Power early warning system based on laboratory operation data

Abstract

The invention discloses an electric power early warning system based on laboratory operation data, and the system comprises a sensing layer which is provided with a multi-mode sensor network, the multi-mode sensor network comprises a current / voltage transformer used for collecting current and voltage data, an infrared thermal imager used for obtaining distribution of a temperature field on the surface of equipment, meteorological monitoring equipment and an Internet of Things terminal. The edge layer is used for constructing an edge computing terminal, a lightweight deep learning model is built in the edge computing terminal, and the functions of the edge computing terminal comprise data cleaning feature extraction and anomaly detection; according to the platform layer, an electric power Internet of Things cloud platform is constructed, and historical fault data, power grid GIS information and external meteorological / geological data are integrated by the electric power Internet of Things cloud platform; and the application layer comprises a power disturbance feature extraction module, a multi-dimensional early warning strategy module and a graded early warning and decision support module, and is used for providing an early warning mechanism, fault tracing and disposal scheme recommendation. According to the invention, better and more accurate power early warning can be realized.

Description

Technical Field

[0001] The present invention relates to the field of early warning systems, and more particularly to a power early warning system based on laboratory operation data. Background Art

[0002] With the continuous improvement of the intelligence and complexity of the power system, the limitations of traditional power early warning technologies have become increasingly prominent, mainly reflected in the following aspects:

[0003] Insufficient adaptability of fixed-threshold early warning: Traditional systems rely on preset thresholds for alarm, but in actual operation, the load characteristics of equipment, environmental conditions, etc. change dynamically, resulting in the fixed threshold being unable to accurately reflect the real risk and easily causing false alarms or missed alarms.

[0004] One-sidedness of single-parameter monitoring: Existing technologies mostly make judgments based on the data of a single sensor, making it difficult to capture faults caused by the coupling of multiple factors and having a high missed detection rate.

[0005] Limited response speed and real-time performance: Traditional systems rely on centralized processing in the cloud, and there are delays in data transmission and analysis, which cannot meet the millisecond-level disposal requirements for emergency scenarios such as arc faults and voltage transient disturbances. Therefore, a power early warning system based on laboratory operation data is proposed. Summary of the Invention

[0006] The present invention solves the above technical problems through the following technical solutions. The present invention includes:

[0007] Perception layer: A multi-modal sensor network is deployed. The multi-modal sensor network includes current / voltage transformers for collecting current and voltage data, an infrared thermal imager for obtaining the surface temperature field distribution of equipment, meteorological monitoring equipment, and an Internet of Things terminal;

[0008] Edge layer: An edge computing terminal is built and a lightweight deep learning model is built-in. The functions of the edge computing terminal include data cleaning, feature extraction, and anomaly detection;

[0009] Platform layer: A power Internet of Things cloud platform is built. The power Internet of Things cloud platform integrates historical fault data, grid GIS information, and external meteorological / geological data;

[0010] Application layer: It includes a power disturbance feature extraction module, a multi-dimensional early warning strategy module, and a hierarchical early warning and decision support module, which are used to provide an early warning mechanism, fault tracing, and recommended disposal solutions.

[0011] Furthermore, the multi-modal sensor network has self-calibration and self-diagnosis functions, and the process content of its self-calibration and self-diagnosis is as follows:

[0012] The current / voltage transformer automatically calibrates the measurement accuracy regularly. When it detects that its own performance has dropped by more than the preset range or a fault occurs, it generates fault diagnosis information and automatically sends it to the platform layer.

[0013] The infrared thermal imager has the functions of automatic image quality evaluation and calibration. It automatically retrieves standard imaging parameters from a pre-established mapping set according to the ambient light and temperature conditions, and automatically adjusts to the standard imaging parameters. At the same time, it conducts real-time diagnosis on its own hardware and software systems, and promptly discovers and reports faults.

[0014] The meteorological monitoring equipment automatically calibrates the measurement error based on historical meteorological data and real-time measurement values, and can monitor the working status of its own sensors in real time. Once an abnormality is found, it immediately sends feedback information.

[0015] The IoT terminal automatically detects the connection status of its own communication module and sensors, and at the same time has the functions of remote configuration and upgrade, and adjusts its acquisition parameters and frequencies according to actual needs.

[0016] Furthermore, the hierarchical early warning of the hierarchical early warning and decision support module in the application layer is based on the improved exponentially weighted moving average method to implement a three-level early warning mechanism. The content of the three-level early warning mechanism includes:

[0017] Level 1 early warning: Triggered when the equipment load rate exceeds 85% or the temperature deviates from the baseline value by ±15°C. The system issues a level 1 early warning, prompting the operation and maintenance personnel that the equipment status has deviated from the normal range and they need to pay attention to the equipment operation.

[0018] Level 2 early warning: Triggered when a fault arc lasting more than 0.5 seconds is detected or the ice coating thickness reaches 80% of the design value. At this time, a level 2 early warning signal is generated, indicating that there is a potential fault risk in the equipment, and the operation and maintenance personnel need to take timely measures for investigation and handling.

[0019] Level 3 early warning: Triggered when the fusing mechanism is triggered or a geological disaster early warning signal is received. A level 3 early warning signal is issued, the system immediately activates the emergency response plan, and at the same time quickly pushes the early warning information to the mobile terminal of the operation and maintenance personnel.

[0020] Furthermore, the power early warning system dynamically adjusts the early warning threshold according to the actual operation situation of the equipment and environmental changes through a dynamic risk assessment model.

[0021] The dynamic risk assessment model includes a threshold dynamic adjustment algorithm based on the improved EWMA and a multivariate Bayesian network model.

[0022] Furthermore, the calculation process of the threshold dynamic adjustment algorithm of the improved EWMA is as follows:

[0023] Threshold(t) = α * x(t) + (1 - α) * Threshold(t - 1);

[0024] Threshold(t) represents the dynamic warning threshold at the t-th moment;

[0025] x(t) represents the measured value of the device operation parameter at the t-th moment (such as the effective value of current, temperature value);

[0026] α is an imaging parameter, and its value range is [0.2, 0.8]. Its value is jointly determined by the fluctuation degree of the device historical operation data and the real-time environmental parameters (including but not limited to temperature and humidity, wind speed, ice coating thickness);

[0027] Threshold(t - 1) represents the warning threshold at the (t - 1)-th moment.

[0028] The multi-variable Bayesian network model includes:

[0029] Arc feature node: Records the arc duration (unit: second), arc energy (unit: joule);

[0030] Device aging node: Includes the device service life (unit: year), remaining life of the insulating material (unit: hour), and mechanical wear degree (quantified as a value from 0 to 100);

[0031] Environmental factor node: Includes air humidity (%RH), environmental temperature (°C), air pressure (hPa); Quantifies the fire or device failure probability through a conditional probability table, and the probability calculation formula is:

[0032]

[0033] where n is the total number of network nodes, is the contribution weight of each node to the failure probability.

[0034] Furthermore, the power disturbance feature extraction module includes:

[0035] Time-frequency analysis unit based on wavelet transform, which uses the db4 wavelet basis to decompose the voltage waveform into 5 layers, can finely analyze the voltage waveform in the time-frequency domain, and extract the distribution characteristics of different frequency components at different times.

[0036] Convolutional neural network (CNN) classifier, which contains 3 convolutional layers and 2 fully connected layers. The input is a 128-point time-frequency matrix, and the output is the probabilities of fault types such as partial discharge and insulation aging. Through learning and training with a large amount of fault waveform data, this classifier can accurately identify different types of power disturbance faults.

[0037] Furthermore, the multi-dimensional early warning strategy of the multi-dimensional early warning module includes:

[0038] Spatial dimension: Based on GIS grid division, precise positioning of the fault area is achieved. The power grid is divided into different grids according to the geographical space. Through the data of sensors in the perception layer and the analysis of the platform layer, the specific grid area where the fault occurs is determined.

[0039] Time dimension: Combining equipment life cycle data to predict the probability of fault occurrence. According to the service life, cumulative operation time and maintenance record information of the equipment, an equipment life cycle model is established to predict the possibility of the equipment failing at different time points.

[0040] Environmental dimension: Introducing meteorological satellite data to predict the impact of extreme weather on the power grid. Using the real-time meteorological information provided by meteorological satellites, such as typhoon paths and rainstorm ranges, to evaluate in advance the impact of extreme weather on power grid equipment.

[0041] The present invention has the following advantages compared with the prior art: This power early warning system based on laboratory operation data realizes comprehensive monitoring and early warning capabilities. It integrates equipment operation parameters, environmental data and historical fault information to achieve three-dimensional monitoring of the power system, effectively capturing early fault signs that are easily missed by traditional single-parameter monitoring. It automatically adjusts the early warning threshold according to the real-time operation state of the equipment and environmental changes, avoiding false alarms or missed alarms caused by fixed thresholds and improving the accuracy of early warning.

[0042] Integrating external data such as meteorology and geology to evaluate in advance the impact of extreme weather or geological disasters on the power grid, realizing multi-disaster coupling early warning and enhancing the robustness of the system.

[0043] Combining the equipment aging model with energy efficiency analysis to optimize the maintenance strategy, extend the service life of the equipment, reduce the full-cycle operation and maintenance cost, making this system more worthy of popularization and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is the system block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] The following details the embodiments of the present invention. These embodiments are implemented on the premise of the technical solution of the present invention, and detailed implementation manners and specific operation processes are given. However, the protection scope of the present invention is not limited to the following embodiments.

[0046] As Figure 1 shown, this embodiment provides a technical solution: A power early warning system based on laboratory operation data, including:

[0047] The perception layer deploys a multi-modal sensor network, including current / voltage transformers, infrared thermal imagers, meteorological monitoring equipment, and Internet of Things terminals, which are used to collect device operation parameters (such as three-phase current effective value, voltage effective value, harmonic components), environmental data (temperature and humidity, wind speed, ice coating thickness, rainfall), and disturbance waveform data (such as voltage transient waveform, current mutation signal) in real time.

[0048] The edge layer constructs an edge computing terminal based on an ARM chip, with a lightweight deep learning model built-in, to achieve data cleaning (removing noise and outliers), feature extraction (such as arc features, temperature change features), and anomaly detection (such as dynamic threshold calculation using the moving average method), and caches the original data and processing results of the most recent 72 hours.

[0049] The platform layer constructs a power Internet of Things cloud platform, integrates historical fault data, power grid GIS information, and external meteorological / geological data, and realizes multi-source data fusion analysis through a high-performance computing cluster, supporting the virtual scene mapping of the power grid three-dimensional GIS model and real-time operation data.

[0050] The application layer develops a visual warning interface, integrating the following core modules: Power disturbance feature extraction module: Analyze the transient disturbance of voltage / current waveforms based on wavelet transform and convolutional neural network to identify early fault features such as partial discharge and insulation aging;

[0051] Multi-dimensional warning strategy module: Combine three-dimensional information of space (GIS grid division), time (equipment life cycle data), and environment (meteorological satellite data) to achieve fault warning;

[0052] Hierarchical warning and decision support module: Provide a three-level warning mechanism, fault tracing, and recommended disposal plan functions, and support multi-terminal real-time interaction.

[0053] The multi-modal sensor network has self-calibration and self-diagnosis functions, and the process content of its self-calibration and self-diagnosis is as follows:

[0054] The current / voltage transformer regularly automatically calibrates the measurement accuracy. When it detects that its own performance has dropped by more than the preset amplitude or a fault occurs, it generates fault diagnosis information and automatically sends it to the platform layer;

[0055] The infrared thermal imager has an automatic image quality evaluation and calibration function. It automatically retrieves standard imaging parameters from a pre-established mapping set according to environmental light and temperature conditions, and automatically adjusts to the standard imaging parameters. At the same time, it conducts real-time diagnosis on its own hardware and software systems, and promptly discovers and reports faults;

[0056] The meteorological monitoring equipment automatically calibrates the measurement error based on historical meteorological data and real-time measurement values, and can monitor the working status of its own sensors in real time. When an anomaly is found, it immediately sends feedback information.

[0057] The IoT terminal automatically detects the connection status between its communication module and sensors, and also has the functions of remote configuration and upgrade. It can adjust its acquisition parameters and frequencies according to actual needs.

[0058] The hierarchical early warning of the hierarchical early warning and decision support module at the application layer is based on an improved exponentially weighted moving average method to implement a three-level early warning mechanism. The content of the three-level early warning mechanism includes:

[0059] Level 1 early warning: Triggered when the device load rate exceeds 85% or the temperature deviates from the baseline value by ±15°C. The system will issue a level 1 early warning, prompting the operation and maintenance personnel that the device status has deviated from the normal range and they need to pay attention to the device operation.

[0060] Level 2 early warning: Triggered when a fault arc lasting more than 0.5 seconds is detected or the ice coating thickness reaches 80% of the design value. At this time, a level 2 early warning signal is generated, indicating that there is a potential fault risk in the device, and the operation and maintenance personnel need to take measures to conduct inspections and handling in a timely manner.

[0061] Level 3 early warning: Triggered when the fusing mechanism is triggered or a geological disaster early warning signal is received. A level 3 early warning signal is issued, and the system will immediately start the emergency response plan, and at the same time quickly push the early warning information to the mobile terminal of the operation and maintenance personnel.

[0062] The power early warning system dynamically adjusts the early warning threshold according to the actual operation conditions of the device and environmental changes through a dynamic risk assessment model;

[0063] The dynamic risk assessment model includes a threshold dynamic adjustment algorithm based on an improved EWMA and a multivariate Bayesian network model.

[0064] The calculation process of the threshold dynamic adjustment algorithm of the improved EWMA is as follows:

[0065] Threshold(t) = α * x(t) + (1 - α) * Threshold(t - 1);

[0066] Threshold(t) represents the dynamic early warning threshold at the t-th moment;

[0067] x(t) represents the measured value of the device operation parameter at the t-th moment (such as the effective value of current, temperature value);

[0068] α is an influence parameter, and its value range is [0.2, 0.8]. Its value is jointly determined by the fluctuation degree of the device historical operation data and real-time environmental parameters (including but not limited to temperature and humidity, wind speed, ice coating thickness);

[0069] Threshold(t - 1) represents the early warning threshold at the (t - 1)-th moment.

[0070] The multi-variable Bayesian network model includes:

[0071] Arc feature node: Records the arc duration (unit: seconds) and arc energy (unit: joules);

[0072] Equipment aging node: Includes the service life of the equipment (unit: years), the remaining life of the insulation material (unit: hours), and the degree of mechanical wear (quantified as a value from 0 to 100);

[0073] Environmental factor node: Includes air humidity (%RH), environmental temperature (°C), and air pressure (hPa); Quantifies the probability of fire or equipment failure through a conditional probability table, and the probability calculation formula is:

[0074]

[0075] where n is the total number of network nodes, is the contribution weight of each node to the failure probability.

[0076] Furthermore, the power disturbance feature extraction module includes:

[0077] Time-frequency analysis unit based on wavelet transform, which uses the db4 wavelet basis to decompose the voltage waveform into 5 layers, can finely analyze the voltage waveform in the time-frequency domain, and extract the distribution characteristics of different frequency components at different times.

[0078] Convolutional neural network (CNN) classifier, which contains 3 convolutional layers and 2 fully connected layers, with the input being a 128-point time-frequency matrix and the output being the probabilities of fault types such as partial discharge and insulation aging. Through learning and training on a large amount of fault waveform data, this classifier can accurately identify different types of power disturbance faults.

[0079] The multi-dimensional early warning strategy of the multi-dimensional early warning module includes:

[0080] Spatial dimension: Achieves precise positioning of the fault area based on GIS grid division. The power grid is divided into different grids according to the geographical space, and through the data of the sensor layer and the analysis of the platform layer, the specific grid area where the fault occurs is determined;

[0081] Time dimension: Combines the equipment life cycle data to predict the probability of fault occurrence. According to the service life of the equipment, the cumulative operation time, and the maintenance record information, an equipment life cycle model is established to predict the possibility of the equipment failing at different time points;

[0082] Environmental dimension: Introduces meteorological satellite data to predict the impact of extreme weather on the power grid. Utilizes the real-time meteorological information provided by meteorological satellites, such as typhoon paths and rainstorm ranges, to evaluate in advance the impact of extreme weather on power grid equipment.

[0083] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0084] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0085] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A power early warning system based on laboratory operation data, characterized in that: include: Perception layer: deploys a multimodal sensor network, which includes current / voltage transformers, infrared thermal imagers, meteorological monitoring equipment, and IoT terminals; Edge layer: Build edge computing terminals with built-in lightweight deep learning models. The functions of the edge computing terminals include data cleaning feature extraction and anomaly detection. Platform layer: A power IoT cloud platform was built, which integrates historical fault data, power grid GIS information, and external meteorological / geological data; Application layer: includes power disturbance feature extraction module, multi-dimensional early warning strategy module and hierarchical early warning and decision support module, which are used to provide early warning mechanism, fault tracing and disposal plan recommendation.

2. The power early warning system based on laboratory operation data according to claim 1 is characterized by: The multimodal sensor network has self-calibration and self-diagnosis functions, and the process of self-calibration and self-diagnosis is as follows: The current / voltage transformer automatically calibrates the measurement accuracy regularly. When it detects that its performance has dropped by more than a preset amount or a fault occurs, fault diagnosis information is generated and automatically sent to the platform layer. The infrared thermal imager has the function of automatic image quality assessment and calibration. It automatically retrieves standard imaging parameters from a pre-established mapping set according to ambient light and temperature conditions, and automatically adjusts to the standard imaging parameters. It also performs real-time diagnosis of its own hardware and software systems to promptly detect and report faults. Meteorological monitoring equipment automatically calibrates measurement errors based on historical meteorological data and real-time measurement values, and can monitor the working status of its own sensors in real time, and send feedback information immediately when an abnormality is found; The IoT terminal automatically detects the connection status between its own communication module and the sensor, and also has remote configuration and upgrade functions, and can adjust its collection parameters and frequency according to actual needs.

3. The power early warning system based on laboratory operation data according to claim 1 is characterized in that: The hierarchical warning of the application layer and the hierarchical warning of the decision support module implement a three-level warning mechanism based on the improved exponentially weighted moving average method. The content of the three-level warning mechanism includes: Level 1 warning: When the equipment load rate exceeds the warning value or the temperature deviates from the preset range of the baseline value, the system will issue a level 1 warning to remind the operation and maintenance personnel that the equipment status has deviated from the normal range and they need to pay attention to the equipment operation; Level 2 warning: When a fault arc is detected that lasts for more than a preset time or the ice thickness reaches a threshold, a level 2 warning signal is generated, indicating that the equipment has a potential fault risk and the operation and maintenance personnel need to take timely measures to investigate and handle it; Level 3 warning: When the fuse mechanism is triggered or a geological disaster warning signal is received, a level 3 warning signal is issued. The system immediately activates the emergency response plan and quickly pushes the warning information to the mobile terminal of the operation and maintenance personnel.

4. The power early warning system based on laboratory operation data according to claim 1 is characterized in that: The power early warning system dynamically adjusts the early warning threshold according to the actual operation of the equipment and environmental changes through a dynamic risk assessment model; The dynamic risk assessment model includes a threshold dynamic adjustment algorithm based on improved EWMA and a multivariate Bayesian network model.

5. The power early warning system based on laboratory operation data according to claim 4 is characterized in that: The calculation process of the threshold dynamic adjustment algorithm of the improved EWMA is as follows: Threshold(t)=α*x(t)+(1-α)*Threshold(t-1); Threshold(t) represents the dynamic warning threshold at time t; x(t) represents the measured value of the equipment operating parameter at time t; α is an image parameter, with a value range of [0.2, 0.8]. Its value is determined by the fluctuation degree of the equipment's historical operation data and the real-time environmental parameters; Threshold(t-1) represents the warning threshold at time t-1; Multivariate Bayesian network models include: Arc feature node: record arc duration and arc energy; Equipment aging node: including equipment service life, remaining life of insulation materials and degree of mechanical wear; Environmental factor nodes: including air humidity, ambient temperature, and air pressure; The probability of fire or equipment failure is quantified by the conditional probability table, and the probability calculation formula is: Where n is the total number of network nodes, is the contribution weight of each node to the failure probability.

6. The power early warning system based on laboratory operation data according to claim 1 is characterized by: The power disturbance feature extraction module comprises: The time-frequency analysis unit based on wavelet transform uses db4 wavelet basis to decompose the voltage waveform into 5 layers, which can make a fine analysis of the voltage waveform in the time-frequency domain and extract the distribution characteristics of different frequency components at different times; Convolutional neural network (CNN) classifier, which includes 3 convolutional layers and 2 fully connected layers, takes a 128-point time-frequency matrix as input and outputs the probability of partial discharge and insulation aging fault types; The classifier identifies different types of power disturbance faults by learning and training a large amount of fault waveform data.

7. The power early warning system based on laboratory operation data according to claim 1 is characterized by: The multi-dimensional warning strategy of the multi-dimensional warning module includes: Spatial dimension: Accurately locate the fault area based on GIS grid division. Divide the power grid into different grids according to geographic space. Determine the specific grid area where the fault occurs through data from sensors at the perception layer and analysis at the platform layer. Time dimension: Combine equipment life cycle data to predict the probability of failure. According to the equipment's service life, cumulative operating time and maintenance record information, establish an equipment life cycle model to predict the possibility of equipment failure at different time points. Environmental dimension: Introduce meteorological satellite data to predict the impact of extreme weather on the power grid, and use the real-time meteorological information provided by meteorological satellites to assess the impact of extreme weather on power grid equipment in advance.

Images