Intelligent sensor information fusion analysis method for electric power infrastructure business

Abstract

The invention relates to the technical field of power industry, and provides an intelligent sensor information fusion analysis method for power infrastructure business, which comprises the following steps: a data acquisition step: acquiring data by using a sensor, and transmitting the data to an edge proxy server for connection so as to acquire multi-modal data and transmit the multi-modal data to a data preprocessing module; a data preprocessing step: performing preprocessing such as data cleaning and standardization on the collected original data, and then transmitting the preprocessed data to a modal coding module; and a modal coding step: considering the specific prior preference of each data modal, respectively coding the data, and converting the data into feature codes which can be understood by the model. According to the method, the integrated multi-modal graph attention neural network is utilized to perform intelligent fusion analysis on various sensor data in the electric power infrastructure scene, so that the relevance among the sensor data of different modals can be captured more accurately, and the construction safety of the electric power infrastructure scene is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of the electric power industry, and in particular to an intelligent sensor information fusion analysis method for electric power infrastructure business. Background Art

[0002] During the construction of power transmission and transformation projects in the power industry, it is crucial to ensure the safety of workers. In order to achieve this goal, multiple key projects need to be monitored and managed in real time. For example, whether the construction workers' work behavior complies with the regulations, whether the working conditions of machinery and equipment, including load, tension and inclination, exceed the safety range, and whether the construction environment meteorological conditions such as temperature, humidity, wind speed and air quality including harmful gas concentration affect construction safety.

[0003] By comprehensively applying the above monitoring items, construction safety can be effectively improved and the life and property safety of workers can be ensured. These monitoring measures not only help prevent accidents, but also improve construction efficiency and quality, and promote the healthy development of the power industry;

[0004] In the existing technical solutions, video surveillance images and other sensor data are collected and processed independently, and are not fully combined. First, the correlation between data is not analyzed, and there is a lot of data redundancy; second, each sensor data is analyzed independently. For example, for video surveillance images, only the target detection algorithm detects irregular operation behaviors; for sensors such as tower tension, inclination, and harmful gases, they only monitor whether their values ​​exceed the safe range. Because the algorithms of various sensors are not combined, it is impossible to find the implicit correlation between data and analyze the potential impact on construction safety. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an intelligent sensor information fusion analysis method for power infrastructure business, which solves the problem that the correlation between data is not analyzed and each sensor data is analyzed independently, and the implicit correlation between data cannot be found, and then the potential impact on construction safety cannot be analyzed.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent sensor information fusion analysis method for power infrastructure business, comprising the following steps:

[0007] Data collection step: using sensors to collect data and transmit it to the edge proxy server to connect and pass the collected multimodal data to the data preprocessing module;

[0008] Data preprocessing step: perform preprocessing such as data cleaning and standardization on the collected raw data, and then pass the preprocessed data to the modal encoding module;

[0009] Modality encoding step: Considering the prior preferences specific to each data modality, encode the data separately and convert it into feature encoding that can be understood by the model;

[0010] Input mapping step: Map the encoded multimodal input features to a feature space suitable for graph structure processing, align them in time series, and unify the dimensions of the input features;

[0011] Backbone network reasoning steps: Based on the graph attention neural network, after the multimodal data is mapped to the unified feature space, a multi-head attention mechanism is introduced to update the node status by calculating the different attention weights between the node and the neighboring nodes multiple times;

[0012] Output prediction result step: According to the needs of downstream tasks, the learned node-level features are converted and integrated to accurately predict business goals.

[0013] Preferably, the sensors include cameras, micro-meteorological sensors, inclination sensors, tension sensors, vibration sensors and gas sensors. The cameras use deep learning algorithms to adapt to different scenes and target features and accurately identify various types of target information.

[0014] Preferably, in the input mapping step, if the original feature dimension is high, the dimension is reduced by projection to reduce computational complexity; if the original feature dimension is low, the dimension can be increased to capture more complex patterns.

[0015] Preferably, the data cleaning in the data preprocessing step includes removing obviously erroneous data, outliers, and invalid data caused by sensor failure.

[0016] Preferably, in the modal encoding step, different encoding methods are used for different types of sensor data to retain key information of the data to the greatest extent.

[0017] Preferably, in the backbone network inference step, the number of heads of the multi-head attention mechanism can be adjusted according to the actual data complexity and computing resources.

[0018] An intelligent sensor information fusion analysis system for power infrastructure business, including:

[0019] Data acquisition module: used to establish connections with various sensors through edge proxy servers to collect multimodal data in power infrastructure scenarios;

[0020] Data preprocessing module: performs preprocessing operations such as data cleaning and standardization on the raw data collected by the data acquisition module to provide high-quality data for subsequent modal coding;

[0021] Modality encoding module: Considering the prior preferences specific to each data modality, the preprocessed data is encoded separately and converted into feature encodings that can be understood by the model;

[0022] Input mapping module: maps the multimodal input features output by the modal encoding module to a feature space suitable for graph structure processing, aligns and unifies the feature dimensions in time series, and can adjust the feature dimension size as needed;

[0023] Backbone network reasoning module: Based on the graph attention neural network, after mapping the multimodal data to the unified feature space, a multi-head attention mechanism is introduced to update the node state by calculating the different attention weights between the node and the neighboring nodes multiple times to accurately capture the correlation between different modal data;

[0024] Output prediction module: According to the downstream task requirements, the node-level features learned by the backbone network reasoning module are converted and integrated to accurately predict business goals and output the prediction results.

[0025] Preferably, the sensors include cameras, micrometeorological sensors, inclination sensors, tension sensors, vibration sensors and gas sensors, wherein the cameras are used for target monitoring based on visual images, the micrometeorological sensors are used for environmental parameter measurement, the inclination and tension sensors are used for transmission line construction safety monitoring, the vibration sensors are used for foundation pit vibration monitoring, and the gas sensors are used for gas monitoring at construction sites.

[0026] The present invention provides an intelligent sensor information fusion analysis method for electric power infrastructure business.

[0027] It has the following beneficial effects:

[0028] 1. The present invention uses an integrated multimodal graph attention neural network to perform intelligent fusion analysis on multiple sensor data in power infrastructure scenarios, which can more accurately capture the correlation between sensor data of different modes and improve the construction safety of power infrastructure scenarios.

[0029] 2. The present invention can more accurately capture the correlation between sensor data of different modalities.

[0030] 3. The present invention has good scalability. Different types of sensors required at different infrastructure sites can be quickly deployed under the same algorithm network model, reducing the occupation of edge proxy server resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments 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.

[0033] Example:

[0034] As one aspect of the present invention, please refer to the attached Figure 1 The embodiment of the present invention provides an intelligent sensor information fusion analysis system for power infrastructure business, including:

[0035] Data acquisition module: used to establish connections with various sensors through edge proxy servers to collect multimodal data in power infrastructure scenarios. Sensors include cameras, micro-meteorological sensors, inclination sensors, tension sensors, vibration sensors and gas sensors. Cameras are used for target monitoring based on visual images, micro-meteorological sensors are used for environmental parameter measurement, inclination and tension sensors are used for transmission line construction safety monitoring, vibration sensors are used for foundation pit vibration monitoring, and gas sensors are used for construction site gas monitoring;

[0036] Data preprocessing module: performs preprocessing operations such as data cleaning and standardization on the raw data collected by the data acquisition module to provide high-quality data for subsequent modal coding;

[0037] Modality encoding module: Considering the prior preferences specific to each data modality, the preprocessed data is encoded separately and converted into feature encodings that can be understood by the model;

[0038] Input mapping module: maps the multimodal input features output by the modal encoding module to a feature space suitable for graph structure processing, aligns and unifies the feature dimensions in time series, and can adjust the feature dimension size as needed;

[0039] Backbone network reasoning module: Based on the graph attention neural network, after mapping the multimodal data to the unified feature space, a multi-head attention mechanism is introduced to update the node state by calculating the different attention weights between the node and the neighboring nodes multiple times to accurately capture the correlation between different modal data;

[0040] Output prediction module: According to the downstream task requirements, the node-level features learned by the backbone network reasoning module are converted and integrated to accurately predict business goals and output the prediction results.

[0041] Based on the above-mentioned intelligent sensor information fusion analysis system for electric power infrastructure business, as another aspect of the present invention, an intelligent sensor information fusion analysis method for electric power infrastructure business is proposed, comprising the following steps:

[0042] Data collection step: Use sensors to collect data and transmit it to the edge proxy server to collect multimodal data and pass it to the data preprocessing module. The sensors include cameras, micro-meteorological sensors, tilt sensors, tension sensors, vibration sensors and gas sensors. The camera uses a deep learning algorithm to adapt to different scenes and target features and accurately identify various target information;

[0043] Data preprocessing step: The collected raw data is preprocessed by data cleaning and standardization, and then the preprocessed data is passed to the modal coding module. Data cleaning includes removing obviously erroneous data, outliers, and invalid data caused by sensor failure;

[0044] Modal encoding step: Consider the prior preferences specific to each data modality, encode the data separately, and convert it into feature encoding that the model can understand. Different encoding methods are used for different types of sensor data to retain the key information of the data to the greatest extent;

[0045] Input mapping step: Map the encoded multimodal input features to a feature space suitable for graph structure processing, align them in time series, and unify the dimensions of the input features. If the original feature dimension is high, reduce the dimension through projection to reduce the computational complexity. If the original feature dimension is low, increase the dimension to capture more complex patterns.

[0046] Backbone network reasoning steps: Based on the graph attention neural network, after the multimodal data is mapped to the unified feature space, a multi-head attention mechanism is introduced to update the node state by calculating the different attention weights between the node and the neighboring nodes multiple times. The number of heads of the multi-head attention mechanism can be adjusted according to the actual data complexity and computing resources;

[0047] Output prediction result step: According to the needs of downstream tasks, the learned node-level features are converted and integrated to accurately predict business goals.

[0048] 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. An intelligent sensor information fusion analysis method for power infrastructure business, characterized in that: The following steps are involved: Data collection step: using sensors to collect data and transmit it to the edge proxy server to connect and pass the collected multimodal data to the data preprocessing module; Data preprocessing step: perform preprocessing such as data cleaning and standardization on the collected raw data, and then pass the preprocessed data to the modal encoding module; Modality encoding step: Considering the prior preferences specific to each data modality, encode the data separately and convert it into feature encoding that can be understood by the model; Input mapping step: Map the encoded multimodal input features to a feature space suitable for graph structure processing, align them in time series, and unify the dimensions of the input features; Backbone network reasoning steps: Based on the graph attention neural network, after the multimodal data is mapped to the unified feature space, a multi-head attention mechanism is introduced to update the node status by calculating the different attention weights between the node and the neighboring nodes multiple times; Output prediction result step: According to the needs of downstream tasks, the learned node-level features are converted and integrated to accurately predict business goals.

2. According to claim 1, a smart sensor information fusion analysis method for power infrastructure business is characterized in that: The sensors include cameras, micro-meteorological sensors, inclination sensors, tension sensors, vibration sensors and gas sensors. The cameras use deep learning algorithms to adapt to different scenes and target features and accurately identify various types of target information.

3. The intelligent sensor information fusion analysis method for power infrastructure business according to claim 1 is characterized in that: In the input mapping step, if the original feature dimension is high, the dimension is reduced by projection to reduce the computational complexity; if the original feature dimension is low, the dimension can be increased to capture more complex patterns.

4. The intelligent sensor information fusion analysis method for power infrastructure business according to claim 1 is characterized in that: The data cleaning in the data preprocessing step includes removing obviously erroneous data, outliers, and invalid data caused by sensor failure.

5. The intelligent sensor information fusion analysis method for power infrastructure business according to claim 1 is characterized in that: In the modal encoding step, different encoding methods are used for different types of sensor data to retain key information of the data to the greatest extent.

6. The intelligent sensor information fusion analysis method for power infrastructure business according to claim 1 is characterized in that: In the backbone network inference step, the number of heads of the multi-head attention mechanism can be adjusted according to the actual data complexity and computing resources.

7. An intelligent sensor information fusion analysis system for electric power infrastructure business, using an intelligent sensor information fusion analysis method for electric power infrastructure business as described in any one of claims 1 to 6, characterized in that: include: Data acquisition module: used to establish connections with various sensors through edge proxy servers to collect multimodal data in power infrastructure scenarios; Data preprocessing module: performs preprocessing operations such as data cleaning and standardization on the raw data collected by the data acquisition module to provide high-quality data for subsequent modal coding; Modality encoding module: Considering the prior preferences specific to each data modality, the preprocessed data is encoded separately and converted into feature encodings that can be understood by the model; Input mapping module: maps the multimodal input features output by the modal encoding module to a feature space suitable for graph structure processing, aligns and unifies the feature dimensions in time series, and can adjust the feature dimension size as needed; Backbone network reasoning module: Based on the graph attention neural network, after mapping the multimodal data to the unified feature space, a multi-head attention mechanism is introduced to update the node state by calculating the different attention weights between the node and the neighboring nodes multiple times to accurately capture the correlation between different modal data; Output prediction module: According to the downstream task requirements, the node-level features learned by the backbone network reasoning module are converted and integrated to accurately predict business goals and output the prediction results.

8. The intelligent sensor information fusion analysis system for power infrastructure business according to claim 7 is characterized in that: The sensors include cameras, micro-meteorological sensors, inclination sensors, tension sensors, vibration sensors and gas sensors, wherein the cameras are used for target monitoring based on visual images, the micro-meteorological sensors are used for environmental parameter measurement, the inclination and tension sensors are used for transmission line construction safety monitoring, the vibration sensors are used for foundation pit vibration monitoring, and the gas sensors are used for construction site gas monitoring.

Images