An energy industry gas turbine intelligent early warning system and method

Abstract

This invention relates to the field of power digitalization technology, providing an intelligent early warning system and method for gas turbines in the energy industry. The system comprises: a cloud layer, consisting of an infrastructure layer, a cloud capability layer, a cloud node core layer, and a cloud-edge interaction management layer; the cloud node core layer consists of a data processing module, a model building module, a training and optimization module, a deployment management module, and an integrated training and inference engine; the cloud-edge interaction management layer consists of an edge node management service module, a model push service module, a cloud heartbeat detection service module, and a push result processing service module; and multiple edge terminals, each consisting of an edge terminal heartbeat detection service module, a model receiving service module, a monitoring service module, a registration service module, a model start / stop service module, and an inference result push service module. The system and method of this invention better reflect the actual conditions of gas turbines in the field, have high model early warning accuracy, and can save on gas turbine operation and maintenance costs.

Description

Technical Field

[0001] This invention relates to the field of power digitalization technology, and in particular to an intelligent early warning system and method for gas turbines in the energy industry. Background Technology

[0002] Gas-fired power generation, with its unique characteristics of being clean, stable, and flexible, is favored by power generation manufacturers. However, the high operation and maintenance costs of gas turbines are a significant obstacle to the development of gas-fired power generation. How to leverage next-generation digital technologies such as artificial intelligence to achieve a new leap in operation and maintenance costs and management decisions amidst the digital revolution has become the most pressing challenge for gas-fired power generation.

[0003] Currently, the management of gas turbine power generation mainly adopts centralized processing, with all processing taking place in the cloud. Training and inference are performed in the cloud. In practical applications, especially during inference, data needs to be sent from the edge to the cloud for inference. Due to the geographical isolation between the edge and the cloud, the following shortcomings exist: 1. The network latency is relatively large, making real-time inference impossible; 2. Security cannot be guaranteed, and real-time accurate early warning cannot be provided.

[0004] Therefore, how to provide a real-time and accurate early warning monitoring method for gas turbines has become an urgent technical problem to be solved. Summary of the Invention

[0005] In view of this, the present invention constructs a cloud-edge collaborative architecture, forming a closed loop of data processing, operator development, cloud training, edge inference, and data feedback, effectively supporting intelligent early warning for gas turbines.

[0006] On one hand, the present invention provides an intelligent early warning system for gas turbines in the energy industry, comprising:

[0007] The cloud platform consists of an infrastructure layer, a cloud capability layer, a cloud node core layer, and a cloud-edge interaction management layer. The infrastructure layer, constructed using Kubernetes, allocates hardware resources when the edge model start / stop service module launches the Docker model image. The cloud node core layer comprises a data processing module, a model building module, a training and optimization module, a deployment management module, and an integrated training and inference engine. It is used to build models based on gas turbine data and to train, evaluate, and test these models to obtain qualified training models. The cloud-edge interaction management layer consists of an edge node management service module, a model push service module, a cloud heartbeat detection service module, and a push result processing service module.

[0008] Multiple edge terminals, each of which consists of an edge terminal heartbeat detection service module, a model receiving service module, a monitoring service module, a registration service module, a model start / stop service module, and an inference result push service module.

[0009] On the other hand, the present invention provides an intelligent early warning method for gas turbines in the energy industry, comprising:

[0010] Step S1: Build a model based on gas turbine data, and train, evaluate and test the model to obtain a qualified training model;

[0011] Step S2: Register the edge device to the remote device through the registration service module;

[0012] Step S3: Use the edge heartbeat detection service module and the cloud heartbeat detection service module to maintain the heartbeat connection between the edge and the cloud;

[0013] Step S4: Push the tested model from the cloud to the edge;

[0014] Step S5: The edge device performs inference using the received model and returns the inference results to the cloud.

[0015] Furthermore, step S1 of the intelligent early warning method for gas turbines in the energy industry of the present invention includes:

[0016] Step S11: The data processing module is used to tag, deduplicatize, fill in missing values, filter fields, and sort the environmental and operational data of the gas turbine to obtain model data, and then send the model data to the corresponding topic in the Kafka in the data processing module.

[0017] Step S12: Use the model building module to extract features from the model data, combine the two operators of device health and device early warning, and build the model.

[0018] Step S13: Using an integrated training and inference engine, based on the features extracted in step S12, subscribe to the corresponding gas turbine model data from the data processing module, adjust the training parameters, perform distributed training on the model, and obtain the training model.

[0019] Step S14: Use the training and optimization module to evaluate the alarm accuracy and recall of the training model. If the alarm accuracy is lower than 90% or the recall is lower than 92%, re-extract features and re-build and train the training model.

[0020] Step S15: Use the deployment management module to test the training models that have passed the evaluation, obtain the training models that have passed the test, and rebuild, train, evaluate and test the models that have not passed the test.

[0021] Furthermore, in step S15 of the intelligent early warning method for gas turbines in the energy industry of the present invention, the deployment management module is used to test the training model that has passed the evaluation to obtain a training model that has passed the test. This includes: sending model input parameters to the training model that has passed the evaluation using the deployment management module; if an inference result is returned, the training module is determined to have passed the test.

[0022] Furthermore, step S2 of the intelligent early warning method for gas turbines in the energy industry of the present invention includes:

[0023] Step S21: Use the edge node management service module to assign codes to the edge devices;

[0024] Step S22: Each edge terminal's registration service module sends a registration request to the edge terminal management service module, carrying the corresponding code;

[0025] Step S23: The edge node management service module receives and verifies the registration request, and at the same time obtains the AK and SK information of each edge node.

[0026] Furthermore, step S3 of the intelligent early warning method for gas turbines in the energy industry of the present invention includes: the edge heartbeat detection service module and the cloud heartbeat detection service module communicate once every 1 minute. If the communication fails, the communication is attempted 3 times consecutively. If the communication still fails, the communication is attempted again after an interval of 5 minutes. If the communication still fails, the connection between the edge heartbeat detection service module and the cloud heartbeat detection service module is disconnected, and an alarm is sent at the same time.

[0027] Furthermore, step S4 of the intelligent early warning method for gas turbines in the energy industry of the present invention includes:

[0028] Step S41: The model push service module encrypts the model file and model parameters in the cloud and sends them to the model receiving service module via HTTPS protocol. The model parameters include the underlying dependency framework information required for edge model packaging and the AK and SK information obtained by the edge node management module.

[0029] Step S42: The model receiving service module receives the model file and model parameters, authenticates the AK and SK, decrypts the model file, and packages it into a Docker model image.

[0030] Furthermore, step S5 of the intelligent early warning method for gas turbines in the energy industry of the present invention includes:

[0031] Step S51: Use the edge model start / stop service module to start the Docker model image according to the hardware resources allocated by Kubernetes, and deploy the started Docker model image;

[0032] Step S52: Subscribe to model data from the corresponding topic in the data processing module, perform inference based on the subscribed model data, and obtain the inference result;

[0033] Step S53: Use the inference result push service module to push the inference result with attached AK and SK information to the push result processing service module in the cloud;

[0034] Step S54: The cloud push result processing service module tags the received inference results and sends them to the data processing module.

[0035] Furthermore, in step S51 of the intelligent early warning method for gas turbines in the energy industry of the present invention, the deployment of the started Docker model image includes: deploying the started Docker model image by means of single instance deployment or multi-instance distributed deployment.

[0036] Furthermore, the intelligent early warning method for gas turbines in the energy industry of the present invention also includes: using a monitoring service module to monitor the operation and inference service call status of edge nodes.

[0037] The intelligent early warning system and method for gas turbines in the energy industry of this invention have the following beneficial effects:

[0038] 1. Model building, training, evaluation, and testing are performed in the cloud, while inference is performed at the edge. The cloud-trained model and edge data are seamlessly integrated, achieving collaboration and data unification between the cloud and the edge. This effectively reduces network latency for inference and enables real-time inference and early warning.

[0039] 2. Secure transmission of model files is achieved across the entire chain through national cryptographic encryption and cloud-edge authentication. By pushing weight files and packaging model images at the edge, the amount of communication between the cloud and the edge is effectively reduced. By employing edge-based inference and alarm processing, personnel and devices at the edge can quickly handle alarm information, which helps to avoid the risks caused by untimely processing.

[0040] 3. Enable data backflow training: After the model at the edge performs inference, the gas turbine data flows back to the cloud data processing module. After cleaning and labeling, it can be used to train the cloud model.

[0041] Based on the above-mentioned beneficial effects, the intelligent early warning system for gas turbines in the energy industry of this invention is more consistent with the actual situation of gas turbines on site, has high model early warning accuracy, and can save gas turbine operation and maintenance costs. Attached Figure Description

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is an architectural diagram of an intelligent early warning system for gas turbines in the energy industry, which is an exemplary first embodiment of the present invention.

[0044] Figure 2 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry, which is an exemplary second embodiment of the present invention.

[0045] Figure 3 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry, which is an exemplary third embodiment of the present invention.

[0046] Figure 4 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry, which is an exemplary fourth embodiment of the present invention.

[0047] Figure 5 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry, which is an exemplary sixth embodiment of the present invention.

[0048] Figure 6 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry, which is an exemplary seventh embodiment of the present invention. Detailed Implementation

[0049] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0050] It should be noted that, in the absence of conflict, the following embodiments and features can be combined with each other; and, based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0051] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0052] The following definitions of terms used in the embodiments are as follows:

[0053] AK: Access Key Id, used to identify the user.

[0054] SK: Secret Access Key, is the key used by the user to encrypt and verify the authentication string. SK must be kept secret.

[0055] Figure 1This is an architectural diagram of an intelligent early warning system for gas turbines in the energy industry according to an exemplary first embodiment of the present invention, such as... Figure 1 As shown, the system in this embodiment includes:

[0056] The cloud platform consists of an infrastructure layer, a cloud capability layer, a cloud node core layer, and a cloud-edge interaction management layer. The infrastructure layer, constructed using Kubernetes, allocates hardware resources when the edge model start / stop service module launches the Docker model image. The cloud node core layer comprises a data processing module, a model building module, a training and optimization module, a deployment management module, and an integrated training and inference engine. It is used to build models based on gas turbine data and to train, evaluate, and test these models to obtain qualified training models. The cloud-edge interaction management layer consists of an edge node management service module, a model push service module, a cloud heartbeat detection service module, and a push result processing service module.

[0057] Multiple edge terminals, each of which consists of an edge terminal heartbeat detection service module, a model receiving service module, a monitoring service module, a registration service module, a model start / stop service module, and an inference result push service module.

[0058] In practical applications, the system in this embodiment performs model building, training, evaluation, and testing in the cloud, and inference at the edge. It seamlessly connects cloud-trained models with edge data, achieving collaboration and data unification between the cloud and edge, effectively reducing network latency for inference and enabling real-time inference and early warning. By employing edge-based inference and alarm processing, personnel and devices at the edge can quickly handle alarm information, mitigating the risks associated with delayed responses.

[0059] Figure 2 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry according to an exemplary second embodiment of the present invention. The method of this embodiment includes:

[0060] Step S1: Build a model based on gas turbine data, and train, evaluate and test the model to obtain a qualified training model;

[0061] Step S2: Register the edge device to the remote device through the registration service module;

[0062] Step S3: Use the edge heartbeat detection service module and the cloud heartbeat detection service module to maintain the heartbeat connection between the edge and the cloud;

[0063] Step S4: Push the tested model from the cloud to the edge;

[0064] Step S5: The edge device performs inference using the received model and returns the inference results to the cloud.

[0065] The intelligent early warning method for gas turbines in the energy industry in this embodiment also includes: using a monitoring service module to monitor the operation and inference service call status of edge nodes.

[0066] Figure 3 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry according to an exemplary third embodiment of the present invention. This embodiment is... Figure 2 Preferred embodiments of the method shown are as follows: Figure 3 As shown, step S1 of the method in this embodiment includes:

[0067] Step S11: The data processing module is used to tag, deduplicatize, fill in missing values, filter fields, and sort the environmental and operational data of the gas turbine to obtain model data, and then send the model data to the corresponding topic in the Kafka in the data processing module.

[0068] Step S12: Use the model building module to extract features from the model data, combine the two operators of device health and device early warning, and build the model.

[0069] Step S13: Using an integrated training and inference engine, based on the features extracted in step S12, subscribe to the corresponding gas turbine model data from the data processing module, adjust the training parameters, perform distributed training on the model, and obtain the training model.

[0070] Step S14: Use the training and optimization module to evaluate the alarm accuracy and recall of the training model. If the alarm accuracy is lower than 90% or the recall is lower than 92%, re-extract features and re-build and train the training model.

[0071] Step S15: Use the deployment management module to test the training models that have passed the evaluation, obtain the training models that have passed the test, and rebuild, train, evaluate and test the models that have not passed the test.

[0072] In step S15 of the method in this embodiment, the deployment management module is used to test the training model that has passed the evaluation to obtain a training model that has passed the test. This includes: sending model input parameters to the training model that has passed the evaluation using the deployment management module; if an inference result is returned, the training module is determined to have passed the test.

[0073] Figure 4 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry according to an exemplary fourth embodiment of the present invention. This embodiment is... Figure 2 Preferred embodiments of the method shown are as follows: Figure 4 As shown, step S2 of the method in this embodiment includes:

[0074] Step S21: Use the edge node management service module to assign codes to the edge devices;

[0075] Step S22: Each edge terminal's registration service module sends a registration request to the edge terminal management service module, carrying the corresponding code;

[0076] Step S23: The edge node management service module receives and verifies the registration request, and at the same time obtains the AK and SK information of each edge node.

[0077] The fifth exemplary embodiment of the present invention provides an intelligent early warning method for gas turbines in the energy industry. This embodiment is... Figure 2 In a preferred embodiment of the method shown, step S3 of the method in this embodiment includes: the edge heartbeat detection service module and the cloud heartbeat detection service module communicate once every 1 minute. If the communication fails, the communication is attempted 3 times consecutively. If the communication still fails, the communication is attempted again after an interval of 5 minutes. If the communication still fails, the connection between the edge heartbeat detection service module and the cloud heartbeat detection service module is disconnected, and an alarm is sent at the same time.

[0078] Figure 5 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry according to an exemplary sixth embodiment of the present invention. This embodiment is... Figure 2 Preferred embodiments of the method shown are as follows: Figure 5 As shown, step S4 of the method in this embodiment includes:

[0079] Step S41: The model push service module encrypts the model file and model parameters in the cloud and sends them to the model receiving service module via HTTPS protocol. The model parameters include the underlying dependency framework information required for edge model packaging and the AK and SK information obtained by the edge node management module.

[0080] Step S42: The model receiving service module receives the model file and model parameters, authenticates the AK and SK, decrypts the model file, and packages it into a Docker model image.

[0081] In practical applications, this embodiment uses national cryptographic encryption and cloud-edge authentication to transmit model files, achieving secure transmission across the entire link. By pushing weight files and packaging model images at the edge, the amount of communication between the cloud and the edge is effectively reduced.

[0082] Figure 6 This is a flowchart of an intelligent early warning method for gas turbines in the energy industry according to an exemplary seventh embodiment of the present invention. This embodiment is... Figure 2 Preferred embodiments of the method shown are as follows: Figure 6 As shown, step S5 of the method in this embodiment includes:

[0083] Step S51: Use the edge model start / stop service module to start the Docker model image according to the hardware resources allocated by Kubernetes, and deploy the started Docker model image;

[0084] Step S52: Subscribe to model data from the corresponding topic in the data processing module, perform inference based on the subscribed model data, and obtain the inference result;

[0085] Step S53: Use the inference result push service module to push the inference result with attached AK and SK information to the push result processing service module in the cloud;

[0086] Step S54: The cloud push result processing service module tags the received inference results and sends them to the data processing module.

[0087] In step S51 of the method in this embodiment, the deployment of the started Docker model image includes: deploying the started Docker model image using a single-instance deployment or a multi-instance distributed deployment method.

[0088] In practical applications, the method of this embodiment can realize data backflow training. After the model at the edge performs inference, the gas turbine data flows back to the cloud data processing module. After cleaning and labeling, it can be used for training the cloud model.

[0089] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent early warning of gas turbines in the energy industry, characterized in that, The method includes: Step S1: Construct a model based on gas turbine data, and train, evaluate, and test the model to obtain a qualified trained model, including: Step S11: The data processing module is used to tag, deduplicatize, fill in missing values, filter fields, and sort the environmental and operational data of the gas turbine to obtain model data, and then send the model data to the corresponding topic in the Kafka in the data processing module. Step S12: Use the model building module to extract features from the model data, combine the two operators of device health and device early warning, and build the model. Step S13: Using an integrated training and inference engine, based on the features extracted in step S12, subscribe to the corresponding gas turbine model data from the data processing module, adjust the training parameters, perform distributed training on the model, and obtain the training model. Step S14: Use the training and optimization module to evaluate the alarm accuracy and recall of the training model. If the alarm accuracy is lower than 90% or the recall is lower than 92%, re-extract features and re-build and train the training model. Step S15: Use the deployment management module to test the training models that have passed the evaluation, obtain the training models that have passed the test, and rebuild, train, evaluate and test the models that have not passed the test. Step S2: Register the edge device to the remote device through the registration service module, including: Step S21: Use the edge node management service module to assign codes to the edge devices; Step S22: Each edge terminal's registration service module sends a registration request to the edge terminal management service module, carrying the corresponding code; Step S23: The edge node management service module receives and verifies the registration request, and at the same time obtains the AK and SK information of each edge node; Step S3: Maintain the heartbeat connection between the edge and the cloud using the edge heartbeat detection service module and the cloud heartbeat detection service module, including: the edge heartbeat detection service module and the cloud heartbeat detection service module communicate once every 1 minute. If the communication fails, try to communicate 3 times consecutively. If the communication still fails, wait 5 minutes before trying to communicate again. If the communication still fails, disconnect the connection between the edge heartbeat detection service module and the cloud heartbeat detection service module and send an alarm. Step S4: Push the tested model from the cloud to the edge, including: Step S41: The model push service module encrypts the cloud model file and model parameters and sends them to the model receiving service module via HTTPS protocol. The model parameters include the underlying dependency framework information required for edge model packaging and the AK and SK information obtained by the edge node management module. Step S42: The model receiving service module receives the model file and model parameters, authenticates the AK and SK, decrypts the model file, and packages it into a Docker model image; Step S5: The edge device performs inference using the received model and returns the inference results to the cloud, including: Step S51: Use the edge model start / stop service module to start the Docker model image according to the hardware resources allocated by Kubernetes, and deploy the started Docker model image; Step S52: Subscribe to model data from the corresponding topic in the data processing module, perform inference based on the subscribed model data, and obtain the inference result; Step S53: Use the inference result push service module to push the inference result with attached AK and SK information to the push result processing service module in the cloud; Step S54: The cloud push result processing service module tags the received inference results and sends them to the data processing module.

2. The intelligent early warning method for gas turbines in the energy industry according to claim 1, characterized in that, In step S15, the deployment management module is used to test the training model that has passed the evaluation to obtain a training model that has passed the test. This includes: sending model input parameters to the training model that has passed the evaluation using the deployment management module; if an inference result is returned, the training model is determined to have passed the test.

3. The intelligent early warning method for gas turbines in the energy industry according to claim 1, characterized in that, In step S51, the launched Docker model image is deployed, including deploying the launched Docker model image using single-instance deployment or multi-instance distributed deployment.

4. The intelligent early warning method for gas turbines in the energy industry according to claim 1, characterized in that, The method also includes: using a monitoring service module to monitor the operation and inference service calls of edge nodes.

5. An intelligent early warning system for gas turbines in the energy industry, characterized in that, The system, when running, implements the steps of the intelligent early warning method for gas turbines in the energy industry as described in any one of claims 1-4, wherein the system includes: The cloud platform consists of an infrastructure layer, a cloud capability layer, a cloud node core layer, and a cloud-edge interaction management layer. The infrastructure layer, constructed using Kubernetes, allocates hardware resources when the edge model start / stop service module launches the Docker model image. The cloud node core layer comprises a data processing module, a model building module, a training and optimization module, a deployment management module, and an integrated training and inference engine. It is used to build models based on gas turbine data and to train, evaluate, and test these models to obtain qualified training models. The cloud-edge interaction management layer consists of an edge node management service module, a model push service module, a cloud heartbeat detection service module, and a push result processing service module. Multiple edge terminals, each of which consists of an edge terminal heartbeat detection service module, a model receiving service module, a monitoring service module, a registration service module, a model start / stop service module, and an inference result push service module.

Images