Variable-frequency starting monitoring system and device for gas generator

Through intelligent sensor nodes and deep learning models, multimodal data acquisition and abnormal detection of the variable frequency start system of the gas generator is realized, solving the problems of fault positioning difficulties and reset button failure, and improving the stability and economic benefits of the system.

CN119070475APending Publication Date: 2024-12-03ZHEJIANG ZHENENG ELECTRIC POWER CO LTD XIAOSHAN POWER PLANT
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411078461.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-07
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing gas generator frequency conversion start system lacks effective monitoring, resulting in difficulty in positioning of faults, failure of reset buttons, frequent power outages and restarts of equipment, affecting equipment safety and economic benefits.

Method used

Intelligent sensor nodes are used to integrate multimodal sensors, and anomaly detection model is formed through data fusion algorithms and deep learning models. Combining adaptive learning rate and regularization technology, the training process is optimized, and abnormal data is analyzed in real time to achieve rapid fault identification and dynamic adjustment.

Benefits of technology

It improves the accuracy and real-time nature of fault detection, reduces the false alarm rate, reduces the time for fault analysis, avoids abnormal equipment damage, and enhances the stability and reliability of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119070475A_ABST
    Figure CN119070475A_ABST
Patent Text Reader

Abstract

The invention discloses a gas generator variable-frequency starting monitoring system and device, and the system comprises a monitoring system, a variable-frequency starting system, and an excitation system, a sensor is connected with a motor of a gas generator, and is used for controlling the frequency and voltage in the starting process of the generator; performing joint training and learning on the multi-modal data based on a recurrent neural network to form a comprehensive anomaly detection model; in combination with an adaptive learning rate and a regularization technology, the training process of the anomaly detection model is optimized, and the generalization ability and stability of the anomaly detection model are improved; and carrying out dynamic analysis on the multi-modal data by utilizing a real-time abnormal data processing algorithm, and identifying abnormal data. According to the invention, a fault source can be quickly positioned and a starting system can be monitored by collecting and analyzing data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of gas generators, and in particular to a gas generator variable frequency starting monitoring system. Background Art

[0002] With the expansion of the production scale of the domestic steel industry and the application of high inertia and high torque motors, the operation of large synchronous motors such as gas generators has received extensive attention from enterprises and power departments. In order to reduce the impact on the power grid when the motor starts and the torque vibration borne by the motor itself, ensure the safe, stable and continuous operation of the power grid and increase the operating life of the motor, metallurgical blast furnace blower units are usually driven by high-power synchronous motors. With the increasing capacity of blast furnaces, the power of the drive motor has also increased; the impact on the system power grid during the motor startup process has received more and more attention, and it has become the first choice to configure large synchronous units with high-power inverters as starting devices.

[0003] At present, most of the variable frequency starting systems of gas turbines do not have the specific status of the monitoring system when driving the gas turbines. Only a few individual operating parameters can be seen, which makes it difficult to accurately locate the root cause of the fault. The original system locates the fault of the gas turbine starting variable frequency system. Engineers need to measure or analyze each operating link before locating the specific fault in the sequential control of the starting variable frequency system. The process is cumbersome and time-consuming.

[0004] During the start-up of the gas turbine, there are many signal exchanges between the variable frequency starting system and the gas turbine control system. At present, after the gas turbine control system selects the variable frequency starting system, it accelerates the drag unit until it exits. If a signal is lost and the acceleration is not dragged, it stays in a certain step of the sequential control sequence or exits due to a fault. The gas turbine control system and the variable frequency starting system cannot quickly understand, and the specific cause is time-consuming and laborious. In addition, if the local reset button of the original system fails, the fault cannot be reset and the power must be cut off and restarted. If the source of the fault is still there, even if the power is cut off and restarted, the fault still exists and cannot be reset. In this way, multiple power cuts and restarts under unknown circumstances will inevitably lead to abnormal damage to the equipment. If the operating personnel need to check the inside of the equipment, they must first emergency stop the fault to open the cabinet door, and then they must cut off the power and restart to reset the fault. The use of variable frequency starting system process monitoring greatly improves the real-time monitoring, and visualization further observes the status feedback of the starting variable frequency system and the command signal sent by the gas turbine control system on the spot. At the same time, the signal recording and storage is conducive to the post-start analysis and troubleshooting of the start, and the abnormal problems are quickly and effectively solved. The redundant reset fault function has been added to avoid multiple power cuts and restarts of the system.

[0005] At present, in the sequential control process of the variable frequency starting system to drive the unit to run, because it is operated as a black box internally, if a specific fault occurs, the root cause cannot be quickly found, and the manufacturer's engineers need to make specific inquiries, which brings huge pressure to the power plant to quickly start the unit. When the sequential control starting variable frequency system fails, it was previously impossible to quickly confirm the specific root cause of the problem, which increases the time for fault analysis and cannot quickly analyze and locate the fault point, which is greatly unfavorable to the normal recovery of the startup, shortens the available hours of the gas turbine, and has an adverse impact on economic benefits.

[0006] Therefore, it is necessary to provide a gas generator variable frequency starting monitoring system and device to solve the above technical problems. Summary of the invention

[0007] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0008] In view of the above existing problems, the present invention is proposed.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions: integrating temperature sensors, sound sensors, pressure sensors, speed sensors, and current sensors in intelligent sensor nodes to perform multi-modal data acquisition and obtain multi-modal data for monitoring key operating parameters of gas generators;

[0010] Using a data fusion algorithm, the multimodal data is fused into a unified data set;

[0011] The sensor is connected to the motor of the gas generator and is used to control the frequency and voltage during the generator startup process;

[0012] Jointly training and learning the multimodal data based on a recurrent neural network to form a comprehensive anomaly detection model;

[0013] Combining adaptive learning rate and regularization technology to optimize the training process of the anomaly detection model and improve the generalization ability and stability of the anomaly detection model;

[0014] Using a real-time abnormal data processing algorithm, dynamically analyzing the multimodal data to identify abnormal data;

[0015] Combined with the abnormal data classification algorithm, the abnormal data is divided to distinguish between real abnormalities and false alarms, and the obtained classification results are fed back to the monitoring system to confirm the real fault code and realize dynamic adjustment and optimization.

[0016] As a preferred solution of the gas generator variable frequency starting monitoring system described in the present invention, the key operating parameters include temperature, pressure, rotation speed and current.

[0017] As a preferred solution of the gas generator variable frequency starting monitoring system described in the present invention, the multimodal data includes speed data, sound data, vibration data, temperature data, humidity data, pressure data, and current / voltage data.

[0018] As a preferred solution of the gas generator variable frequency start monitoring system of the present invention, the multi-modal data is fused by using the data fusion algorithm, which is to perform linear weighted summation of the data collected by different sensors according to the weights. The mathematical expression formula is as follows:

[0019]

[0020] Among them, D 融合 is the fused data, D i is the data collected by the i-th sensor, ω i is the weight corresponding to the data collected by the i-th sensor, and N represents the number of sensors;

[0021] The fused multimodal data is then preprocessed, and the preprocessing at least includes denoising and normalization.

[0022] As a preferred solution of the gas generator variable frequency starting monitoring system described in the present invention, the preset algorithm includes a dynamic adjustment strategy, which can automatically optimize the output frequency and voltage of the inverter according to the real-time monitored operating parameters to achieve smooth starting and operation.

[0023] During the model training process, the model parameters are updated by minimizing the loss function to optimize the model's prediction effect and generalization ability.

[0024] As a preferred solution of the gas generator variable frequency startup monitoring system of the present invention, the mathematical expression for optimizing the training process of the anomaly detection model is as follows:

[0025]

[0026] Among them, θ * is the optimal model parameter, y n is the actual label of the nth training sample, x n is the feature vector of the nth training sample, f(x n ; θ) is a deep learning model used to predict the label of the nth training sample, L is the number of training samples, J(y n ,f(x n; θ)) is the objective function of the loss function, which is used to measure the difference between the model prediction value and the actual value.

[0027] As a preferred solution of the gas generator variable frequency starting monitoring system described in the present invention, the variable frequency starting system includes an alarm module, which is communicated with the CPU and is used to send an alarm signal when abnormal operating conditions are detected. The alarm module can send alarm information to the user through sound, light or electronic communication.

[0028] As a preferred solution of the gas generator variable frequency starting monitoring device of the present invention, it includes:

[0029] Touch screen human-machine interface device IPC HMI; monitors the commands received and feedback sent during the startup of the variable frequency starting system, as well as the status of each switch during the startup process;

[0030] The IPC HMI and PLC use the MPI communication method, and several different screens or areas are made on the IPC HMI to display the instructions and status of the startup process, so that relevant personnel can observe on the spot during startup;

[0031] a memory storing operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations;

[0032] The signals between the gas turbine control system and the variable frequency starting system and between the excitation and variable frequency starting system are hard wired, including the gas turbine control system giving the variable frequency starting system command and the variable frequency starting system status feedback to the gas turbine control system, and the variable frequency starting system giving the excitation command and the excitation status feedback to the variable frequency starting system;

[0033] The status of the starting frequency conversion system, excitation system, gas turbine control system and all auxiliary equipment is collected and sent to the human-machine interface device for visualization in real time through communication.

[0034] The beneficial effects of the present invention are as follows: all external collection quantities are sampled on a large scale by means of communication, and each control step in the sequential control process is vividly reflected through the motion control screen, so that the root cause of the fault can be vividly seen when a certain step fails. The present invention improves the accuracy, real-time and stability of abnormal detection, reduces the false alarm rate, and ensures the safety and efficiency of gas generators by adopting multimodal data acquisition, data fusion, deep learning training, optimization of training process, real-time abnormal processing and abnormal data classification. When the sequential control starting frequency conversion system fails, the specific root cause of the problem can be quickly confirmed, the time for analyzing the fault can be reduced, and the fault point can be quickly analyzed and located. The frequency conversion starting system process monitoring is greatly improved by the human-machine interface device, and the visualization further observes the state feedback of the starting frequency conversion system and the command signal sent by the gas turbine control system on site. At the same time, the signal recording and storage is conducive to the post-start analysis and troubleshooting of the start, and the abnormal problem is quickly and effectively solved. The redundant reset fault function is added to avoid multiple power outages and restarts of the system.

[0035] The startup process monitoring and recording of the variable frequency starting system is well compatible with the original control system and does not affect the original control logic. During the operation of the variable frequency starting system to drive the unit, the status relationship between the gas turbine control system, the excitation system and the variable frequency starting system, the status of the variable frequency starting system itself, and the sequence of the variable frequency starting system sequence control program are displayed in real time. At the same time, the opening and closing status of the relevant circuit breakers and the action signal of the generator protection can also be displayed. If the control strategy and signal are changed in the future, the monitoring screen is also very convenient to modify according to the actual situation, increase or decrease the signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0037] in:

[0038] Figure 1 It is a flow chart of the gas generator variable frequency starting monitoring system and device shown in the present invention. DETAILED DESCRIPTION

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0040] Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without making any creative work should fall within the scope of protection of the present invention.

[0041] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0042] Example 1

[0043] According to an embodiment of the present invention, Figure 1 The flowchart shown is a gas generator variable frequency starting monitoring system, which specifically includes:

[0044] Integrate temperature sensors, sound sensors, pressure sensors, speed sensors, and current sensors in intelligent sensor nodes to collect data in a multi-modal manner and obtain multi-modal data for monitoring key operating parameters of gas generators;

[0045] Use data fusion algorithms to fuse multimodal data into a unified data set;

[0046] The sensor is connected to the motor of the gas generator and is used to control the frequency and voltage during the generator startup process;

[0047] Jointly train and learn multimodal data based on recurrent neural networks to form a comprehensive anomaly detection model;

[0048] Combine adaptive learning rate and regularization technology to optimize the training process of the anomaly detection model and improve the generalization ability and stability of the anomaly detection model;

[0049] Use real-time abnormal data processing algorithms to dynamically analyze multimodal data and identify abnormal data;

[0050] Combined with the abnormal data classification algorithm, the abnormal data is divided to distinguish between real abnormalities and false alarms, and the classification results are fed back to the monitoring system to confirm the real fault code and achieve dynamic adjustment and optimization.

[0051] The following describes in more detail the implementation process and / or effects of certain embodiments of the present invention in conjunction with some preferred or optional examples of the present invention.

[0052]

Integrated multimodal sensor

[0053] In an optional implementation, the multimodal data includes rotation speed data, sound data, vibration data, temperature data, humidity data, pressure data, and current / voltage data.

[0054] As an example, the rotation speed data is image or video data captured by a rotation speed sensor and is used to observe the actual operation of the gas generator, such as detecting whether there is abnormal movement of the equipment or whether there are abnormal objects.

[0055] As an example, sound data is a sound signal collected by a sound sensor, which is used for the sound produced when a gas generator is started, such as mechanical running sounds and abnormal sounds, to help identify whether the machine is operating normally.

[0056] As an example, vibration data is vibration signal data collected by a vibration sensor, which is used to monitor the vibration conditions of the gas generator when it rotates, such as vibration amplitude and frequency, and is used to detect whether there is abnormal vibration or vibration of the equipment.

[0057] As an example, temperature data is temperature information collected by a temperature sensor and is used to describe the temperature conditions of various components of a gas generator. Abnormal temperatures may indicate a malfunction or overheating of the equipment.

[0058] As an example, humidity data is humidity information collected by a humidity sensor and is used to monitor humidity changes in the workplace of a gas generator, which may affect the operating status of the equipment.

[0059] As an example, pressure data is pressure information collected by a pressure sensor and is used to detect pressure changes when a gas generator is working, such as pressure changes in a gas pressure system, so as to detect abnormal conditions in a timely manner.

[0060] As an example, current / voltage data is current and voltage information collected by a current sensor and is used to monitor the working status of the gas generator electrical system, such as the current change of the motor, in order to detect electrical faults.

[0061] In summary, by comprehensively considering the data of rotation speed, sound, vibration, and other environmental parameters (such as temperature, humidity, pressure, current, etc.), multimodal data collection of intelligent sensor nodes can be realized, thereby comprehensively monitoring the operating status of smart mining equipment and detecting abnormal situations in a timely manner.

[0062] Preferably, by integrating multiple sensors such as speed sensors, sound sensors, vibration sensors, etc. into the intelligent sensor node, multi-dimensional data collection of the gas generator is achieved. This multimodal data collection method makes the perception of the operating status of the gas generator more comprehensive and helps to accurately capture various abnormal conditions in the operation of the gas generator.

[0063]

Data Fusion

[0064] Furthermore, the embodiment of the present invention utilizes a data fusion algorithm to fuse multimodal data, which is to perform linear weighted summation of the data collected by different sensors according to weights. The mathematical expression formula is as follows:

[0065]

[0066] Among them, D 融合 is the fused data, D i is the data collected by the i-th sensor, ω i is the weight corresponding to the data collected by the i-th sensor, and N represents the number of sensors;

[0067] The fused multimodal data is then preprocessed, and the preprocessing at least includes denoising and normalization.

[0068] As an example, denoising can be performed by filtering the signal, and its mathematical expression is as follows:

[0069]

[0070] Among them, x filtered (t) is the signal after filtering, x(k) is the original signal, and Q is the filter window size.

[0071] As an example, data collected by different sensors are normalized to have a uniform scale, such as scaling the data to a range of [0, 1] or [-1.1].

[0072] As an example, the moving average method is used to smooth the data to reduce the volatility of the data and improve the stability of the data.

[0073] Preferably, the embodiment of the present invention utilizes a data fusion algorithm to integrate the data collected by different sensors into a unified data set, so that the correlation between the data is clearer, the possibility of data redundancy and information loss is reduced, and it is beneficial to improve the training effect and accuracy of subsequent anomaly detection models.

[0074]

Joint training of recurrent neural networks

[0075] Furthermore, the embodiment of the present invention performs joint training and learning on multimodal data based on a recurrent neural network to form a comprehensive anomaly detection model, and its mathematical expression formula is as follows:

[0076] Z=Data Fusion (D)

[0077] X = data preprocessing (Z)

[0078]

[0079]

[0080] Where D is the multimodal data set, Z is the data set after data fusion, X is the preprocessed multimodal data, and Y is the abnormal label used to indicate the abnormal state of the multimodal data. represents the abnormal label predicted by the model, θ is the parameter of the deep learning model, is the loss function used to measure the prediction effect of the model, η is the learning rate used to control the update speed of the model parameters, R(θ) is the regularization term used to constrain the complexity of the model, and λ is the regularization parameter used to penalize the model complexity and prevent overfitting.

[0081] It should be noted that during the model training process, the model parameters are updated by minimizing the loss function to optimize the model's prediction effect and generalization ability.

[0082] In a preferred embodiment, by adjusting the learning rate and controlling the step size of parameter updates, the model convergence speed is accelerated, the training efficiency is improved, and the oscillation phenomenon caused by excessive parameter updates is avoided.

[0083] It is not difficult to understand that by jointly training and learning multimodal data based on recurrent neural networks, a comprehensive anomaly detection model is formed, which can better mine the intrinsic relationships and features between data, improve the generalization ability and accuracy of the anomaly detection model, and enable it to better adapt to different working conditions.

[0084] Optimize the training process

[0085] The mathematical expression for the training process of optimizing the anomaly detection model is as follows:

[0086]

[0087] Among them, θ * is the optimal model parameter, y n is the actual label of the nth training sample, x n is the feature vector of the nth training sample, f(x n ; θ) is a deep learning model used to predict the label of the nth training sample, L is the number of training samples, J(y n ,f(x n ; θ)) is the objective function of the loss function, which is used to measure the difference between the model prediction value and the actual value.

[0088] In a preferred embodiment, the optimal model parameters are obtained by minimizing the loss function, so that the model can better fit the training data and improve the accuracy and generalization ability of the anomaly detection model.

[0089] In a preferred embodiment, the complexity of the model is controlled by regularization parameters, overfitting is prevented, the generalization ability of the model is improved, and the adaptability of the model to unknown data is enhanced.

[0090] In a preferred implementation, all training samples are fully utilized to improve the training effect of the model and reduce the deviation in the training process, thereby obtaining a more accurate model.

[0091] In a preferred embodiment, the range of model parameters is limited by regularization terms to avoid overfitting problems caused by excessive parameters, improve the generalization ability of the model, and make the model more universal.

[0092] It should also be noted that the embodiments of the present invention optimize the training process of the anomaly detection model by combining adaptive learning rate and regularization technology. The application of adaptive learning rate and regularization technology helps to avoid the problems of overfitting or underfitting in the model training process, improves the generalization ability and stability of the model, and makes the anomaly detection model more reliable.

[0093] In summary, by optimizing the training process of the anomaly detection model, a more accurate, stable and generalized anomaly detection model is obtained, thereby improving the anomaly detection effect during the startup of the gas generator and ensuring the safe operation of the gas generator.

[0094]

Real-time abnormal data processing

[0095] The embodiment of the present invention uses a real-time abnormal data processing algorithm to dynamically analyze multimodal data and identify abnormal data, which specifically includes the following steps:

[0096] Define the multimodal data set as D = {D1, D2, ..., D i}, where D i is the data collected by the i-th sensor;

[0097] Using the anomaly detection function f(D i ) Perform anomaly detection on the data collected by each sensor and output anomaly scores;

[0098] The anomaly score of the integrated multi-sensor data is the weighted sum of the anomaly scores of each sensor, and its mathematical expression is as follows:

[0099]

[0100] Among them, ω i represents the weight of the i-th sensor, that is, the importance of different sensors for anomaly detection, f(D i ) is an anomaly detection function;

[0101] By dynamically adjusting the weight ωi And the anomaly detection function f(D i ) parameters to identify abnormal data.

[0102] It should be noted that the value range of this formula is the real number domain. The larger the anomaly score S is, the higher the degree of data anomaly is, and the smaller the anomaly score S is, the lower the degree of data anomaly is.

[0103] In an optional implementation, an adaptive learning algorithm is used to dynamically adjust weights and parameters to adapt to abnormal situations under different environments and conditions, thereby improving the accuracy and stability of anomaly detection.

[0104] Preferably, the embodiments of the present invention improve the comprehensiveness and accuracy of anomaly detection by taking into account the abnormal conditions of each sensor through weighted summation of the anomaly scores of each sensor; and dynamically adjust the weights and parameters of the anomaly detection function through an adaptive learning algorithm, adaptively adjust the model according to real-time data and environmental conditions, so that it adapts to different abnormal conditions, thereby improving the adaptability and stability of anomaly detection.

[0105] The embodiments of the present invention combine anomaly detection functions with real-time data processing algorithms to dynamically analyze multimodal data and identify abnormal data in real time, thereby promptly discovering and responding to abnormal situations and ensuring the safe and stable operation of equipment and the environment. The embodiments of the present invention comprehensively consider multiple sensor data, dynamically adjust weights and parameters, and process real-time abnormal data to improve the accuracy and stability of anomaly detection, effectively reduce the false alarm rate, and improve detection efficiency.

[0106] Preferably, the real-time abnormal data processing algorithm combined with the adaptive learning algorithm can effectively and comprehensively utilize multi-sensor data, dynamically adjust model parameters, realize real-time abnormal data processing and dynamic analysis, thereby improving the accuracy and stability of anomaly detection and ensuring the safe operation of equipment and environment.

[0107] Preferably, the embodiment of the present invention combines an abnormal data classification algorithm to divide the abnormal data to distinguish between real anomalies and false alarms. Such a classification method helps to accurately identify abnormal situations and feed back the classification results to the monitoring system to achieve dynamic adjustment and optimization; and by classifying the abnormal data, it is possible to effectively reduce the false alarm rate, improve the performance and stability of the anomaly detection system, and provide a strong guarantee for the safe operation of the equipment.

[0108] Furthermore, in order to better illustrate the authenticity of the technical effects of the embodiments of the present invention, this embodiment is implemented in a gas generator variable frequency starting monitoring system according to the aforementioned invention content to verify its effect in practical applications.

[0109] (1) Experimental preparation

[0110] Data collection: Prepare a data set containing speed, sound, vibration, temperature, humidity, pressure, current / voltage data to simulate multimodal data in the operating environment of the gas generator.

[0111] Data preprocessing: De-noise the collected raw data, use filters to denoise the signals, and perform normalization and data smoothing to reduce data volatility and improve stability.

[0112] Data fusion: Integrate data from different sensors into a unified data set, use linear weighted fusion algorithm and principal component analysis (PCA) algorithm to fuse the data, extract the main features, and achieve data fusion.

[0113] Model training: Based on the recurrent neural network (RNN), multimodal data is jointly trained and learned to form a comprehensive anomaly detection model. During the training process, adaptive learning rate and regularization technology are used to optimize the training process of the anomaly detection model and improve the generalization ability and stability of the model.

[0114] Abnormal data processing: Use real-time abnormal data processing algorithms to dynamically analyze multimodal data, identify abnormal data, and divide abnormal data in combination with abnormal data classification algorithms to distinguish between real anomalies and false alarms.

[0115] (2) Experimental process

[0116] Set up intelligent sensor nodes: which integrate multi-modal sensors such as speed sensors, sound sensors, and vibration sensors and are deployed on gas generators;

[0117] Data collection: Each sensor continuously collects data, including speed data, sound data, vibration data, temperature data, humidity data, pressure data, and current / voltage data;

[0118] Data integration and preprocessing: Use data fusion algorithm to integrate data from different sensors and perform preprocessing, including denoising and normalization, to divide the data into three groups, namely experimental sample 1, experimental sample 2, and experimental sample 3;

[0119] Anomaly detection and analysis: Use recurrent neural network models to jointly train and learn multimodal data to form a comprehensive anomaly detection model, and combine real-time anomaly data processing algorithms to dynamically analyze data and identify abnormal data;

[0120] Abnormal data classification: Combined with the abnormal data classification algorithm, the abnormal data is divided to distinguish between real anomalies and false alarms.

[0121] The data processing module uses the DataHub DH5000 model, which receives the data sent by the intelligent sensor nodes, performs data integration, preprocessing, anomaly detection and analysis, and finally outputs the anomaly score, as shown in the following table:

[0122] Table 1. Statistics of data anomaly scores.

[0123]

[0124] Referring to Table 1, by comparing the anomaly scores of experimental samples 1, 2, and 3, we can clearly see the anomaly detection performance of different samples in various data types; for example, the vibration data anomaly score of experimental sample 1 is low, while the temperature data anomaly score is high, indicating that the vibration data in this sample is relatively stable, while the temperature data has anomalies; for another example, the pressure data anomaly score of experimental sample 2 is high, while the temperature data anomaly score is low, indicating that the temperature data in this sample is relatively stable; these data scores effectively reflect the performance and superiority of this method in anomaly monitoring.

[0125] The startup process monitoring and recording of the variable frequency starting system is well compatible with the original control system and does not affect the original control logic. During the operation of the variable frequency starting system to drag the unit, the status relationship between the gas turbine control system, the excitation system and the variable frequency starting system, the status of the variable frequency starting system itself, and the sequence of the variable frequency starting system sequence are displayed in real time. At the same time, the opening and closing status of the relevant circuit breakers and the action signal of the generator protection can also be displayed. In the event that the control strategy and signal are changed in the future, the monitoring screen is also very convenient to modify according to the actual situation, and increase or decrease the signal. The present invention greatly shortens the fault analysis time and has a great improvement on the economy and reliability of operation.

[0126] It should be noted that some other aspects disclosed in the embodiments of the present invention further provide a gas generator variable frequency starting monitoring device, including: a touch screen human-machine interface device IPC HMI; monitoring the instructions received and the feedback sent during the variable frequency starting system startup process and the state of each switch opening and closing during the startup process;

[0127] The IPC HMI and PLC use the MPI communication method, and several different screens or areas are made on the IPC HMI to display the instructions and status of the startup process, so that relevant personnel can observe on the spot during startup;

[0128] a memory storing operable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations;

[0129] The signals between the gas turbine control system and the variable frequency starting system and between the excitation and variable frequency starting system are hard wired, including the gas turbine control system giving the variable frequency starting system command and the variable frequency starting system status feedback to the gas turbine control system, and the variable frequency starting system giving the excitation command and the excitation status feedback to the variable frequency starting system;

[0130] The status of the starting frequency conversion system, excitation system, gas turbine control system and all auxiliary equipment is collected and sent to the human-machine interface device for visualization in real time through communication.

[0131] Real-time acquisition transfers all external data that need to be collected and analyzed into the self-developed screen, and uses the screen to vividly reflect all relevant quantities of the starting frequency conversion system in the starting sequence control, and makes it into a motion control screen. The monitoring screen is developed by using Siemens Botu software, which collects the status of the starting frequency conversion system, excitation system, gas turbine control system and all auxiliary equipment, and sends it to the industrial control panel device for visualization in real time through communication.

[0132] The control part of Siemens start-up frequency conversion system consists of S7-300 series PLC and CU320. Profibus DP communication is used between the two. The signals between the gas turbine control system and the frequency conversion start-up system and between the excitation and frequency conversion start-up system are hard-wired, including the gas turbine control system giving the frequency conversion start-up system command and the frequency conversion start-up system status feedback to the gas turbine control system, and the frequency conversion start-up system giving the excitation command and the excitation status feedback to the frequency conversion start-up system. In order to monitor the frequency conversion start-up system process, a touch screen human-machine interface device IPC HMI is added to monitor the instructions received and the feedback issued during the start-up process of the frequency conversion start-up system and the state of each switch during the start-up process. The installation position can be opened in the control cabinet door according to the corresponding size. MPI communication is used between IPC HMI and PLC. Several different screens or areas are made in IPC HMI to display the instructions and status of the start-up process, which is convenient for relevant personnel to observe on the spot during the start-up. All variables displayed on the monitoring screen record the signals when the display starts recording and save them in the memory card, so that they can be consulted later to assist in troubleshooting and problem solving. Solved the problem of the frequency conversion system being unable to be monitored in real time during startup

[0133] In summary, the present invention samples all external acquisition quantities on a large scale through communication, and vividly reflects each control step in the sequential control process through the motion control screen, so that the root cause of the fault can be vividly seen when a fault occurs in a certain step.

[0134] It should be appreciated that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory.

[0135] The method may be implemented in a computer program using standard programming techniques, including a non-transitory computer-readable storage medium configured with a computer program, wherein the storage medium so configured causes a computer to operate in a specific and predefined manner.

[0136] Each program may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system, however, the program can be implemented in assembly or machine language if desired; in any case, the language may be a compiled or interpreted language.

[0137] Furthermore, the program can be run on an application specific integrated circuit programmed for this purpose.

[0138] Additionally, the operations of the processes described herein can be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context.

[0139] The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that executes collectively on one or more processors, by hardware, or a combination thereof.

[0140] The computer program includes a plurality of instructions executable by one or more processors.

[0141] Further, the method may be implemented in any type of computing platform operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device.

[0142] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A gas generator variable frequency starting monitoring system, characterized in that: include: Monitoring system, variable frequency starting system, excitation system, Integrate temperature sensors, sound sensors, pressure sensors, speed sensors, and current sensors in intelligent sensor nodes to collect data in a multi-modal manner and obtain multi-modal data for monitoring key operating parameters of gas generators; Using a data fusion algorithm, the multimodal data is fused into a unified data set; The sensor is connected to the motor of the gas generator and is used to control the frequency and voltage during the generator startup process; Jointly training and learning the multimodal data based on a recurrent neural network to form a comprehensive anomaly detection model; Combining adaptive learning rate and regularization technology to optimize the training process of the anomaly detection model and improve the generalization ability and stability of the anomaly detection model; Using a real-time abnormal data processing algorithm, dynamically analyzing the multimodal data to identify abnormal data; Combined with the abnormal data classification algorithm, the abnormal data is divided to distinguish between real abnormalities and false alarms, and the obtained classification results are fed back to the monitoring system to confirm the real fault code and realize dynamic adjustment and optimization.

2. The gas generator variable frequency starting monitoring system according to claim 1 is characterized in that: The key operating parameters include temperature, pressure, rotation speed and current.

3. The gas generator variable frequency starting monitoring system according to claim 1 is characterized in that: The multimodal data includes rotation speed data, sound data, vibration data, temperature data, humidity data, pressure data, and current / voltage data.

4. The gas generator variable frequency starting monitoring system according to claim 1 is characterized in that: The multimodal data is fused using the data fusion algorithm, which is to perform linear weighted summation of the data collected by different sensors according to the weights. The mathematical expression formula is as follows: Among them, D 融合 is the fused data, D i is the data collected by the i-th sensor, ω i is the weight corresponding to the data collected by the i-th sensor, and N represents the number of sensors; The fused multimodal data is then preprocessed, and the preprocessing at least includes denoising and normalization.

5. The gas generator variable frequency starting monitoring system according to claim 1, characterized in that: It also includes preset algorithms containing dynamic adjustment strategies that can automatically optimize the inverter's output frequency and voltage based on real-time monitored operating parameters to achieve smooth startup and operation; During the model training process, the model parameters are updated by minimizing the loss function to optimize the model's prediction effect and generalization ability.

6. The gas generator variable frequency starting monitoring system according to claim 1, characterized in that: The mathematical expression for optimizing the training process of the anomaly detection model is as follows: Among them, θ * is the optimal model parameter, y n is the actual label of the nth training sample, x n is the feature vector of the nth training sample, f(x n ; θ) is a deep learning model used to predict the label of the nth training sample, L is the number of training samples, J(y n ,f(x n ; θ)) is the objective function of the loss function, which is used to measure the difference between the model prediction value and the actual value.

7. The gas generator variable frequency starting monitoring system according to claim 1 is characterized in that: The variable frequency starting system includes an alarm module, which is in communication with the CPU and is used to send an alarm signal when an abnormal operating condition is detected. The alarm module can send alarm information to the user through sound, light or electronic communication.

8. A monitoring device, comprising the gas generator variable frequency starting monitoring system according to any one of claims 1 to 7, characterized in that: include: Touch screen human-machine interface device IPC HMI; Monitor the whole process of the received instructions and feedback of the variable frequency starting system during startup, as well as the status of each switch during startup; The IPC HMI and PLC use the MPI communication method, and several different screens or areas are made on the IPC HMI to display the instructions and status of the startup process, so that relevant personnel can observe on the spot during startup; a memory storing operable instructions that, when executed by one or more processors, cause the one or more processors to perform operations; The signals between the gas turbine control system and the variable frequency starting system and between the excitation and variable frequency starting system are hard wired, including the gas turbine control system giving the variable frequency starting system command and the variable frequency starting system status feedback to the gas turbine control system, and the variable frequency starting system giving the excitation command and the excitation status feedback to the variable frequency starting system; The status of the starting frequency conversion system, excitation system, gas turbine control system and all auxiliary equipment is collected and sent to the human-machine interface device for visualization in real time through communication.

Citation Information

Cited By

  • Intelligent welding method for machine tool manufacturing

    CN120055614A

  • Gas generator set performance optimization method and system based on deep learning

    CN120428575A