Power failure detection method

Abstract

The invention discloses a power fault detection method, which belongs to the field of power fault detection, can improve the accuracy and response speed of fault detection, can enhance the reliability and operation and maintenance efficiency of a power grid, and has wide application prospects and remarkable economic benefits. The method mainly comprises the following steps: collecting real-time operation data from each monitoring point of a power grid; cleaning and standardizing the collected data; key features in the data are extracted by using a deep learning model; training a deep learning model by using the labeled fault data set; inputting real-time data into the trained deep learning model, and carrying out fault detection; determining the position and the type of the fault based on the output of the model; and according to a diagnosis result, proper measures are taken for troubleshooting or isolating. The method is mainly used for power fault detection.

Description

Technical Field

[0001] The present invention relates to the field of power fault detection, and in particular to a power fault detection method. Background Art

[0002] As the complexity and scale of power systems continue to expand, traditional fault detection methods have many deficiencies in accuracy and response speed. Deep learning technology has achieved remarkable results in image recognition, natural language processing and other fields with its powerful feature extraction and model generalization capabilities. This paper proposes a power grid fault detection method based on deep learning, which aims to improve the accuracy and efficiency of fault detection and ensure the safe and stable operation of the power grid. Summary of the invention

[0003] The purpose of the present invention is to provide a power fault detection method, which can improve the accuracy and response speed of fault detection, and can also enhance the reliability and operation and maintenance efficiency of the power grid, and has broad application prospects and significant economic benefits; through data standardization, feature extraction, model training and other steps, the high efficiency and accuracy of fault detection are ensured, and strong technical support is provided for the safe and stable operation of the power grid.

[0004] The present invention is achieved through the following technical solutions:

[0005] A method for detecting power failures comprises the following steps:

[0006] a) Collect real-time operation data from various monitoring points of the power grid;

[0007] b) Clean and standardize the collected data;

[0008] c) Use deep learning models to extract key features from data;

[0009] d) Use the labeled fault dataset to train the deep learning model;

[0010] e) Input real-time data into the trained deep learning model for fault detection;

[0011] f) Based on the output of the model, determine the location and type of fault;

[0012] g) Based on the diagnosis results, take appropriate measures to troubleshoot or isolate the problem.

[0013] Furthermore, the step a) includes deploying various sensors and remote terminal units at key locations of the power grid, continuously collecting power grid operation data, including voltage, current, frequency, and temperature parameters, through sensors and RTUs, and transmitting the collected data to a central database or cloud platform in real time to ensure the integrity and real-time nature of the data.

[0014] Furthermore, the step b) includes filling missing values ​​using linear interpolation or polynomial interpolation, detecting and removing outliers using isolation forest method; converting data of different units into the same unit, scaling the data to between 0 and 1 and converting it into a standard normal distribution.

[0015] Furthermore, the step c) includes selecting a convolutional neural network according to the characteristics of the data and task requirements, extracting the time series features of voltage and current, extracting local and global features in the data through multi-layer convolution and pooling operations, and selecting the most representative and discriminative features based on the output of the model for subsequent model training.

[0016] Furthermore, the step d) includes using historical fault data to annotate the data, marking the time and type of the fault; using the TensorFlow deep learning framework: using the annotated fault data set to supervise the deep learning model, optimizing the model parameters through the back propagation algorithm, using the loss function to evaluate the performance of the model, using the validation set to tune the hyperparameters, and selecting the optimal model parameters; using an independent validation set to verify the trained model and evaluate the performance indicators of the model.

[0017] Furthermore, the step e) includes inputting the real-time data into a trained deep learning model after preprocessing; calculating the output result through forward propagation of the model to obtain the probability of fault occurrence or classification result; setting a threshold value according to the probability value output by the model, and determining it as a fault if the threshold value is exceeded.

[0018] Furthermore, the step f) includes identifying the mode that may cause the fault based on the output of the model, and determining the specific location of the fault in combination with the topology of the power grid and the equipment configuration; analyzing the characteristics of the fault mode and determining the type of fault; and generating a detailed fault diagnosis report.

[0019] Furthermore, the step g) includes formulating an emergency response plan according to the severity and scope of the fault; notifying relevant personnel to operate according to the emergency response plan; and after the maintenance personnel arrive at the scene, they perform troubleshooting and repair the damaged equipment.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. The deep learning model can adapt to the dynamic changes of the power grid and handle complex and changing fault modes.

[0022] 2. Through the analysis of historical fault data, the fault tree model is continuously optimized to improve the self-learning and adaptive capabilities of the system.

[0023] 3. Rationally dispatch maintenance resources according to the severity and impact scope of the fault to improve operation and maintenance efficiency.

[0024] 4. Rapidly detect and handle faults, reduce grid downtime, and ensure the continuity of power supply. DETAILED DESCRIPTION

[0025] The present invention is further described below.

[0026] Embodiment: A method for detecting a power failure comprises the following steps:

[0027] a) Collect real-time operation data from various monitoring points of the power grid;

[0028] b) Clean and standardize the collected data;

[0029] c) Use deep learning models to extract key features from data;

[0030] d) Use the labeled fault dataset to train the deep learning model;

[0031] e) Input real-time data into the trained deep learning model for fault detection;

[0032] f) Based on the output of the model, determine the location and type of fault;

[0033] g) Based on the diagnosis results, take appropriate measures to troubleshoot or isolate the problem.

[0034] The step a) includes deploying various sensors and remote terminal units at key locations of the power grid, continuously collecting power grid operation data including voltage, current, frequency, and temperature parameters through sensors and RTUs, and transmitting the collected data to a central database or cloud platform in real time to ensure the integrity and real-time nature of the data;

[0035] The step b) includes filling missing values ​​using linear interpolation or polynomial interpolation, detecting and removing outliers using the isolation forest method; converting data in different units to the same unit, scaling the data to between 0 and 1 and converting it to a standard normal distribution;

[0036] The step c) includes selecting a convolutional neural network according to the characteristics of the data and the task requirements, extracting the time series features of the voltage and current, extracting the local and global features in the data through multi-layer convolution and pooling operations, and selecting the most representative and discriminative features according to the output of the model for subsequent model training;

[0037] The step d) includes using historical fault data to annotate the data, marking the time and type of the fault; using the TensorFlow deep learning framework: using the annotated fault data set to supervise the deep learning model, optimizing the model parameters through the back propagation algorithm, using the loss function to evaluate the performance of the model, using the validation set to perform hyperparameter tuning, and selecting the optimal model parameters; using an independent validation set to verify the trained model and evaluate the performance indicators of the model;

[0038] The step e) includes inputting the real-time data into a trained deep learning model after preprocessing; calculating the output result through the forward propagation of the model to obtain the probability of the fault or the classification result; setting a threshold value according to the probability value output by the model, and determining it as a fault if it exceeds the threshold value;

[0039] The step f) includes identifying the mode that may cause the fault based on the output of the model, and determining the specific location of the fault in combination with the topology of the power grid and the equipment configuration; analyzing the characteristics of the fault mode and determining the type of the fault; and generating a detailed fault diagnosis report;

[0040] The step g) includes formulating an emergency response plan according to the severity and impact scope of the fault; notifying relevant personnel to operate according to the emergency response plan; and after the maintenance personnel arrive at the scene, they perform troubleshooting and repair the damaged equipment.

Claims

1. A method for detecting power failure, characterized in that: The steps include: a) Collect real-time operation data from various monitoring points of the power grid; b) Clean and standardize the collected data; c) Use deep learning models to extract key features from data; d) Use the labeled fault dataset to train the deep learning model; e) Input real-time data into the trained deep learning model for fault detection; f) Based on the output of the model, determine the location and type of fault; g) Based on the diagnosis results, take appropriate measures to troubleshoot or isolate the problem.

2. The power failure detection method according to claim 1, characterized in that: The step a) includes deploying various sensors and remote terminal units at key locations of the power grid, continuously collecting power grid operation data including voltage, current, frequency, and temperature parameters through sensors and RTUs, and transmitting the collected data to a central database or cloud platform in real time to ensure the integrity and real-time nature of the data.

3. The power failure detection method according to claim 1, characterized in that: The step b) includes filling missing values ​​using linear interpolation or polynomial interpolation, detecting and removing outliers using the isolation forest method; converting data in different units into the same unit, scaling the data to between 0 and 1 and converting it into a standard normal distribution.

4. The power failure detection method according to claim 1, characterized in that: The step c) includes selecting a convolutional neural network according to the characteristics of the data and the task requirements, extracting the time series features of the voltage and current, extracting the local and global features in the data through multi-layer convolution and pooling operations, and selecting the most representative and discriminative features according to the output of the model for subsequent model training.

5. The power failure detection method according to claim 1, characterized in that: The step d) includes using historical fault data to annotate the data, marking the time and type of the fault; Use the TensorFlow deep learning framework: Use the labeled fault dataset to supervise the deep learning model, optimize the model parameters through the back propagation algorithm, use the loss function to evaluate the performance of the model, use the validation set to tune the hyperparameters and select the optimal model parameters; use an independent validation set to verify the trained model and evaluate the performance indicators of the model.

6. The power failure detection method according to claim 1, characterized in that: The step e) includes inputting the real-time data into a trained deep learning model after preprocessing; The output result is calculated through the forward propagation of the model to obtain the probability of fault occurrence or classification result; a threshold is set according to the probability value output by the model, and a fault is determined if it exceeds the threshold.

7. The power failure detection method according to claim 1, characterized in that: The step f) includes identifying the mode that may cause the fault based on the output of the model, and determining the specific location of the fault in combination with the topology of the power grid and the equipment configuration; analyzing the characteristics of the fault mode and determining the type of fault; and generating a detailed fault diagnosis report.

8. The power failure detection method according to claim 1, characterized in that: The step g) includes formulating an emergency response plan according to the severity and impact scope of the fault; Notify relevant personnel to follow the emergency response plan; After the maintenance personnel arrive at the scene, they carry out troubleshooting and repair the damaged equipment.