Pole tower corrosion state monitoring system and method thereof

Abstract

The invention discloses a tower corrosion state monitoring system and method. The tower corrosion state monitoring system comprises a sensor module, a data acquisition and transmission module, a data processing and analysis module, an early warning and decision module and a visual interface module. The system has the advantages that the corrosion state and the inclination condition of the tower are comprehensively monitored from different angles through multi-sensor fusion, the accuracy and the reliability of the monitoring system are improved through the multi-sensor fusion technology, stable transmission and real-time monitoring of data are achieved in combination with the low-power wide area network (LPWAN) technology and the 5G technology, and the system is suitable for popularization and application. The method is advantaged in that different environmental conditions and monitoring requirements are adapted, flexibility and adaptability of the system are improved, deep learning algorithms such as a convolutional neural network (CNN), a recurrent neural network (RNN) and variants thereof and a multi-modal fusion model are utilized to carry out deep analysis and processing on monitoring data, the intelligent level of the monitoring system is improved, and the monitoring system has a wide application prospect. And the corrosion state, the inclination condition and the development trend of the tower can be accurately evaluated.

Description

Technical Field

[0001] The present invention relates to the technical field of power facility monitoring, and in particular to a system and method for monitoring the corrosion status of a tower. Background Art

[0002] Traditional tower corrosion monitoring methods mainly rely on manual inspections and regular testing. Manual inspections usually require a lot of manpower, time and cost, and due to the large number of towers and their wide distribution, it is difficult to achieve full coverage, and it is easy to miss inspections. Although some existing tower corrosion monitoring systems have improved the shortcomings of traditional methods to a certain extent, there are still many problems. On the one hand, some monitoring systems have a single technology and rely only on a certain type of sensor for monitoring, which cannot fully and accurately reflect the corrosion status of the tower. On the other hand, unstable data transmission technology is also a common problem. The data transmission method used by some monitoring systems is easily affected by the environment, resulting in data loss or transmission delays, which affects the real-time and accuracy of monitoring. Some monitoring systems only use simple statistical analysis methods, which cannot deeply explore the potential laws and characteristics in the data, and it is difficult to make accurate judgments on complex corrosion situations. Summary of the invention

[0003] The present invention provides a system that can monitor the corrosion status of poles and towers in real time and accurately, and also has powerful data analysis and processing capabilities, which can provide scientific decision support for maintenance personnel. At the same time, the monitoring system should also have good adaptability and be able to operate stably under different environmental conditions and pole tower types to meet the increasingly complex monitoring needs of power facilities. The specific plan is as follows:

[0004] A tower corrosion status monitoring system and method thereof, comprising a sensor module, a data acquisition and transmission module, a data processing and analysis module, an early warning and decision module and a visual interface module;

[0005] The sensor module is used to collect data reflecting the corrosion status of the tower in real time from different angles, including: electrochemical sensor, intelligent temperature and humidity sensor, stress and strain sensor, image sensor;

[0006] Electrochemical sensors determine the corrosion status of the tower by measuring the corrosion current and potential between the tower metal and the surrounding environment;

[0007] Intelligent temperature and humidity sensors use humidity sensing elements and temperature measuring elements to measure the temperature and humidity in the environment;

[0008] Stress-strain sensors work by means of strain gauges or fiber gratings. Strain gauges are sensors based on the resistance strain effect. When the tower is subjected to external force, the strain gauge will deform, causing its resistance value to change. Fiber grating sensors use the sensitivity of the grating in the optical fiber to external stress. When the external stress changes, the wavelength of the grating will change.

[0009] The image sensor obtains the surface condition information of the tower by taking images of the tower's appearance.

[0010] The data acquisition and transmission module adopts a combination of low-power wide area network (LPWAN) technology and 5G technology to ensure stable data transmission and real-time performance.

[0011] The data acquisition and transmission module is responsible for receiving and sending data in the system, and specifically includes: data acquisition terminal and communication transmission module: the data acquisition terminal is responsible for receiving data from each sensor, and performing preliminary processing and storage. The data acquisition terminal can be installed on the tower or in a nearby distribution box;

[0012] The data processing and analysis module specifically includes: data storage module and data analysis module; the data storage module uses MySQL, which is suitable for the storage and query of structured data, and also uses the non-relational database MongoDB, which is suitable for large-scale, high-concurrency data storage and processing;

[0013] Data storage adopts distributed storage mode, storing data on multiple nodes;

[0014] The data analysis module uses a convolutional neural network (CNN) to process image data to identify the corrosion area and degree of the tower. It also uses a multimodal fusion model to fuse different types of data to improve the accuracy and reliability of the monitoring system.

[0015] Preprocessing

[0016] Before data fusion, the data of each modality needs to be preprocessed to ensure its validity and consistency;

[0017] Data collection:

[0018] Electrochemical sensor data is represented as: (X_{\text{electro}}=[V,I]), where (V) is voltage and (I) is current;

[0019] The data of the intelligent temperature and humidity sensor is expressed as: (X_{\text{env}}=[T,H]), where (T) is temperature and (H) is humidity.

[0020] The stress-strain sensor data is expressed as: (X_{\text{strain}}=[\sigma,

[0021] \epsilon]), where (\sigma) is stress and (\epsilon) is strain;

[0022] Image sensor data: (X_{\text{image}}=\text{RGB image})

[0023] Denoising: Filtering the sensor data, including mean filtering or Kalman filtering;

[0024] Normalization: The data normalization formula is:

[0025] [X_i^{norm}=\frac{X_i-\min(X_i)}{\max(X_i)-\min(X_i)},]

[0026] Among them, (X_i) represents the original data, (\min(X_i)) and (\max(X_i)) are the minimum and maximum values ​​of the data respectively.

[0027] Interpolation: Interpolate missing data to ensure data continuity.

[0028] Multimodal Data Fusion Strategies In the multimodal fusion model, fusion strategies include feature-level fusion, decision-level fusion, and hybrid-level fusion;

[0029] The early warning and decision-making module evaluates and predicts the corrosion status based on the output of the data processing and analysis module; when potential corrosion risks are detected, the system will automatically generate early warning information;

[0030] The visual interface module displays real-time monitoring data, analysis results and warning information for users;

[0031] Furthermore, feature extraction and fusion are performed on preprocessed multimodal data using different deep learning models:

[0032] Image data ((I_{img})): Use convolutional neural network (CNN) to extract local features and corrosion area information of the image;

[0033] Electrochemical data ((E_{ec},I_{ec})): Use multi-layer perceptron (MLP) or recurrent neural network (RNN) to analyze time series trends;

[0034] Temperature and humidity data ((T_{env},H_{env})): Use MLP for simple regression analysis;

[0035] Stress-strain data ((S_{strain},\epsilon_{strain})): Use MLP or support vector machine (SVM) to model the tower structure changes;

[0036] Each data modality is characterized by extracting features through the above model to obtain a feature vector:

[0037] [\mathbf{f}{img},\mathbf{f}{ec},\mathbf{f}{env},\mathbf{f}{strain}]

[0038] Feature-level fusion

[0039] The feature vectors of different data modalities are fused to form a comprehensive feature vector (\mathbf{f}_{fusion}), which can be expressed as:

[0040] [\mathbf{f}{fusion}=f{concat}(\mathbf{f}{img},\mathbf{f}{ec},

[0041] \mathbf{f}{env},\mathbf{f}{strain}),]

[0042] Where (f_{concat}) represents the feature connection function (such as vector concatenation or weighted summation).

[0043] Furthermore, the multimodal data fusion model training inputs the fused feature vector (\mathbf{f}{fusion}) into the deep learning model (such as multi-layer perceptron, convolutional neural network, etc.) for training. Assume that the loss function is the cross entropy loss (\mathcal{L}{CE}):

[0044] [\mathcal{L}{CE}=-\sum{c=1}^{C}y_c\log(\hat{y}_c),]

[0045] Among them, (y_c) is the actual category label, (\hat{y}_c) is the predicted probability, and (C) is the total number of categories.

[0046] Update the model parameters (\theta) using the Adam optimizer:

[0047] [\mathbf{m}t=\beta_1\mathbf{m}{t-1}+(1-\beta_1)

[0048] \nabla_\theta\mathcal{L}t,]

[0049] [\mathbf{v}t=\beta_2\mathbf{v}{t-1}+(1-\beta_2)

[0050] (\nabla\theta\mathcal{L}_t)^2,]

[0051] [\hat{\mathbf{m}}_t=\frac{\mathbf{m}_t}{1-\beta_1^t},

[0052] \quad\hat{\mathbf{v}}_t=\frac{\mathbf{v}t}{1-\beta_2^t},]

[0053] [\theta{t+1}=\theta_t-

[0054] \eta\frac{\hat{\mathbf{m}}_t}{\sqrt{\hat{\mathbf{v}}_t}+\epsilon}.]

[0055] Among them, (\mathbf{m}_t) and (\mathbf{v}_t) are the first-order and second-order moment estimates of the gradient, (\beta_1,\beta_2) are exponential decay rates, (\eta) is the learning rate, and (\epsilon) is a numerical stability constant;

[0056] Decision-making and early warning

[0057] After the training is completed, the newly collected multimodal data is input into the trained model to output the prediction result of the corrosion state of the tower (\hat{y}). According to the warning threshold judgment result, if the corrosion state reaches the warning standard, the warning module is triggered to notify the relevant maintenance personnel.

[0058] Explanation of main parameters 1. (E_{ec}, I_{ec}): Potential and current data measured by electrochemical sensors, reflecting metal corrosion current and corrosion rate.

[0059] (T_{env},H_{env}): Ambient temperature and humidity, which affect the corrosion rate of metals.

[0060] (S_{strain},\epsilon_{strain}): stress and strain data used to evaluate the load condition of the tower structure.

[0061] (I_{img}): Image data, containing visual information of the tower surface, used to detect corrosion areas and extent.

[0062] (\mathbf{f}{img},\mathbf{f}{ec},\mathbf{f}{env},

[0063] \mathbf{f}{strain}): Feature vectors extracted from image, electrochemical, environmental and stress data, respectively.

[0064] (\mathbf{f}_{fusion}): Fusion feature vector, integrating information from all data modalities.

[0065] (\mathca l{L}_{CE}): Cross entropy loss function, used for training multimodal fusion models.

[0066] Here we selected convolutional neural network (CNN) for image data processing, and used the feature vector after multimodal data fusion as the model input to predict the corrosion state of the tower.

[0067] Model structure design The structure of the model includes the following layers:

[0068] Input layer: receives the feature vector after multimodal fusion (\mathbf{f}_{\text{fusion}}).

[0069] Hidden layer: multi-layer fully connected layer, with ReLU as the activation function, used to extract high-level features.

[0070] Output layer: outputs the prediction result of the corrosion state, which is a binary classification of corrosion / no corrosion. The following is the application process of the multimodal fusion model in this system;

[0071] Furthermore, the early warning module can set different early warning levels according to the corrosion status and development trend of the tower, including mild warning, moderate warning and severe warning, which correspond to different degrees of corrosion and risk respectively; when the monitoring system detects that the corrosion status of the tower reaches the corresponding early warning level, it will automatically issue an early warning signal to remind relevant personnel to take measures.

[0072] Early warning method selection: There are many ways to warn, such as SMS, email, sound and light alarm. According to different user needs and scenarios, choose the appropriate early warning method to ensure that the early warning information can be conveyed to relevant personnel in a timely and accurate manner. At the same time, pay attention to the content and format of the early warning information to make it easy to understand and process.

[0073] Furthermore, the user interface module uses charts, reports, and maps to visualize the monitoring data; bar charts, line charts, pie charts and other charts can be used to display different types of data, such as corrosion current, temperature and humidity, stress and strain, etc.; reports are used to display detailed data information and analysis results; maps are used to mark the location of each tower, and different colors are used to represent different corrosion states; through visual display, users can intuitively understand the corrosion state and development trend of the tower.

[0074] Real-time monitoring interface: Design a real-time monitoring interface through which users can view the data and image information of each sensor in real time. A large screen display or multi-window display can be used to facilitate users to view the monitoring data of multiple towers at the same time. In the real-time monitoring interface, an early warning function should be set up so that users can be reminded in time when the monitoring system issues an early warning signal.

[0075] Parameter setting and management

[0076] User authority management: Set different levels of user authority to ensure the security and reliability of the system. It can be divided into multiple levels such as administrators, operators and ordinary users, with different operation permissions and data access permissions. Administrators can comprehensively manage and maintain the system, operators can perform operations such as data collection and analysis, and ordinary users can view monitoring data and early warning information.

[0077] System parameter settings: Users can set and manage the parameters of the monitoring system through the user interface module. You can set the sensor's acquisition frequency, warning threshold, communication parameters, etc., and you can also adjust and optimize the parameters of the deep learning algorithm. When setting parameters, pay attention to the rationality and effectiveness of the parameters to ensure the normal operation of the system.

[0078] A method for using a tower corrosion condition monitoring system, the steps are as follows:

[0079] S1. Sensor data collection

[0080] Each sensor collects tower data in real time at a preset collection frequency, including corrosion current, potential, temperature and humidity, stress and strain, image and tower tilt angle;

[0081] S2. The data acquisition terminal receives the data from the sensor and performs preliminary processing and storage;

[0082] S3. Data transfer:

[0083] 1) The data acquisition terminal transmits the processed data to the data processing module through the communication network;

[0084] 2) The data processing module stores the received data in a database;

[0085] S4. Use deep learning algorithms to analyze and process data, extract effective features, and evaluate the corrosion status, tilt and development trend of towers;

[0086] Based on the analysis results, determine whether the corrosion status and tilt of the tower have reached the warning level.

[0087] If the warning level is reached, a corresponding warning signal will be issued and a corresponding maintenance plan will be formulated.

[0088] S5 user interface display, which visualizes the monitoring data and analysis results through the user interface module;

[0089] Users can view the corrosion status, tilt, warning information and maintenance plan of the tower through the user interface module, and can also set and manage system parameters.

[0090] The advantage of the present invention is that multi-sensor fusion comprehensively monitors the corrosion state and tilt of the tower from different angles. Through multi-sensor fusion technology, the accuracy and reliability of the monitoring system are improved. In combination with low-power wide area network (LPWAN) technology and 5G technology, stable data transmission and real-time monitoring are achieved, and different environmental conditions and monitoring requirements are adapted to improve the flexibility and adaptability of the system. Deep learning algorithms such as convolutional neural network (CNN), recurrent neural network (RNN) and its variants and multimodal fusion models are used to deeply analyze and process the monitoring data, improve the intelligence level of the monitoring system, accurately evaluate the corrosion state, tilt and development trend of the tower, establish a complete early warning mechanism, issue early warning signals in time, remind relevant personnel to take measures, provide scientific decision-making support, formulate reasonable maintenance plans, reduce maintenance costs, and improve the reliability and safety of the power system. Intuitive user interface, design a friendly user interface, display monitoring data and analysis results in the form of intuitive charts, reports, maps, etc., convenient for users to view and manage, and improve user experience. DETAILED DESCRIPTION

[0091] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0092] The preferred embodiments of the present invention are described below with reference to the present invention. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0093] A tower corrosion status monitoring system and method thereof, comprising a sensor module, a data acquisition and transmission module, a data processing and analysis module, an early warning and decision module and a visual interface module;

[0094] The sensor module is used to collect data reflecting the corrosion status of the tower in real time from different angles, including: electrochemical sensor, intelligent temperature and humidity sensor, stress and strain sensor, image sensor;

[0095] Electrochemical sensors determine the corrosion status of the tower by measuring the corrosion current and potential between the tower metal and the surrounding environment;

[0096] Intelligent temperature and humidity sensors use humidity sensing elements and temperature measuring elements to measure the temperature and humidity in the environment;

[0097] Stress-strain sensors work by means of strain gauges or fiber gratings. Strain gauges are sensors based on the resistance strain effect. When the tower is subjected to external force, the strain gauge will deform, causing its resistance value to change. Fiber grating sensors use the sensitivity of the grating in the optical fiber to external stress. When the external stress changes, the wavelength of the grating will change.

[0098] The image sensor obtains the surface condition information of the tower by taking images of the tower's appearance.

[0099] The data acquisition and transmission module adopts a combination of low-power wide area network (LPWAN) technology and 5G technology to ensure stable data transmission and real-time performance.

[0100] The data acquisition and transmission module is responsible for receiving and sending data in the system, and specifically includes: data acquisition terminal and communication transmission module: the data acquisition terminal is responsible for receiving data from each sensor, and performing preliminary processing and storage. The data acquisition terminal can be installed on the tower or in a nearby distribution box;

[0101] The data processing and analysis module specifically includes: data storage module and data analysis module; the data storage module uses MySQL, which is suitable for the storage and query of structured data, and also uses the non-relational database MongoDB, which is suitable for large-scale, high-concurrency data storage and processing;

[0102] Data storage adopts distributed storage mode, storing data on multiple nodes;

[0103] The data analysis module uses a convolutional neural network (CNN) to process image data to identify the corrosion area and degree of the tower. It also uses a multimodal fusion model to fuse different types of data to improve the accuracy and reliability of the monitoring system.

[0104] Preprocessing

[0105] Before data fusion, the data of each modality needs to be preprocessed to ensure its validity and consistency;

[0106] Data collection:

[0107] Electrochemical sensor data is represented as: (X_{\text{electro}}=[V,I]), where (V) is voltage and (I) is current;

[0108] The data of the intelligent temperature and humidity sensor is expressed as: (X_{\text{env}}=[T,H]), where (T) is temperature and (H) is humidity.

[0109] The stress-strain sensor data is expressed as: (X_{\text{strain}}=[\sigma,

[0110] \epsilon]), where (\sigma) is stress and (\epsilon) is strain;

[0111] Image sensor data: (X_{\text{image}}=\text{RGB image})

[0112] Denoising: Filtering the sensor data, including mean filtering or Kalman filtering;

[0113] Normalization: The data normalization formula is:

[0114] [X_i^{norm}=\frac{X_i-\min(X_i)}{\max(X_i)-\min(X_i)},]

[0115] Among them, (X_i) represents the original data, (\min(X_i)) and (\max(X_i)) are the minimum and maximum values ​​of the data respectively.

[0116] Interpolation: Interpolate missing data to ensure data continuity.

[0117] Multimodal Data Fusion Strategies In the multimodal fusion model, fusion strategies include feature-level fusion, decision-level fusion, and hybrid-level fusion;

[0118] The early warning and decision-making module evaluates and predicts the corrosion status based on the output of the data processing and analysis module; when potential corrosion risks are detected, the system will automatically generate early warning information;

[0119] The visual interface module displays real-time monitoring data, analysis results and warning information for users;

[0120] Furthermore, feature extraction and fusion are performed on preprocessed multimodal data using different deep learning models:

[0121] Image data ((I_{img})): Use convolutional neural network (CNN) to extract local features and corrosion area information of the image;

[0122] Electrochemical data ((E_{ec},I_{ec})): Use multi-layer perceptron (MLP) or recurrent neural network (RNN) to analyze time series trends;

[0123] Temperature and humidity data ((T_{env},H_{env})): Use MLP for simple regression analysis;

[0124] Stress-strain data ((S_{strain},\epsilon_{strain})): Use MLP or support vector machine (SVM) to model the tower structure changes;

[0125] Each data modality is characterized by extracting features through the above model to obtain a feature vector:

[0126] [\mathbf{f}{img},\mathbf{f}{ec},\mathbf{f}{env},\mathbf{f}{strain}]

[0127] Feature-level fusion

[0128] The feature vectors of different data modalities are fused to form a comprehensive feature vector (\mathbf{f}_{fusion}), which can be expressed as:

[0129] [\mathbf{f}{fusion}=f{concat}(\mathbf{f}{img},\mathbf{f}{ec},

[0130] \mathbf{f}{env},\mathbf{f}{strain}),]

[0131] Where (f_{concat}) represents the feature connection function (such as vector concatenation or weighted summation).

[0132] Furthermore, the multimodal data fusion model training inputs the fused feature vector (\mathbf{f}{fusion}) into the deep learning model (such as multi-layer perceptron, convolutional neural network, etc.) for training. Assume that the loss function is the cross entropy loss (\mathcal{L}{CE}):

[0133] [\mathcal{L}{CE}=-\sum{c=1}^{C}y_c\log(\hat{y}_c),]

[0134] Among them, (y_c) is the actual category label, (\hat{y}_c) is the predicted probability, and (C) is the total number of categories.

[0135] Update the model parameters (\theta) using the Adam optimizer:

[0136] [\mathbf{m}t=\beta_1\mathbf{m}{t-1}+(1-\beta_1)

[0137] \nabla_\theta\mathcal{L}t,]

[0138] [\mathbf{v}t=\beta_2\mathbf{v}{t-1}+(1-\beta_2)

[0139] (\nabla\theta\mathcal{L}_t)^2,]

[0140] [\hat{\mathbf{m}}_t=\frac{\mathbf{m}_t}{1-\beta_1^t},

[0141] \quad\hat{\mathbf{v}}_t=\frac{\mathbf{v}t}{1-\beta_2^t},]

[0142] [\theta{t+1}=\theta_t-

[0143] \eta\frac{\hat{\mathbf{m}}_t}{\sqrt{\hat{\mathbf{v}}_t}+\epsilon}.]

[0144] Among them, (\mathbf{m}_t) and (\mathbf{v}_t) are the first-order and second-order moment estimates of the gradient, (\beta_1,\beta_2) are exponential decay rates, (\eta) is the learning rate, and (\epsilon) is a numerical stability constant;

[0145] Decision-making and early warning

[0146] After the training is completed, the newly collected multimodal data is input into the trained model to output the prediction result of the corrosion state of the tower (\hat{y}). According to the warning threshold judgment result, if the corrosion state reaches the warning standard, the warning module is triggered to notify the relevant maintenance personnel.

[0147] Explanation of main parameters 1. (E_{ec}, I_{ec}): Potential and current data measured by electrochemical sensors, reflecting metal corrosion current and corrosion rate.

[0148] (T_{env},H_{env}): Ambient temperature and humidity, which affect the corrosion rate of metals.

[0149] (S_{strain},\epsilon_{strain}): stress and strain data used to evaluate the load condition of the tower structure.

[0150] (I_{img}): Image data, containing visual information of the tower surface, used to detect corrosion areas and extent.

[0151] (\mathbf{f}{img},\mathbf{f}{ec},\mathbf{f}{env},

[0152] \mathbf{f}{strain}): Feature vectors extracted from image, electrochemical, environmental and stress data, respectively.

[0153] (\mathbf{f}_{fusion}): Fusion feature vector, integrating information from all data modalities.

[0154] (\mathcal{L}_{CE}): Cross entropy loss function, used for training multimodal fusion models.

[0155] Here we selected convolutional neural network (CNN) for image data processing, and used the feature vector after multimodal data fusion as the model input to predict the corrosion state of the tower.

[0156] Model structure design The structure of the model includes the following layers:

[0157] Input layer: Receives the feature vector after multimodal fusion (\mathbf{f}_{\text{fusion}}).

[0158] Hidden layer: multi-layer fully connected layer, with ReLU as the activation function, used to extract high-level features.

[0159] Output layer: outputs the prediction result of the corrosion state, which is a binary classification of corrosion / no corrosion. The following is the application process of the multimodal fusion model in this system;

[0160] Furthermore, the early warning module can set different early warning levels according to the corrosion status and development trend of the tower, including mild warning, moderate warning and severe warning, which correspond to different degrees of corrosion and risk respectively; when the monitoring system detects that the corrosion status of the tower reaches the corresponding early warning level, it will automatically issue an early warning signal to remind relevant personnel to take measures.

[0161] Early warning method selection: There are many ways to warn, such as SMS, email, sound and light alarm. According to different user needs and scenarios, choose the appropriate early warning method to ensure that the early warning information can be conveyed to relevant personnel in a timely and accurate manner. At the same time, pay attention to the content and format of the early warning information to make it easy to understand and process.

[0162] Furthermore, the user interface module uses charts, reports, and maps to visualize the monitoring data; bar charts, line charts, pie charts and other charts can be used to display different types of data, such as corrosion current, temperature and humidity, stress and strain, etc.; reports are used to display detailed data information and analysis results; maps are used to mark the location of each tower, and different colors are used to represent different corrosion states; through visual display, users can intuitively understand the corrosion state and development trend of the tower.

[0163] Real-time monitoring interface: Design a real-time monitoring interface through which users can view the data and image information of each sensor in real time. A large screen display or multi-window display can be used to facilitate users to view the monitoring data of multiple towers at the same time. In the real-time monitoring interface, an early warning function should be set up so that users can be reminded in time when the monitoring system issues an early warning signal.

[0164] Parameter setting and management

[0165] User authority management: Set different levels of user authority to ensure the security and reliability of the system. It can be divided into multiple levels such as administrators, operators and ordinary users, with different operation permissions and data access permissions. Administrators can comprehensively manage and maintain the system, operators can perform operations such as data collection and analysis, and ordinary users can view monitoring data and early warning information.

[0166] System parameter settings: Users can set and manage the parameters of the monitoring system through the user interface module. You can set the sensor's acquisition frequency, warning threshold, communication parameters, etc., and you can also adjust and optimize the parameters of the deep learning algorithm. When setting parameters, pay attention to the rationality and effectiveness of the parameters to ensure the normal operation of the system.

[0167] A method for using a tower corrosion condition monitoring system, the steps are as follows:

[0168] S1. Sensor data collection

[0169] Each sensor collects tower data in real time at a preset collection frequency, including corrosion current, potential, temperature and humidity, stress and strain, image and tower tilt angle;

[0170] S2. The data acquisition terminal receives the data from the sensor and performs preliminary processing and storage;

[0171] S3. Data transfer:

[0172] 1) The data acquisition terminal transmits the processed data to the data processing module through the communication network;

[0173] 2) The data processing module stores the received data in a database;

[0174] S4. Use deep learning algorithms to analyze and process data, extract effective features, and evaluate the corrosion status, tilt and development trend of towers;

[0175] Based on the analysis results, determine whether the corrosion status and tilt of the tower have reached the warning level.

[0176] If the warning level is reached, a corresponding warning signal will be issued and a corresponding maintenance plan will be formulated.

[0177] S5 user interface display, which visualizes the monitoring data and analysis results through the user interface module;

[0178] Users can view the corrosion status, tilt, warning information and maintenance plan of the tower through the user interface module, and can also set and manage system parameters.

[0179] The advantages of the present invention are that the advantages and innovations of the monitoring system are that multi-sensor fusion comprehensively monitors the corrosion state and tilt of the tower from different angles. Through multi-sensor fusion technology, the accuracy and reliability of the monitoring system are improved. In combination with low-power wide area network (LPWAN) technology and 5G technology, stable data transmission and real-time monitoring are achieved, and the flexibility and adaptability of the system are improved. Deep learning algorithms such as convolutional neural network (CNN), recurrent neural network (RNN) and its variants and multimodal fusion models are used to deeply analyze and process the monitoring data, improve the intelligence level of the monitoring system, accurately evaluate the corrosion state, tilt and development trend of the tower, establish a complete early warning mechanism, issue early warning signals in time, remind relevant personnel to take measures, provide scientific decision-making support, formulate reasonable maintenance plans, reduce maintenance costs, and improve the reliability and safety of the power system. Intuitive user interface, design a friendly user interface, display monitoring data and analysis results in the form of intuitive charts, reports, maps, etc., convenient for users to view and manage, and improve user experience.

[0180] When the metal of the tower corrodes, a tiny electrochemical cell is formed between the metal and the environment, generating a corrosion current. The electrochemical sensor can detect this corrosion current and convert it into an electrical signal output. At the same time, the measurement of potential can help determine the corrosion potential of the tower metal in the environment, thereby evaluating the possibility and speed of corrosion.

[0181] Installation location: Electrochemical sensors are installed at the bottom and joints of the tower. These locations are easily exposed to moisture, oxygen and other corrosive substances, and are high-incidence areas for tower corrosion. By installing electrochemical sensors at these locations, the corrosion of the tower can be monitored more accurately.

[0182] Data collection frequency: The collection frequency of the electrochemical sensor can be set according to the corrosion rate of the tower and the monitoring requirements. For towers with faster corrosion rates, the collection frequency can be increased to timely grasp the changes in the corrosion state; for towers with slower corrosion rates, the collection frequency can be appropriately reduced to save energy and reduce the amount of data transmission.

[0183] Data accuracy and reliability: To ensure the data accuracy and reliability of electrochemical sensors, sensors need to be calibrated and maintained regularly. Calibration can be performed using standard solutions or reference electrodes with known corrosion potential to ensure the measurement accuracy of the sensor. At the same time, attention should be paid to the waterproof, moisture-proof and corrosion-proof measures of the sensor to prevent the sensor from being damaged by the external environment.

[0184] Working principle: Intelligent temperature and humidity sensors use humidity sensing elements and temperature measuring elements to measure the temperature and humidity in the environment. Humidity sensing elements usually use materials such as humidity-sensitive capacitors or humidity-sensitive resistors, and their capacitance or resistance value will change with changes in ambient humidity. Temperature measuring elements can use thermistors, thermocouples, etc., and determine the ambient temperature by measuring the resistance or potential changes caused by temperature.

[0185] Installation location: The intelligent temperature and humidity sensor can be installed in the environment around the tower to monitor the temperature and humidity changes in the environment where the tower is located. You can choose to install the sensor on the ground near the tower, on the tower bracket or other suitable locations to ensure that the sensor can accurately reflect the ambient temperature and humidity around the tower.

[0186] Data collection frequency: The changes in temperature and humidity are relatively slow, so the collection frequency of the intelligent temperature and humidity sensor can be appropriately reduced. Generally speaking, temperature and humidity data can be collected every few minutes or tens of minutes, and the specific collection frequency can be adjusted according to actual conditions.

[0187] Data accuracy and reliability: In order to improve the data accuracy and reliability of the intelligent temperature and humidity sensor, high-precision humidity sensing elements and temperature measuring elements can be used, and the sensors can be calibrated and error compensated. At the same time, attention should be paid to the protection measures of the sensors to prevent the sensors from being affected by dust, moisture and other contaminants and reducing the measurement accuracy.

[0188] Working principle: Stress and strain sensors usually use strain gauges or fiber Bragg grating technologies. Strain gauges are sensors based on the resistance strain effect. When the tower is subjected to external force, the strain gauge will deform, causing its resistance value to change. Fiber Bragg grating sensors use the sensitivity of the grating in the optical fiber to external stress. When the external stress changes, the wavelength of the grating will change.

[0189] Installation location: Stress and strain sensors can be installed at key stress-bearing parts of the tower, such as the tower's pillars, beams, etc. By monitoring the stress and strain changes in these parts, we can understand the stress conditions of the tower, determine whether the tower has structural damage or deformation, and then evaluate the corrosion status of the tower.

[0190] Data collection frequency: Stress and strain change relatively quickly, especially in severe weather conditions such as wind, rain, and snow, the stress condition of the tower may change greatly. Therefore, it is necessary to increase the collection frequency of stress and strain sensors in order to grasp the stress condition of the tower in a timely manner. Generally speaking, stress and strain data can be collected every few seconds or minutes.

[0191] Data accuracy and reliability: In order to ensure the data accuracy and reliability of stress and strain sensors, it is necessary to select high-quality sensors and install and debug the sensors. During the installation process, it is necessary to ensure that the sensor is in good contact with the tower to avoid looseness or displacement. At the same time, the sensor should be calibrated and maintained regularly to ensure the measurement accuracy of the sensor.

[0192] Working principle: The image sensor obtains the surface condition information of the tower by taking images of the tower's appearance. Visible light cameras, infrared cameras or other types of image sensors can be used, and the appropriate sensor type can be selected according to different monitoring needs. For example, a visible light camera can capture information such as the surface color and texture of the tower, and an infrared camera can detect the temperature distribution of the tower to find potential corrosion areas or hot spots.

[0193] Installation location: The image sensor can be installed at different locations on the tower to obtain images of the tower's appearance from different angles. You can choose to install the image sensor at the top, middle, and bottom of the tower, or install the camera at a fixed location around the tower to take a full-scale picture of the tower.

[0194] Data collection frequency: The frequency of image data collection can be adjusted according to actual conditions. Generally speaking, image data can be collected every few hours or days. For important towers or areas with high corrosion risks, the collection frequency can be appropriately increased.

[0195] Data accuracy and reliability: To improve the data accuracy and reliability of image sensors, it is necessary to select high-resolution image sensors and preprocess and analyze the images. Technologies such as image enhancement, denoising, and edge detection can be used to improve the quality and clarity of images. At the same time, attention should be paid to the protection of image sensors to prevent the sensors from being affected by dust, rain, and other contaminants that may reduce image quality.

[0196] Data collection

[0197] Data collection method: Distributed data collection method is used to transmit the data of each sensor to the data collection terminal through wired or wireless methods. For electrochemical sensors, intelligent temperature and humidity sensors, stress and strain sensors, etc., wired connection method can be used to transmit the sensor data to a nearby collection terminal. For image sensors, wireless transmission method can be used to transmit image data to a remote data processing center.

[0198] Data acquisition terminal: The data acquisition terminal is responsible for receiving the data from each sensor and performing preliminary processing and storage. A low-power, high-performance microprocessor or embedded system can be selected as the data acquisition terminal to ensure the stability and reliability of data acquisition. The data acquisition terminal can be installed on the tower or in a nearby distribution box to facilitate connection with the sensor and data transmission.

[0199] Data caching and preprocessing: In the data acquisition terminal, the data caching function can be set up. When the communication network fails, the data will be temporarily stored and transmitted after the network is restored. At the same time, the collected data can be preprocessed, such as data cleaning, removal of outliers and noise, etc., to improve the quality and reliability of the data.

[0200] Data Transfer

[0201] Communication technology selection: According to the distribution of towers and monitoring requirements, select appropriate communication technology for data transmission. Low-power wide area network (LPWAN) technology, such as LoRa, NB-IoT, etc., can be used for long-distance, low-power data transmission; 5G technology can also be used for high-speed data transmission and real-time monitoring scenarios. For remote areas or areas with weak communication infrastructure, LPWAN technology can be preferred; for areas that require real-time video monitoring or large-scale data transmission, 5G technology can be selected.

[0202] Network architecture: Build a stable and reliable communication network architecture to ensure smooth data transmission. You can use star, mesh or tree network topologies, depending on the actual situation. In the network, you can set up multiple relay nodes to enhance the signal coverage and transmission stability. At the same time, pay attention to the security and confidentiality of the network, and use encryption technology and access control measures to prevent data from being stolen or tampered with.

[0203] Data transmission protocol: Select a suitable data transmission protocol to ensure accurate data transmission and parsing. Lightweight IoT protocols such as MQTT and CoAP can be used, which are suitable for low-power and resource-constrained devices. These protocols are simple, efficient, and reliable, and can meet the data transmission requirements of the tower corrosion status monitoring system.

[0204] (III) Data processing and analysis module

[0205] Data storage

[0206] Database selection: Choose a suitable database management system to store and manage the collected data. You can use relational databases, such as MySQL, Oracle, etc., which are suitable for the storage and query of structured data; you can also use non-relational databases, such as MongoDB, Redis, etc., which are suitable for large-scale, high-concurrency data storage and processing. According to the characteristics and needs of the data, choose the appropriate database type to ensure data storage security and query efficiency.

[0207] Data storage method: Use distributed storage to store data on multiple nodes to improve data reliability and availability. Cloud storage services can be used to store data on cloud servers to facilitate data management and sharing. At the same time, pay attention to data backup and recovery strategies, and back up data regularly to prevent data loss.

[0208] Data analysis

[0209] Deep learning algorithm selection: According to the characteristics of the data and monitoring requirements, select the appropriate deep learning algorithm for data analysis. Convolutional neural network (CNN) can be used to process image data to identify the corrosion area and degree of the tower; recurrent neural network (RNN) and its variants can be used to analyze time series data to predict the development trend of tower corrosion; multimodal fusion model can be used to fuse different types of data to improve the accuracy and reliability of the monitoring system.

[0210] Model training and optimization: Use a large amount of labeled data to train the deep learning model, and continuously adjust the model's parameters and structure to improve the model's performance. Techniques such as data enhancement and regularization can be used to prevent overfitting and improve the model's generalization ability. At the same time, attention should be paid to the interpretability and stability of the model to ensure that the model's results are reliable and interpretable.

[0211] Data analysis process: Develop a complete data analysis process, including data preprocessing, feature extraction, model training, model evaluation and application. In each link, the quality and accuracy of the data must be strictly controlled to ensure the reliability of the analysis results. At the same time, the analysis process must be continuously optimized to improve the efficiency and accuracy of data analysis.

[0212] (IV) Early warning and decision-making module

[0213] Early warning mechanism

[0214] Warning level setting: Different warning levels are set according to the corrosion status and development trend of the tower. It can be divided into light warning, medium warning and severe warning, corresponding to different corrosion degrees and risk levels. When the monitoring system detects that the corrosion status of the tower reaches the corresponding warning level, it will automatically send out a warning signal to remind relevant personnel to take measures.

[0215] Early warning method selection: There are many ways to warn, such as SMS, email, sound and light alarm, etc. According to different user needs and scenarios, choose the appropriate early warning method to ensure that the early warning information can be conveyed to relevant personnel in a timely and accurate manner. At the same time, pay attention to the content and format of the early warning information to make it easy to understand and process.

[0216] Decision support

[0217] Maintenance plan formulation: According to the corrosion status and warning level of the tower, formulate a corresponding maintenance plan. It can include surface treatment, anti-corrosion coating repair, local reinforcement, replacement of parts and other maintenance measures. When formulating the maintenance plan, it is necessary to comprehensively consider factors such as maintenance cost, construction difficulty, and impact on the operation of the transmission line, and select the best maintenance plan.

[0218] Cost-benefit analysis: Conduct cost-benefit analysis on different maintenance plans to evaluate the economic and social benefits of each plan. You can use methods such as life cycle cost analysis to calculate the total cost and total benefit of each plan and select the plan with the best cost-effectiveness. At the same time, pay attention to future maintenance needs and risks to ensure the sustainability of the maintenance plan.

[0219] (V) User Interface Module

[0220] Visualization

[0221] Data visualization: Use charts, reports, maps and other forms to visualize the monitoring data. You can use bar charts, line charts, pie charts and other charts to display different types of data, such as corrosion current, temperature and humidity, stress and strain, etc. Use reports to display detailed data information and analysis results; use maps to mark the location of each tower, and use different colors to represent different corrosion states. Through visualization, users can intuitively understand the corrosion status and development trend of the tower.

[0222] Real-time monitoring interface: Design a real-time monitoring interface through which users can view the data and image information of each sensor in real time. A large screen display or multi-window display can be used to facilitate users to view the monitoring data of multiple towers at the same time. In the real-time monitoring interface, an early warning function should be set up so that users can be reminded in time when the monitoring system issues an early warning signal.

[0223] Parameter setting and management

[0224] User authority management: Set different levels of user authority to ensure the security and reliability of the system. It can be divided into multiple levels such as administrators, operators and ordinary users, with different operation permissions and data access permissions. Administrators can comprehensively manage and maintain the system, operators can perform operations such as data collection and analysis, and ordinary users can view monitoring data and early warning information.

[0225] System parameter settings: Users can set and manage the parameters of the monitoring system through the user interface module. You can set the sensor's acquisition frequency, warning threshold, communication parameters, etc., and you can also adjust and optimize the parameters of the deep learning algorithm. When setting parameters, pay attention to the rationality and effectiveness of the parameters to ensure the normal operation of the system.

[0226] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments, but it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A tower corrosion status monitoring system, characterized in that: It includes sensor module, data acquisition and transmission module, data processing and analysis module, early warning and decision module and visual interface module; The sensor module is used to collect data reflecting the corrosion status of the tower in real time from different angles, including: electrochemical sensor, intelligent temperature and humidity sensor, stress and strain sensor, image sensor; The data acquisition and transmission module is responsible for receiving and sending data in the system, and specifically includes: data acquisition terminal and communication transmission module: the data acquisition terminal is responsible for receiving data from each sensor, and performs preliminary processing and storage. The data acquisition terminal is installed on the tower or in a nearby distribution box; The data processing and analysis module specifically includes: data storage module and data analysis module; the data storage module uses MySQL, which is suitable for the storage and query of structured data, and also uses the non-relational database MongoDB, which is suitable for large-scale, high-concurrency data storage and processing; The data analysis module uses a convolutional neural network (CNN) to process image data, identify the corrosion area and degree of the tower, and also uses a multimodal fusion model to fuse different types of data; The collection and processing of multimodal data requires preprocessing before data fusion: Data collection: Electrochemical sensor data is represented as: (X_{\text{electro}}=[V,I]), where (V) is voltage and (I) is current; The data of the intelligent temperature and humidity sensor is represented as: (X_{\text{env}}=[T,H]), where (T) is temperature and (H) is humidity; The stress-strain sensor data is expressed as: (X_{\text{strain}}=[\sigma, \epsilon]), where (\sigma) is stress and (\epsilon) is strain; Image sensor data: (X_{\text{image}}=\text{RGB image}) Denoising: Filtering the sensor data; Normalization: The data normalization formula is: [X_i^{norm}=\frac{X_i-\min(X_i)}{\max(X_i)-\min(X_i)},] Among them, (X_i) represents the original data, (\min(X_i)) and (\max(X_i)) are the minimum and maximum values ​​of the data respectively; Interpolation: interpolation of missing data; Multimodal Data Fusion Strategies In the multimodal fusion model, fusion strategies include feature-level fusion, decision-level fusion, and hybrid-level fusion; The early warning and decision-making module evaluates and predicts the corrosion status based on the output of the data processing and analysis module; when potential corrosion risks are detected, the system will automatically generate early warning information; The visualization interface module displays real-time monitoring data, analysis results and warning information for users.

2. A tower corrosion status monitoring system according to claim 1, characterized in that: The feature extraction and fusion is based on the preprocessed multimodal data, using different deep learning models to extract features: Image data ((I_{img})): Use convolutional neural network (CNN) to extract local features and corrosion area information of the image; Electrochemical data ((E_{ec},I_{ec})): Use multi-layer perceptron (MLP) or recurrent neural network (RNN) to analyze time series trends; Temperature and humidity data ((T_{env},H_{env})): Use MLP for simple regression analysis; Stress-strain data ((S_{strain},\epsilon_{strain})): Use MLP or support vector machine (SVM) to model the tower structure changes; Each data modality is characterized by extracting features through the above model to obtain a feature vector: [\mathbf{f}{img},\mathbf{f}{ec},\mathbf{f}{env},\mathbf{f}{strain}] feature level fusion The feature vectors of different data modes are combined to form a comprehensive feature vector (\mathbf{f}_{fusion}), which can be expressed as: [\mathbf{f}{fusion}=f{concat}(\mathbf{f}{img},\mathbf{f}{ec}, \mathbf{f}{env},\mathbf{f}{strain}),] Where (f_{concat}) represents the feature connection function (such as vector concatenation or weighted summation).

3. A tower corrosion status monitoring system according to claim 2, characterized in that: The multimodal data fusion model training inputs the fused feature vector (\mathbf{f}{fusion}) into a deep learning model (such as a multi-layer perceptron, a convolutional neural network, etc.) for training. Assume that the loss function is the cross entropy loss (\mathcal{L}{CE}): [\mathcal{L}{CE}=-\sum{c=1}^{C}y_c\log(\hat{y}_c),] Among them, (y_c) is the actual category label, (\hat{y}_c) is the predicted probability, and (C) is the total number of categories. Update the model parameters (\theta) using the Adam optimizer: [\mathbf{m}t=\beta_1\mathbf{m}{t-1}+(1-\beta_1)\nabla_\theta\mathcal{L}t,] [\mathbf{v}t=\beta_2\mathbf{v}{t-1}+(1-\beta_2) (\nabla\theta\mathcal{L}_t)^2,] [\hat{\mathbf{m}}_t=\frac{\mathbf{m}_t}{1-\beta_1^t}, \quad\hat{\mathbf{v}}_t=\frac{\mathbf{v}t}{1-\beta_2^t},] [\theta{t+1}=\theta_t- \eta\frac{\hat{\mathbf{m}}_t}{\sqrt{\hat{\mathbf{v}}_t}+\epsi lon}.] where (\mathbf{m}_t) and (\mathbf{v}_t) are the first-order and second-order moment estimates of the gradient, respectively, (\beta_1,\beta_2) are exponential decay rates, (\eta) is the learning rate, and (\epsi lon) is a numerical stability constant; Decision-making and early warning After the training is completed, the newly collected multimodal data is input into the trained model to output the prediction result of the corrosion state of the tower (\hat{y}); according to the warning threshold judgment result, if the corrosion state reaches the warning standard, the warning module is triggered to notify the relevant maintenance personnel; 1. (E_{ec}, I_{ec}): Potential and current data measured by electrochemical sensors, reflecting metal corrosion current and corrosion rate. (T_{env},H_{env}): Ambient temperature and humidity, which affect the corrosion rate of metals. (S_{strain},\epsilon_{strain}): stress and strain data, used to evaluate the load condition of the tower structure; (I_{img}): image data, containing visual information of the tower surface, used to detect corrosion areas and extent; (\mathbf{f}{img},\mathbf{f}{ec},\mathbf{f}{env}, \mathbf{f}{strain}): Feature vectors extracted from image, electrochemical, environmental and stress data, respectively; (\mathbf{f}_{fusion}): fusion feature vector, integrating the information of all data modalities; (\mathcal{L}_{CE}): cross entropy loss function, used for training multimodal fusion models; Here we selected convolutional neural network (CNN) for image data processing and used the feature vector after multimodal data fusion as model input to predict the corrosion state of the tower; 1. Model structure design The structure of the model includes the following layers: Input layer: Receives the feature vector after multimodal fusion (\mathbf{f}_{\text{fusion}}). Hidden layer: multi-layer fully connected layer, with ReLU as the activation function, used to extract high-level features. Output layer: outputs the prediction result of the corrosion state, which is a binary classification of corrosion / no corrosion. The following is the application process of the multimodal fusion model in this system.

4. The tower corrosion status monitoring system according to claim 1 is characterized in that: The warning module can set different warning levels according to the corrosion status and development trend of the tower, including light warning, medium warning and severe warning, which correspond to different corrosion degrees and risk levels respectively; When the monitoring system detects that the corrosion status of the tower reaches the corresponding warning level, it will automatically send out a warning signal to remind relevant personnel to take measures; Selection of warning methods: A variety of warning methods can be used, such as SMS, email, and sound and light alarms.

5. A tower corrosion status monitoring system and method according to claim 1, characterized in that: The user interface module uses charts, reports, and maps to visualize the monitoring data; uses bar charts, line charts, pie charts, and other charts to display different types of data, such as corrosion current, temperature and humidity, stress and strain, etc.; uses reports to display detailed data information and analysis results; uses maps to mark the location of each tower, and uses different colors to represent different corrosion states; through visual display, users can intuitively understand the corrosion state and development trend of the tower; Real-time monitoring interface: Design a real-time monitoring interface through which users can view the data and image information of each sensor in real time. A large screen display or multi-window display can be used to facilitate users to view the monitoring data of multiple towers at the same time. In the real-time monitoring interface, an early warning function should be set up so that users can be reminded in time when the monitoring system issues an early warning signal; Parameter setting and management User authority management: set different levels of user authority; it can be divided into multiple levels such as administrator, operator and ordinary user, with different operation authority and data access authority; administrator can conduct comprehensive management and maintenance of the system, operator can conduct data collection and analysis, and ordinary user can view monitoring data and early warning information; System parameter setting: Users can set and manage the parameters of the monitoring system through the user interface module, including the sensor's collection frequency, warning threshold, and communication parameters.

6. A method for using a tower corrosion status monitoring system, characterized in that: Here are the steps: S1. Sensor data collection Each sensor collects tower data in real time at a preset collection frequency, including corrosion current, potential, temperature and humidity, stress and strain, images and tower tilt angle. S2. The data acquisition terminal receives the data from the sensor and performs preliminary processing and storage; S3. Data transfer: 1) The data acquisition terminal transmits the processed data to the data processing module through the communication network; 2) The data processing module stores the received data in a database; S4. Use deep learning algorithms to analyze and process data, extract effective features, and evaluate the corrosion status, tilt and development trend of towers; Based on the analysis results, determine whether the corrosion status and tilt of the tower have reached the warning level; If the warning level is reached, a corresponding warning signal will be issued and a corresponding maintenance plan will be formulated; S5 user interface display, which visualizes the monitoring data and analysis results through the user interface module; Users can view the corrosion status, tilt, warning information and maintenance plan of the tower through the user interface module, and can also set and manage system parameters.