Gas turbine state intelligent early warning method and system based on dynamic neural network

Abstract

The invention provides a gas turbine state intelligent early warning method and system based on a dynamic neural network. The method comprises the steps of gas turbine historical operation data acquisition, characteristic parameter screening, neural network prediction model construction and training, gas turbine real-time data acquisition, gas turbine key operation state characteristic parameter prediction and analysis, dynamic updating of a gas turbine historical database and a neural network prediction model and the like. And real-time monitoring and early warning of the running state of the gas turbine are realized. According to the method, through boundary condition-operation state characteristic parameter classification and neural network prediction model dynamic modeling and training, the problems of limitation of single-parameter early warning in traditional gas turbine state monitoring, lack of prediction model updating mechanisms and the like are solved, and a technical means can be provided for intelligent operation and maintenance of the gas turbine.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for gas turbines, and in particular to an intelligent early warning method and system for gas turbine status based on dynamic neural networks. Background Technology

[0002] Gas turbines, as core power equipment, are widely used in aviation, shipbuilding, and power generation. Their operating status directly affects the safety and reliability of the entire system. With the development of intelligent manufacturing and predictive maintenance technologies, gas turbine condition early warning technology has become a key means to ensure the safe operation of equipment. However, existing technologies mostly use single-variable threshold judgment or single-parameter prediction models, which make it difficult to capture the coupling relationship between parameters. When multiple parameters show slight anomalies simultaneously, the system may not be able to trigger an early warning in time. In addition, current early warning models are mostly trained on historical operating data and remain fixed once deployed, while the performance of gas turbines will continue to decline due to factors such as component aging and changes in fuel characteristics, leading to a gradual increase in model prediction bias. Summary of the Invention

[0003] This invention aims to at least solve the technical problems of limitations in single-parameter early warning and lack of prediction model update mechanism in the prior art. In particular, it innovatively proposes a method and system for intelligent early warning of gas turbine status based on dynamic neural network.

[0004] On one hand, the present invention provides a method for intelligent early warning of gas turbine status based on dynamic neural networks, the method comprising: Collect historical operating data of gas turbines, including atmospheric environment data, operating status data and monitoring data, and construct a historical operating database of gas turbines; Gas turbine operating data is divided into boundary condition characteristic parameters and state characteristic parameters. Boundary condition characteristic parameters are parameters that characterize the boundary conditions of gas turbine operation, while state characteristic parameters are parameters that characterize the operating state of gas turbine. Using boundary condition feature parameters as input and state feature parameters as output, the entire operating data of a single "start-up-run-shutdown" process is used as a set of samples, and samples from the gas turbine historical operation database are used as the training set to train a neural network prediction model. Collect real-time operating data of the gas turbine, including atmospheric environment data, operating status data, and monitoring data; The trained neural network prediction model is used to predict the gas turbine state characteristic parameters during the "start-up-operation-shutdown" process. The deviation between the predicted value and the actual value is calculated. When the deviation exceeds the set threshold, the current gas turbine operating state is considered abnormal. After the gas turbine is shut down, the data from this operation is added to the historical operation database, and the data from the earliest start-up and shutdown process in the historical operation database is deleted. The model is then retrained using the updated dataset.

[0005] As an optional embodiment of the present invention, the sampling accuracy of the historical running data may be no less than the second level.

[0006] As an optional embodiment of the present invention, the training process of the neural network prediction model may be as follows: A feedforward neural network model was established, using three hidden layers, each containing 32, 24, and 16 neurons respectively. Use the Bayesian regularization training algorithm ('trainbr') to prevent overfitting; Configure neural network parameters: The data was randomly partitioned, with the training set accounting for a certain proportion. (0< <1), the validation set ratio is (0< <1, and + <1), the proportion of the test set is The maximum number of training rounds is set to N, and the mean squared error is used as the loss function. Set the activation function, where the hidden layer uses the hyperbolic tangent function and the output layer uses a linear function.

[0007] As an optional embodiment of the present invention, the time span of the historical operation data is not less than one year; The atmospheric environment data includes atmospheric temperature, pressure, and humidity; the operating status data includes gas turbine power and speed; and the monitoring data includes gas turbine temperature, pressure, vibration, flow rate, and valve opening.

[0008] As an optional embodiment of the present invention, the boundary condition characteristic parameters may include atmospheric environmental data and monitoring data, and the state characteristic parameters may include operational state data.

[0009] As an optional embodiment of the present invention, the sampling accuracy of the historical operating data is consistent with the sampling accuracy of the real-time operating data during the process of collecting historical operating data and real-time operating data of the gas turbine.

[0010] On the other hand, the present invention also provides a gas turbine condition intelligent early warning system based on dynamic neural network, which adopts the gas turbine condition intelligent early warning method based on dynamic neural network described above. The system includes a data acquisition and preprocessing module, a condition analysis and early warning module, and a result display module. The data acquisition and preprocessing module is used to acquire gas turbine operating data and perform normalization and feature engineering data preprocessing. The state analysis and early warning module is used to execute the intelligent early warning method for gas turbine state based on dynamic neural network. The results display module is used to display the gas turbine condition analysis results, including the predicted values, actual values, and relative deviations of the operating condition characteristic parameters.

[0011] On the other hand, the present invention also provides a computer-readable storage medium, comprising: A memory on which computer programs are stored; A processor is configured to execute the program in the memory to implement the aforementioned intelligent early warning method for gas turbine status based on dynamic neural networks.

[0012] On the other hand, the present invention also provides an electronic device, comprising: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the intelligent early warning method for gas turbine status based on dynamic neural networks as described above.

[0013] The beneficial effects of this invention are as follows: This invention achieves real-time monitoring and early warning of gas turbine operating status through steps such as historical data collection, feature parameter screening, neural network prediction model construction and training, real-time data collection, prediction and analysis of key operating status feature parameters of the gas turbine, and dynamic updating of the historical database and neural network prediction model of the gas turbine. Specifically, this invention first collects nearly one year of historical operating data of the gas turbine and constructs a historical operating database of the gas turbine, ensuring the integrity of the gas turbine operating data; then, based on the historical operating data of the gas turbine, it screens key feature parameters to characterize the operating conditions of the gas turbine, achieving feature dimensionality reduction of the gas turbine operating status; next, it establishes and trains a neural network prediction model to predict key operating parameters of the gas turbine in real time, and analyzes whether the gas turbine operating status is normal by comparing the deviation between the real-time values ​​and the predicted values; finally, it dynamically updates the historical operating database and neural network prediction model when the gas turbine is shut down. This invention, through boundary condition-operating status feature parameter classification and dynamic modeling and training of the neural network prediction model, solves the limitations of single-parameter early warning in traditional gas turbine condition monitoring and the lack of a prediction model update mechanism, and can provide technical means for intelligent operation and maintenance of gas turbines.

[0014] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0015] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of an intelligent early warning method for gas turbine status based on dynamic neural networks provided by the present invention; Figure 2 This is a structural block diagram of a gas turbine condition intelligent early warning system based on dynamic neural networks provided by the present invention. Detailed Implementation

[0016] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0017] Example 1 like Figure 1 As shown, a method for intelligent early warning of gas turbine status based on dynamic neural networks is disclosed, the method comprising: S1. Collect recent historical operating data of the gas turbine (the time span of historical operating data shall not be less than 1 year), including atmospheric environmental data such as atmospheric temperature, pressure and humidity, operating status data such as gas turbine power and speed, and monitoring data such as gas turbine temperature, pressure, vibration, flow rate and valve opening, and construct a historical operating database of the gas turbine.

[0018] In this embodiment, the historical operation data spans one year.

[0019] S2. Divide the gas turbine operating data into boundary condition characteristic parameters and state characteristic parameters. The boundary condition characteristic parameters are parameters that characterize the boundary conditions of the gas turbine operation, and the state characteristic parameters are parameters that characterize the operating state of the gas turbine.

[0020] S3. Using boundary condition feature parameters as input and state feature parameters as output, take all the operating data of a "start-up-run-shutdown" process as a set of samples, and use samples from the gas turbine historical operation database as the training set to train the neural network prediction model.

[0021] S4. Collect real-time operating data of the gas turbine, including atmospheric environmental data such as atmospheric temperature, pressure, and humidity, operating status data such as gas turbine power and speed, and monitoring data such as gas turbine temperature, pressure, vibration, flow rate, and valve opening.

[0022] It should be noted in step S4 that in the real-time acquisition of gas turbine operating data, the sampling accuracy of the real-time operating data should be consistent with the sampling accuracy of the historical operating data in S1.

[0023] S5. Using a trained neural network prediction model, predict the gas turbine state characteristic parameters during the "start-up-operation-shutdown" process, calculate the deviation between the predicted value and the actual value, and consider the current gas turbine operating state to be abnormal when the deviation exceeds the set threshold.

[0024] S6. After the gas turbine is shut down, the data from this operation is added to the historical operation database, and the data from the earliest start-up and shutdown process in the historical operation database is deleted. The model is then retrained using the updated dataset.

[0025] In this embodiment, the sampling accuracy of the historical running data in step S1 is on the order of seconds.

[0026] In this embodiment, the boundary condition characteristic parameters mentioned in step S2 include atmospheric environmental data and monitoring data, such as atmospheric temperature, atmospheric pressure, atmospheric humidity, unit speed, IGV opening, etc. The operating status characteristic parameters include operating status data, such as compressor efficiency, combustion efficiency, turbine efficiency, gas turbine power, speed, etc.

[0027] As an optional embodiment of the present invention, the training process of the neural network prediction model in step S3 is optionally as follows: S301. Establish a feedforward neural network model using three hidden layers, each containing 32, 24, and 16 neurons respectively.

[0028] S302. Use the Bayesian regularization training algorithm ('trainbr') to prevent overfitting.

[0029] S303, Configure neural network parameters: The data was randomly partitioned, with the training set accounting for 0.85, the validation set for 0.1, and the test set for 0.05. The maximum number of training iterations was set to 1000, and the mean squared error was used as the loss function.

[0030] S304. Set the activation function, where the hidden layer uses the hyperbolic tangent function and the output layer uses a linear function.

[0031] Example 2 like Figure 2 As shown, a gas turbine condition intelligent early warning system based on dynamic neural network is used to implement the gas turbine condition intelligent early warning method based on dynamic neural network described in Example 1.

[0032] The system includes a data acquisition and preprocessing module, a status analysis and early warning module, and a result display module. The data acquisition and preprocessing module collects gas turbine operating data and performs data preprocessing tasks such as normalization and feature engineering. The status analysis and early warning module executes the intelligent early warning method for gas turbine status based on a dynamic neural network. The result display module displays the gas turbine status analysis results, including predicted values, actual values, and relative deviations of operating status characteristic parameters.

[0033] In this embodiment, the principle of intelligent early warning for gas turbine status based on dynamic neural networks is as follows: The system first acquires the operating data of the gas turbine through the data acquisition module. Subsequently, the data preprocessing module cleans, filters, and extracts features from these raw signals to eliminate noise and interference, and extracts useful feature information for screening similar operating conditions. The preprocessed signals are sent to the status analysis and early warning module, which uses an intelligent early warning method for gas turbine status based on dynamic neural networks to predict the gas turbine status characteristic parameters in real time during the "start-up-operation-shutdown" process. By comparing the deviation between the predicted values ​​and the actual values, the module assesses whether there are any abnormalities in the gas turbine's operating status. Finally, the results display module presents the gas turbine status analysis results in an intuitive way, including the predicted values, actual values, and relative deviations of the operating status characteristic parameters.

[0034] Example 3 A computer-readable storage medium comprising: A memory that stores computer programs.

[0035] A processor is configured to execute the program in the memory to implement the intelligent early warning method for gas turbine status based on a dynamic neural network as described in Embodiment 1. The computer-readable medium may be included in the apparatus, device, or system of the present invention, or it may exist independently.

[0036] The computer-readable storage medium may be any tangible medium that contains or stores a program, and may be an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device. More specific examples include, but are not limited to, electrical connections having one or more wires, portable computer disks, hard disks, optical fibers, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0037] The computer-readable storage medium may also include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code, specific examples of which include, but are not limited to, electromagnetic signals, optical signals, or any suitable combination thereof.

[0038] This disclosure also provides an electronic device including one or more processors.

[0039] A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the intelligent early warning method for gas turbine status based on dynamic neural networks as described above.

[0040] It should be noted here that the electronic device in the embodiments of this disclosure may also include an input device and an output device. The processor, memory, input device, and output device may be connected via a bus or other means, without specific limitations herein.

[0041] As a computer-readable storage medium, memory can be used to store software programs, computer-executable programs, and various modules, such as the program or module corresponding to the intelligent early warning method for gas turbine status based on dynamic neural networks in this disclosure. The processor executes various functional applications and data processing of the electronic device by running the software programs or modules stored in the memory.

[0042] Input devices can be used to receive input digital numbers or signals. These signals can be key signals related to user settings and function control of the device / terminal / server. Output devices can include display devices such as screens.

[0043] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for intelligent early warning of gas turbine status based on dynamic neural networks, characterized in that, The method includes: Collect historical operating data of gas turbines, including atmospheric environment data, operating status data and monitoring data, and construct a historical operating database of gas turbines; Gas turbine operating data is divided into boundary condition characteristic parameters and state characteristic parameters. Boundary condition characteristic parameters are parameters that characterize the boundary conditions of gas turbine operation, while state characteristic parameters are parameters that characterize the operating state of gas turbine. Using boundary condition feature parameters as input and state feature parameters as output, the entire operating data of a single "start-up-run-shutdown" process is used as a set of samples, and samples from the gas turbine historical operation database are used as the training set to train a neural network prediction model. Collect real-time operating data of the gas turbine, including atmospheric environment data, operating status data, and monitoring data; The trained neural network prediction model is used to predict the gas turbine state characteristic parameters during the "start-up-operation-shutdown" process. The deviation between the predicted value and the actual value is calculated. When the deviation exceeds the set threshold, the current gas turbine operating state is considered abnormal. After the gas turbine is shut down, the data from this operation is added to the historical operation database, and the data from the earliest start-up and shutdown process in the historical operation database is deleted. The model is then retrained using the updated dataset.

2. The intelligent early warning method for gas turbine status based on dynamic neural networks according to claim 1, characterized in that, The sampling accuracy of the historical operating data is no less than the second level.

3. The intelligent early warning method for gas turbine status based on dynamic neural networks according to claim 1, characterized in that, The training process of the neural network prediction model is as follows: A feedforward neural network model was established, using three hidden layers, each containing 32, 24, and 16 neurons respectively. Use the Bayesian regularization training algorithm ('trainbr') to prevent overfitting; Configure neural network parameters: The data was randomly partitioned, with the training set accounting for a certain proportion. (0< <1), the validation set ratio is (0< <1, and + <1), the proportion of the test set is The maximum number of training rounds is set to N, and the mean squared error is used as the loss function. Set the activation function, where the hidden layer uses the hyperbolic tangent function and the output layer uses a linear function.

4. The intelligent early warning method for gas turbine status based on dynamic neural network according to claim 1, characterized in that, The time span of the historical operational data shall not be less than one year; The atmospheric environment data includes atmospheric temperature, pressure, and humidity; the operating status data includes gas turbine power and speed; and the monitoring data includes gas turbine temperature, pressure, vibration, flow rate, and valve opening.

5. The intelligent early warning method for gas turbine status based on dynamic neural networks according to claim 1, characterized in that, The boundary condition characteristic parameters include atmospheric environmental data and monitoring data, and the state characteristic parameters include operational state data.

6. The intelligent early warning method for gas turbine status based on dynamic neural networks according to any one of claims 1 to 5, characterized in that, During the collection of historical and real-time operating data of the gas turbine, the sampling accuracy of the historical operating data is consistent with that of the real-time operating data.

7. A gas turbine condition intelligent early warning system based on dynamic neural networks, characterized in that, The intelligent early warning method for gas turbine status based on dynamic neural networks according to any one of claims 1 to 6, wherein the system includes a data acquisition and preprocessing module, a status analysis and early warning module, and a result display module; The data acquisition and preprocessing module is used to acquire gas turbine operating data and perform normalization and feature engineering data preprocessing. The state analysis and early warning module is used to execute the intelligent early warning method for gas turbine state based on dynamic neural network. The results display module is used to display the gas turbine condition analysis results, including the predicted values, actual values, and relative deviation information of the operating condition characteristic parameters.

8. A computer-readable storage medium, comprising: A memory on which computer programs are stored; A processor is configured to execute the program in the memory to implement the intelligent early warning method for gas turbine status based on a dynamic neural network according to any one of claims 1 to 6.

9. An electronic device, characterized in that, include: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the intelligent early warning method for gas turbine status based on dynamic neural networks according to any one of claims 1 to 6.

Images