Energy router anomaly detection method

Abstract

The invention relates to the technical field of routing energy detection, and discloses an energy router anomaly detection method, which comprises the following steps of: deploying a sensor at a key node to acquire data such as voltage, current, power, temperature and the like in a data acquisition stage so as to ensure real-time accuracy, and ensuring high precision by adopting an optical fiber and a Hall sensor; then, data preprocessing denoising and normalization are carried out, and missing values are filled through linear interpolation; and then extracting complex features by constructing a convolutional neural network containing a pooling layer, inputting feature vectors into an anomaly detection model based on a support vector data description algorithm to judge anomaly, and automatically giving an alarm when the anomaly occurs. Model training adopts cross validation, system initialization is performed before detection, reasons are analyzed after detection, and the model is periodically updated to adapt to changes. The method has the advantages that the defects of an existing detection means in the aspects of real-time performance, accuracy and data processing capacity are overcome, and accurate monitoring of the running state of the energy router and timely discovery of abnormity are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of routing energy detection, and specifically refers to an abnormal detection method for an energy router. Background Art

[0002] With the rapid development of the smart grid, as a core device for realizing efficient power distribution, conversion, and management, the operation stability of the energy router is of crucial importance. The energy router operates in a complex power environment and is affected by various factors such as voltage fluctuations, load mutations, and equipment aging, making it extremely prone to abnormal conditions. Once an abnormality occurs in the energy router, it will not only affect the stability and reliability of local power supply, but may also trigger a chain reaction, causing a serious impact on the entire power system.

[0003] Traditional abnormal detection means for energy routers mainly rely on manual inspections and simple threshold judgment methods. Manual inspections are inefficient and difficult to detect potential problems in real time, with obvious lag. And simple threshold judgment methods can only monitor a few specific parameters and are often difficult to accurately identify complex and non-linear abnormal situations. For example, when there is slight component aging or local overheating inside the energy router, these early abnormal signals may not be captured in time through conventional threshold detection, leading to the gradual deterioration of the fault. In addition, most of the existing detection methods cannot effectively process and analyze a large amount of operation data, and it is difficult to fully explore the device operation status information hidden behind the data, unable to meet the requirements of the smart grid for high-precision and real-time abnormal detection of energy routers. Summary of the Invention

[0004] To solve the above various problems, the present invention proposes an abnormal detection method for an energy router, which solves the deficiencies of the existing detection means in terms of real-time performance, accuracy, and data processing ability, and realizes precise monitoring of the operation status of the energy router and timely discovery of abnormalities.

[0005] To solve the above technical problems, the technical solution proposed by the present invention is: an abnormal detection method for an energy router, including the following steps:

[0006] Step 1: Data collection. Use sensors deployed at key nodes of the energy router to collect data related to energy transmission and conversion in real time, covering types such as voltage, current, power, and temperature, and perform high-frequency collection to ensure real-time and accurate data.

[0007] Step 2: Data preprocessing. Use a denoising algorithm based on wavelet transform to denoise the original data, and then normalize data with different dimensions to the [0,1] interval for convenient subsequent analysis.

[0008] Step 3: Feature extraction. With the help of deep learning algorithms, by constructing a convolutional neural network with multiple hidden layers, complex features in the preprocessed data are automatically extracted to generate a feature vector representing the operating state of the energy router.

[0009] Step 4: Anomaly judgment. The extracted feature vector is input into a pre-trained anomaly detection model constructed based on the support vector data description algorithm. According to the distance between the feature vector and the normal data region in the model, it is judged whether the energy router is abnormal. If the distance exceeds the set threshold, it is determined as abnormal, and the threshold is obtained from the statistics of a large amount of normal operation data.

[0010] Preferably, during data acquisition, an optical fiber sensor is used to collect temperature, and a Hall sensor is used to collect voltage, current, and power to ensure high-precision measurement.

[0011] Preferably, after denoising in data preprocessing, the linear interpolation method is used to fill in the missing data points to ensure data integrity.

[0012] Preferably, a pooling layer is provided in the hidden layer of the convolutional neural network used for feature extraction, and max pooling is used to reduce the dimension of the feature map, reduce the calculation amount, and retain key features.

[0013] Preferably, after determining an anomaly in anomaly judgment, the system automatically sends an alarm to the monitoring center, and the content includes the time, type, and specific node location of the anomaly occurrence, which is transmitted according to a specific communication protocol through a wireless communication module.

[0014] Preferably, during the training of the anomaly detection model, cross-validation is used. The training data set is divided into multiple subsets, and the subsets are alternately used as the validation set and the training set to select the model with the optimal average performance.

[0015] Preferably, there is a system initialization step before data acquisition to set the parameters of the sensors, acquisition devices, and anomaly detection system to ensure the normal operation of the system.

[0016] Preferably, there is an anomaly cause analysis step after anomaly judgment. After detecting an anomaly, the data before and after the anomaly is analyzed associatively, and combined with the fault tree analysis method, the cause of the anomaly is determined.

[0017] Preferably, the method also includes a model update step. Operational data is regularly collected. When the amount of new data reaches the standard, the anomaly detection model is retrained to update the parameters to adapt to the changes in the operating state.

[0018] The advantages of the present invention compared with the prior art are as follows:

[0019] Improve detection accuracy: By using deep learning algorithms for feature extraction, it can automatically learn complex features in the data. Compared with traditional methods, it can identify abnormal situations more accurately. For example, for complex electrical and thermal faults inside the energy router, it can more accurately determine the type and degree of the faults.

[0020] Enhance real-time performance: Collect data in real-time at high frequency and through a fast algorithm processing flow, it can detect anomalies in a timely manner at the first moment when they occur, greatly shortening the time for anomaly discovery and winning valuable time for taking timely measures to avoid the expansion of faults.

[0021] Improve data processing ability: It can effectively process a large amount of operation data, fully mine the potential information in the data, conduct a comprehensive and in-depth analysis of the operating state of the energy router, and provide strong support for the maintenance and management of the equipment. Brief Description of the Drawings

[0022] Figure 1 It is the principle flowchart of the present invention. Detailed Embodiment

[0023] The present invention will be further described in detail below with reference to the drawings.

[0024] Embodiment 1

[0025] In a regional substation of a smart grid, multiple energy routers are installed. First, at each key node of the energy router, such as input and output ports, power conversion modules, etc., high-precision voltage, current, power sensors and fiber optic temperature sensors are installed. The sensors continuously collect data according to the set high frequency. The collected raw data is quickly transmitted to the data processing unit, where the data is denoised using a denoising algorithm based on wavelet transform to remove the noise generated by electromagnetic interference and other factors. Subsequently, the denoised data is normalized. Then, the preprocessed data is input into the constructed convolutional neural network model, which has been trained with a large amount of historical data and can accurately extract the features in the data. The extracted feature vectors are input into an anomaly detection model based on the support vector data description algorithm. During an operation, the anomaly detection model detects that the distance between the feature vectors of a certain energy router and the normal data area exceeds the set threshold. The system immediately determines that the energy router is abnormal and automatically sends an alarm to the monitoring center, including the time of the anomaly, the type of the anomaly, and the specific node location where the anomaly occurs. After receiving the alarm, the monitoring personnel quickly check the energy router and find that it is due to the aging of a component in a power conversion module that causes the anomaly, and they replace it in time to avoid the further expansion of the fault.

[0026] Embodiment 2

[0027] In a microgrid system with distributed energy access, the energy router undertakes the important task of coordinating the access and distribution of multiple energy sources. When implementing the anomaly detection method of the present invention, the data acquisition link ensures comprehensive coverage of each energy access point and the key parts inside the energy router. During the data preprocessing process, aiming at the unstable data characteristics generated by distributed energy, the parameters of the wavelet transform denoising algorithm are optimized to better adapt to such data. In the feature extraction stage, the convolutional neural network model is adjusted in structure and optimized in parameters according to the characteristics of the operation data of the energy router in the microgrid system, so that it can extract features more effectively. In terms of anomaly judgment, through the statistical analysis of a large amount of normal operation data of the energy router in the microgrid system, the anomaly judgment threshold suitable for this scenario is determined. During one operation, the system detected abnormal fluctuations in the power data of the energy router at a certain moment. After the anomaly detection process, it was accurately judged that due to the photovoltaic array in the distributed energy being blocked by clouds, the input power mutated, which in turn caused the abnormal operation of the energy router. The system timely adjusted the energy distribution strategy to ensure the stable operation of the microgrid.

[0028] Embodiment 3

[0029] In the power supply system of a large commercial complex, there are multiple energy routers supplying power to electrical equipment in different areas. When the anomaly detection method of the present invention is implemented in this scenario, the data acquisition pays attention to the influence of the load characteristics in different areas on the operation data of the energy router, and sensors are reasonably arranged at the key nodes of the energy routers in different areas. In the data preprocessing stage, aiming at the complex and diverse electrical equipment and rich types of data noise generated in the commercial complex, an improved wavelet transform denoising algorithm and a more accurate normalization method are adopted. The convolutional neural network model used for feature extraction is trained for the commercial power supply scenario, and can accurately capture the operation characteristics of the energy router related to commercial power loads. The anomaly judgment model sets a dynamic anomaly judgment threshold through the statistics of the normal operation data of the commercial complex at different times. During a peak evening power consumption period, the system detected an anomaly in an energy router. After analysis, it was found that due to the addition of a large number of high-power lighting devices in a certain area, the load of the energy router was too large. The system timely issued an alarm and adjusted the power supply strategy for this area through interaction with the smart grid to ensure the normal power consumption of the commercial complex.

[0030] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered within the protection scope of the present invention.

Claims

1. An abnormal detection method for an energy router, characterized in that It includes the following steps: Step 1: Data collection. Sensors deployed at key nodes of the energy router are used to collect data related to energy transmission and conversion in real time, covering types such as voltage, current, power, and temperature. High-frequency collection ensures real-time and accurate data. Step 2: Data preprocessing. A denoising algorithm based on wavelet transform is used to denoise the original data, and then data with different dimensions is normalized to the interval [0, 1] for subsequent analysis. Step 3: Feature extraction. With the help of deep learning algorithms, a convolutional neural network with multiple hidden layers is constructed to automatically extract complex features from the preprocessed data, generating a feature vector representing the operating state of the energy router. Step 4: Anomaly judgment. The extracted feature vector is input into an anomaly detection model pre-trained based on the support vector data description algorithm. According to the distance between the feature vector and the normal data region in the model, it is judged whether the energy router is abnormal. If the distance exceeds the set threshold, it is determined as abnormal. The threshold is obtained from the statistics of a large amount of normally operating data.

2. The abnormal detection method of an energy router according to claim 1, characterized in that: During data collection, an optical fiber sensor is used to collect temperature, and Hall sensors are used to collect voltage, current, and power to ensure high-precision measurement.

3. The abnormal detection method of an energy router according to claim 1, characterized in that: After denoising in data preprocessing, the linear interpolation method is used to fill in missing data points to ensure data integrity.

4. The abnormal detection method of an energy router according to claim 1, characterized in that: In the hidden layer of the convolutional neural network used for feature extraction, there is a pooling layer. Max pooling is used to reduce the dimension of the feature map, reducing the computational amount and retaining key features.

5. The abnormal detection method of an energy router according to claim 1, characterized in that: After anomaly judgment determines an anomaly, the system automatically sends an alarm to the monitoring center, including the time, type, and specific node location of the anomaly, which is transmitted through a wireless communication module according to a specific communication protocol.

6. The abnormal detection method of an energy router according to claim 1, characterized in that: During the training of the anomaly detection model, cross-validation is used. The training data set is divided into multiple subsets, and the subsets are alternately used as the validation set and the training set to select the model with the optimal average performance.

7. The abnormal detection method of an energy router according to claim 1, characterized in that: Before data collection, there is a system initialization step to set the parameters of sensors, acquisition devices, and the anomaly detection system to ensure the normal operation of the system.

8. The abnormal detection method of an energy router according to claim 1, wherein: After anomaly judgment, there is an anomaly cause analysis step. After detecting an anomaly, the data before and after the anomaly is analyzed associatively, and combined with the fault tree analysis method, the cause of the anomaly is determined.

9. The abnormal detection method of an energy router according to claim 1, wherein: This method also includes a model update step. Operational data is collected regularly. When the amount of new data reaches the standard, the anomaly detection model is retrained to update the parameters to adapt to changes in the operating state.

Images