Local model training: Each protection device collects fault data locally and independently trains a deep learning model for regional fault detection. After the model parameters are updated, they are sent to the central server;

[0081] Model parameter aggregation: The central server receives the model parameters uploaded by each protection device and performs aggregation calculation to form a global model. The aggregation formula is:

[0082]

[0083] Among them, W k is the local model weight of the kth protection device, n k is the data volume of the device, and N is the total number of protection devices participating in federated learning;

[0084] Model update and distribution: The central server distributes the aggregated global model to each protection device, so that all devices have the latest fault detection capabilities synchronously.

[0085] In this embodiment, a distributed communication mechanism is constructed through blockchain to achieve data sharing and decision-making coordination among protection devices at all levels to ensure decision consistency and information transparency.

[0086] Multi-layer smart contract design: Multi-layer smart contracts in the blockchain ensure the independence and coordination of protection devices at all levels:

[0087] High-level smart contracts are used for fault data exchange and protection coordination at the high-voltage layer;

[0088] The middle-level smart contract is used for regional protection coordination at the medium-voltage layer;

[0089] Low-level smart contracts are used for rapid fault isolation and terminal protection decisions at the low-voltage layer.

[0090] Consistency protocol: Use the RAFT consistency protocol to ensure synchronization of decisions among protection devices when a fault occurs. The leadership election formula of the RAFT protocol is:

[0091]

[0092] Wherein, δ(i,j) represents the voting value of node i for node j, and N is the total number of voting nodes. Based on blockchain and consistency protocol, the decision synchronization of protection devices can be achieved when a fault occurs, avoiding false operation or over-tripping due to inconsistent communication.

[0093] In this embodiment, dynamic optimization of protection parameters is achieved through spatiotemporal data analysis to adapt to different loads and environmental conditions.

[0094] Spatiotemporal data collection: collect current, voltage, load, temperature and other data from each node of the distribution network, perform spatiotemporal data preprocessing, and convert them into multi-dimensional feature vectors;

[0095] Load pattern and fault pattern identification: Use time series analysis models and spatial clustering algorithms to analyze load patterns and fault patterns and identify areas with high fault incidence;

[0096] Dynamic protection parameter adjustment based on Bayesian optimization: The optimal protection parameter combination is determined through Bayesian optimization to improve the system's adaptability to load changes. The parameter adjustment target formula is:

[0097] S optimal = argmax S P(accurate detection|S)

[0098] Where S is the protection sensitivity parameter.

[0099] Dynamic protection optimization driven by spatiotemporal data enhances the system's adaptability under different loads and environments and reduces the probability of over-tripping.

[0100] In this embodiment, after a fault is detected, the system automatically performs topology reconstruction and self-healing processes, and switches to island mode when necessary to maintain system stability;

[0101] Topology reconstruction: After fault isolation, the power transmission path is optimized through the shortest path algorithm, and the optimization goal is;

[0102]

[0103] Among them, P loss (i) is the power loss on the i-th line.

[0104] Genetic algorithm optimization path: Use genetic algorithm to optimize the power supply path to ensure that the system can still meet the load demand after fault isolation and achieve dynamic load balance;

[0105] Islanding mode switching: In the event of a serious fault, the system automatically detects the total current in the fault area and when the islanding condition is met:

[0106]

[0107] The island mode is automatically triggered to enable the affected area to operate independently, reducing the impact on other areas. Through self-healing functions, topology reconstruction and island mode switching, the system can quickly restore to a stable state after a fault occurs, ensuring the continuous power supply capacity of the distribution network.

[0108] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A distribution network over-tripping protection method based on a distributed network, characterized in that: The steps include: S1. Multiple layers of intelligent distributed protection devices are arranged at different nodes of the distribution network, and each layer of protection devices has different fault detection and protection decision-making capabilities; S2. localized data processing and fault detection are realized through the edge computing unit of each protection device, and the fault detection is based on real-time fault identification based on a deep learning model; S3. Use the federated learning mechanism to share the fault detection model among the protection devices and dynamically optimize the protection parameters of each protection device; S4, multi-layer smart contracts and consistency protocols built on blockchain to establish distributed communication and decision-making coordination among various protection devices; S5. Dynamically optimize protection parameters through spatiotemporal data analysis to adapt to changes in the environment and load and prevent over-tripping; S6. After detecting a fault, the system automatically reconstructs the topology and self-heals, and switches to island mode if necessary to maintain system stability.

2. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: The multi-layer intelligent distributed protection device in step S1 includes: The high-voltage layer protection device uses a combined model of convolutional neural network (CNN) and long short-term memory network (LSTM) to extract the spatial and temporal characteristics of current and voltage signals to achieve full network fault detection; The medium voltage layer protection device is based on the LSTM model and uses time series analysis methods to perform real-time monitoring and fault prediction of electrical parameters within the region; The low-voltage layer protection device adopts a lightweight neural network model or a decision tree model to quickly detect terminal short circuit and overload faults.

3. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: The fault detection in the edge computing unit in step S2 includes data preprocessing, feature extraction and real-time classification, wherein; The feature extraction is based on a calculation formula of a convolutional neural network (CNN) model; F(X)=W·X+b Among them, F(X) is the feature matrix, X is the input signal data, W is the weight matrix, and b is the bias term; The fault classification uses the softmax activation function to calculate the classification result. Among them, Z is the model output vector and y is the fault category.

4. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: The step S3 specifically includes: Each protection device trains its fault detection model locally and sends model parameter updates to the central server; The central server aggregates and calculates the received model parameters to form a global model, and sends the global model to each protection device; The model parameter aggregation calculation formula is: Among them, W k is the local model weight of the kth protection device, n k is the data volume, and N is the total number of devices.

5. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: The multi-layer smart contract constructed based on blockchain in step S4 includes the following contents: High-level smart contracts are used for data exchange and protection coordination between high-voltage layer protection devices; The middle-level smart contract is used for local protection coordination between medium-voltage protection devices; Low-level smart contracts are used for rapid fault isolation and terminal protection decision-making of low-voltage layer protection devices.

6. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: In step S4, the decision synchronization and consistency of each protection device are achieved through the RAFT consensus protocol, and the leader election of the RAFT protocol is calculated by the following formula: Among them, δ(i,j) represents the voting value of node i for node j, and N is the total number of nodes participating in the vote.

7. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: The dynamic optimization of protection parameters performed by spatiotemporal data analysis in step S5 includes: Identify load patterns and fault modes through time series analysis and clustering algorithms; Use Bayesian optimization algorithm to determine the best combination of protection parameters to improve the system's adaptability to load changes; The protection parameter adjustment target formula is: S optimal =argmax S P(accurate detection|S) Among them, S is the protection sensitivity parameter.

8. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: In step S6, after the fault is isolated, the system automatically performs topology reconstruction and optimizes the power transmission path through the shortest path algorithm. The optimization goal of the topology reconstruction is: Among them, P loss (i) is the power loss on the i-th line.

9. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: The topology reconstruction further optimizes the power supply path through a genetic algorithm, so that the system can still meet the load demand after fault isolation and achieve dynamic balance of the load.

10. The method for over-tripping protection of a distribution network based on a distributed network according to claim 1, characterized in that: In the step S6, in the case of a serious fault, the system automatically switches to the island mode, and the judgment condition of the island detection is: When I island =0, the island mode is automatically triggered to ensure that the faulty area can operate in an independent mode and reduce the impact on other areas.

Images