Generator State Detection Method and System Based on Convolutional Neural Network

Abstract

This application provides a generator state detection method and system based on convolutional neural networks. The method includes: collecting state information of the generator's cooling system and the corresponding health level to obtain initial training samples; performing data missingness processing, outlier detection, data reconstruction, and sample assignment on the initial training samples to obtain intermediate training samples; classifying the intermediate training samples and performing secondary classification on the classified intermediate training samples using a preset classification code to obtain target training samples consisting of a training set and a test set; constructing an initial generator state detection model for detecting generator state information, training the initial generator state detection model based on the target training samples to obtain a trained generator state detection model; and performing state detection on the generator's cooling system based on the generator state detection model. This application can improve the efficiency and effectiveness of generator state detection model training.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a generator state detection method and system based on convolutional neural networks. Background Technology

[0002] Generators are important electrical devices widely used in power systems, ships, wind power generation, and other fields. Generator condition monitoring refers to the periodic or real-time monitoring and analysis of generators to ensure their normal operation and prevent malfunctions.

[0003] However, in traditional generator condition monitoring, manual inspection and recording are usually used. This method relies too much on the experience and ability of the inspectors and is affected by many external factors. For example, the condition monitoring standards for different parts of the generator may be different under different operating conditions. The influence of missing data or outlier data can lead to inconsistent and inaccurate test data. As a result, the above methods have problems such as low efficiency, low accuracy and susceptibility to human interference.

[0004] Therefore, a generator state detection method and system based on convolutional neural networks needs to be proposed to solve this technical problem. Summary of the Invention

[0005] The purpose of this invention is to provide a generator state detection method and system based on convolutional neural networks. This application embodiment can perform efficient and accurate state detection operations on the generator based on a state detection model combined with the generator's state under different operating conditions. The specific technical solution is as follows:

[0006] In a first aspect of the present invention, a generator state detection method based on a convolutional neural network is provided, applied to a generator cooling system state detection model, the method comprising:

[0007] The status information of the generator's cooling system and the corresponding health status are collected to obtain the initial training samples;

[0008] The initial training samples are processed by data cleaning, feature extraction, and data classification and fusion to obtain intermediate training samples;

[0009] The intermediate training samples are classified according to three different operating conditions of the generator: no-load power, partial load power, and full load power. The intermediate training samples are then classified a second time using a preset classification code to obtain the target training samples consisting of a training set and a test set.

[0010] An initial generator state detection model is constructed to detect the state information, and the initial generator state detection model is trained based on the target training samples to obtain a trained generator state detection model.

[0011] Optionally, the step of collecting the status information of the generator's cooling system and the corresponding health level of the status information to obtain initial training samples includes:

[0012] Based on the status information, the fault type of the generator's cooling system is obtained;

[0013] Define a corresponding severity level for each fault type, and assign a corresponding health level to each severity level;

[0014] For each health level, an output state corresponding to the initial generator state detection model is matched to obtain the initial training sample.

[0015] Optionally, the process of cleaning, extracting features, and classifying and fusing the initial training samples to obtain intermediate training samples includes:

[0016] If some sample intervals in the initial training samples have missing data, the data in the sample intervals will be supplemented or deleted. If some sample intervals in the initial training samples have outlier data, the three sigma criterion will be used to delete some of the outlier data to obtain the cleaned initial training samples.

[0017] The initial training samples after cleaning are sampled to save the data format for subsequent modeling calls;

[0018] The initial training samples after cleaning are reconstructed by defining the sliding window length, sliding window step size, number of data sequence segments, original current signal slices of the generator cooling system, and original voltage signal slices of the generator cooling system to obtain the intermediate training samples.

[0019] Optionally, classifying the intermediate training samples according to the generator's operating conditions includes:

[0020] The cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature, and average pressure within the sliding window step size are obtained under different operating conditions.

[0021] Optionally, training the initial generator state detection model based on the target training samples includes:

[0022] The cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature, and average pressure within the sliding window step size are input into the initial generator state detection model. The health of the generator's cooling system corresponding to the input data is output. The initial generator state detection model is trained by combining the labels of the input data.

[0023] Optionally, the method further includes:

[0024] When the accuracy of the initial generator state detection model in predicting the health of the generator's cooling system is greater than or equal to a preset threshold, the training is complete, and the trained generator state detection model is obtained.

[0025] In another aspect of the present invention, a model training system for detecting generator status is provided, applied to a generator of a target device, the system comprising:

[0026] The initial training sample acquisition module is used to collect the status information of the generator's cooling system and the health status corresponding to the status information to obtain the initial training sample.

[0027] The sample processing module is used to perform data cleaning, feature extraction, and data classification and fusion on the initial training samples to obtain intermediate training samples.

[0028] The sample classification module is used to classify the intermediate training samples according to three different operating conditions of the generator: no-load power, partial load power, and full load power. The intermediate training samples are then classified a second time using a preset classification code to obtain the target training samples consisting of a training set and a test set.

[0029] The model training module is used to construct an initial generator state detection model for detecting the state information, and to train the initial generator state detection model based on the target training samples to obtain a trained generator state detection model.

[0030] The status detection module predicts whether there is a fault in the generator's cooling system based on the generator status detection model.

[0031] Optionally, the initial training sample acquisition module is further specifically used for:

[0032] Based on the status information, the fault type of the generator's cooling system is obtained;

[0033] Define a corresponding severity level for each fault type, and assign a corresponding health level to each severity level;

[0034] For each health level, an output state corresponding to the initial generator state detection model is matched to obtain the initial training sample.

[0035] In another aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed, implements the steps of the method described above.

[0036] In another aspect of the present invention, a computer device is provided, including a processor, a memory, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described above.

[0037] As can be seen from the above, the application embodiment can train a generator state detection model by using training samples composed of the state information and health status of the generator's cooling system under different operating conditions. This enables the generator to perform powerful fault detection functions based on the generator state detection model, without the need for manual intervention or complex detection operations, thereby improving the training efficiency and accuracy of the generator state detection model. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram illustrating an application scenario of the generator state detection system based on a convolutional neural network provided in this application embodiment;

[0040] Figure 2 This is a schematic flowchart of the generator state detection method based on a convolutional neural network provided in the embodiments of this application;

[0041] Figure 3 This is a schematic diagram of the generator state detection system based on a convolutional neural network provided in an embodiment of this application;

[0042] Figure 4 This is an internal structural diagram of the computer device provided in the embodiments of this application. Detailed Implementation

[0043] To more clearly illustrate the technical solutions of the embodiments in this specification, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of this specification. For those skilled in the art, these drawings can be applied to other similar scenarios without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0044] This application provides a generator state detection method and system based on a convolutional neural network. Please refer to... Figure 1 , Figure 1 This diagram illustrates an application scenario of the generator state detection system based on a convolutional neural network provided in this application. The system may include a terminal 10 and a server 11. The generator state detection method based on a convolutional neural network provided in this application can be implemented through the terminal 10 or the server 11.

[0045] like Figure 1 As shown, terminal 10 and server 11 are connected via a network, such as a wired or wireless network. Terminal 10 may include, but is not limited to, portable terminals such as mobile phones and tablets with various network platform applications installed, as well as fixed terminals such as computers, kiosks, and advertising machines. Server 11 provides users with various business services, including a service push server and a user recommendation server.

[0046] It should be noted that, Figure 1 The schematic diagram of the application scenario of the generator state detection system based on convolutional neural network shown is merely an example. The terminal, server, and application scenario described in the embodiments of this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of the system and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0047] Terminal 10 can be used for:

[0048] The process involves collecting the status information of the generator's cooling system and the corresponding health level to obtain initial training samples. These initial training samples undergo data cleaning, feature extraction, and data classification fusion to obtain intermediate training samples. The intermediate training samples are then classified according to three different operating conditions: no-load power, partial load power, and full load power. A secondary classification is performed on the classified intermediate training samples using a preset classification code to obtain target training samples consisting of a training set and a test set. An initial generator state detection model for detecting the status information is constructed, and this model is trained based on the target training samples to obtain a trained generator state detection model. The steps of the generator state detection method based on a convolutional neural network, executed by terminal 10, can also be executed by server 11.

[0049] Figure 2 This paper illustrates a flowchart of a generator state detection method and system based on a convolutional neural network, as provided in an embodiment of this application. Figure 2 As shown, a generator state detection method and system based on convolutional neural networks includes the following steps:

[0050] Step 201: Collect the status information of the generator's cooling system and the corresponding health status to obtain the initial training samples.

[0051] Feature extraction and processing are performed on the state information and health status to obtain training sample features and labels. Then, machine learning algorithms (such as deep neural networks, support vector machines, etc.) can be used to train the training samples, and the model can be validated and evaluated using a test set. Finally, this model can be used to predict the health status of the generator rotor. This can be expressed as the following formula:

[0052] Sample collection: D = {(x1, y1), (x2, y2), ..., (x...} n y n )}

[0053] Where D is the initial training sample, x n Let y be the feature vector of the nth sample. n Let be the label of the nth sample.

[0054] Feature extraction and processing: X = {f(x1), f(x2), ..., f(x...} n )},Y={y1,y2,...,y n}

[0055] Where f is the feature extraction function, X is the training sample feature, and Y is the training sample label.

[0056] Model training: h = f{θ}(X)

[0057] Where h is the model output and f{θ} is a machine learning model with parameter θ.

[0058] Model validation and evaluation: E(h,Y)

[0059] Here, E is the model's loss function, which can be used to evaluate the accuracy and robustness of the model's prediction results.

[0060] Optionally, step 201 may include:

[0061] Based on the status information, the fault type of the generator's cooling system is obtained;

[0062] Define a corresponding severity level for each fault type, and assign a corresponding health level to each severity level;

[0063] For each health level, an output state corresponding to the initial generator state detection model is matched to obtain the initial training sample.

[0064] In some embodiments, the fault types of the generator's cooling system may include abnormal cooling system temperature and coolant leakage. The corresponding severity levels may be mild, moderate, and severe, each corresponding to a different health level. For example, a health level of 0-60% is considered severe, 61-80% is severe, and 81-100% is mild. Correspondingly, the output states for mild, severe, and severe faults are Normal, Warning, and Error, respectively.

[0065] Step 202: Perform data cleaning, feature extraction, and data classification and fusion on the initial training samples to obtain intermediate training samples.

[0066] Optionally, step 202 may include:

[0067] If some sample intervals in the initial training samples have missing data, the data in the sample intervals will be supplemented or deleted. If some sample intervals in the initial training samples have outlier data, the three sigma criterion will be used to delete some of the outlier data to obtain the cleaned initial training samples.

[0068] The initial training samples after cleaning are sampled to save the data format for subsequent modeling calls;

[0069] The initial training samples after cleaning are reconstructed by defining the sliding window length, sliding window step size, number of data sequence segments, original current signal slices of the generator cooling system, and original voltage signal slices of the generator cooling system to obtain the intermediate training samples.

[0070] In some embodiments, the relevant data of the above intermediate training samples can be represented as:

[0071]

[0072]

[0073]

[0074] Where L is the sliding window length; s is the sliding window step size; and k represents the k-th data sequence segment. w_in,1 ,T w_in,2 ,…,T w_in,k} represents a signal slice for the cooling water inlet temperature; Signal slices for cooling water outlet temperature; {P1,P2,…,P k} represents a slice of the original pressure signal from the cooling system; Δf i This represents a first-order difference operation, Δf i =f i -f i-1 .

[0075] In some embodiments, if outlier data exists in a portion of the initial training samples, the three Sigma criterion is used to remove some of the outlier data to obtain a cleaned initial training sample. Specifically, the selection can be based on the cause of the outlier data and the application purpose, and the three Sigma criterion, or "3σ" criterion, can be applied to process and remove some of the outlier data.

[0076] In some embodiments, the cleaned initial training samples can be sampled to preserve the data format for subsequent modeling calls. For example, the initial training samples can be sampled into an electrical signal format that can be used for subsequent modeling.

[0077] Step 203: Classify the intermediate training samples according to the three different operating conditions of the generator being at no-load power, partial load power, and full load power. Then, perform secondary classification on the classified intermediate training samples using a preset classification code to obtain target training samples composed of training set and test set.

[0078] Optionally, step 203 may include:

[0079] The cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature, and average pressure within the sliding window step size are obtained under different operating conditions.

[0080] In some embodiments, different operating conditions may include three categories of generator operating conditions based on power: no-load, partial-load, and full-load. The dataset can be divided into training and test sets in an 8:2 or 7:3 ratio.

[0081] In some embodiments, the preset classification code can be represented as:

[0082] for j in range(len(data_index)):

[0083] data = data_index[j]

[0084] x = data[:,:-1]

[0085] y = data[:, -1]

[0086] x_train,x_test,y_train,y_test=train_test_split(x,y,test_size=0,2,random_state=66).

[0087] In some embodiments, the peak current, average temperature, voltage fluctuation characteristics, and generator cooling system temperature within the sliding window step can be expressed as:

[0088]

[0089] The average pressure is P. mean The average temperature of the cooling water inlet is T. w_in,mean The average temperature at the cooling water outlet is T. w_out,mean .

[0090] Step 204: Construct an initial generator state detection model for detecting the state information, and train the initial generator state detection model based on the target training samples to obtain a trained generator state detection model.

[0091] Optionally, step 204 may also include:

[0092] The cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature, and average pressure within the sliding window step size are input into the initial generator state detection model. The health of the generator's cooling system corresponding to the input data is output. The initial generator state detection model is trained by combining the labels of the input data.

[0093] Optionally, the method of this application further includes:

[0094] When the accuracy of the initial fault detection model in predicting the health of the generator's cooling system is greater than or equal to a preset threshold, the training is complete, and the trained target fault detection model is obtained.

[0095] In some embodiments, the input data and output data can be represented as:

[0096]

[0097]

[0098] in, y cl ∈[0,100%], X cl Y is the input for the generator condition detection model. cl Output a health label for the generator condition detection model, where g represents the g-th sample.

[0099] After completing the dataset preprocessing, the fault data are mixed and then divided into training and test sets according to a specified ratio. During model training, the output of the fault detection model is the health status of the tested component, and the predicted state of the generator's cooling system is based on the corresponding health status. Finally, the initial fault detection model is trained and validated. Generally, if the validation accuracy reaches a preset threshold, such as 98% or 99%, the initial fault detection model can be considered usable, and the target fault detection model is obtained.

[0100] As can be seen from the above, the application embodiment can train a fault detection model by using training samples composed of the state information and health status of the generator's cooling system under different operating conditions. Thus, based on the fault detection model, the generator's powerful fault detection function can be improved without manual intervention or complex detection operations, thereby enhancing the efficiency and accuracy of the generator's detection model.

[0101] To implement the above-described method embodiments, this application also provides a generator state detection system based on a convolutional neural network. Figure 3 This illustration shows a schematic diagram of a generator state detection system based on a convolutional neural network according to an embodiment of this application. The system includes:

[0102] The initial training sample acquisition module 301 is used to collect the status information of the generator's cooling system and the health status corresponding to the status information to obtain the initial training sample.

[0103] The sample processing module 302 is used to perform data cleaning, feature extraction and data classification fusion on the initial training samples to obtain intermediate training samples;

[0104] The sample classification module 303 is used to classify the intermediate training samples according to three different operating conditions of the generator being at no-load power, partial load power, and full load power. The intermediate training samples after classification are further classified by a preset classification code to obtain the target training samples composed of the training set and the test set.

[0105] The model training module 304 is used to construct an initial generator state detection model for detecting the state information, and to train the initial generator state detection model based on the target training sample to obtain a trained generator state detection model.

[0106] The status detection module 305 predicts whether there is a fault in the generator's cooling system based on the generator status detection model.

[0107] Optionally, the initial training sample acquisition module 302 is further specifically used for:

[0108] Based on the status information, the fault type of the generator's cooling system is obtained;

[0109] Define a corresponding severity level for each fault type, and assign a corresponding health level to each severity level;

[0110] For each health level, an output state corresponding to the initial generator state detection model is matched to obtain the initial training sample.

[0111] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the modules / units / subunits / components in the above-described system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0112] As can be seen from the above, the application embodiment can train a fault detection model by using training samples composed of the state information and health status of the generator's cooling system under different operating conditions. Thus, based on the fault detection model, the generator's powerful fault detection function can be improved without manual intervention or complex detection operations, thereby enhancing the efficiency and accuracy of the generator's detection model.

[0113] In some embodiments, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores relevant data from the image acquisition device. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a generator state detection method and system based on a convolutional neural network.

[0114] In some embodiments, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, communication interface, display screen, and input system connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a generator state detection method and system based on a convolutional neural network. The display screen can be an LCD screen or an e-ink screen. The input system can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0115] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0116] In summary, this application provides a generator state detection method based on a convolutional neural network, the method comprising:

[0117] The status information of the generator's cooling system and the corresponding health status are collected to obtain the initial training samples;

[0118] The initial training samples are processed by data cleaning, feature extraction, and data classification and fusion to obtain intermediate training samples;

[0119] The intermediate training samples are classified according to three different operating conditions of the generator: no-load power, partial load power, and full load power. The intermediate training samples are then classified a second time using a preset classification code to obtain the target training samples consisting of a training set and a test set.

[0120] An initial generator state detection model is constructed to detect the state information, and the initial generator state detection model is trained based on the target training samples to obtain a trained generator state detection model.

[0121] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A generator state detection method based on convolutional neural networks, characterized in that, The method, applied to a generator condition detection model, includes: The process of collecting the status information of the generator's cooling system and the corresponding health level of the status information to obtain initial training samples includes: obtaining the fault type of the generator's cooling system based on the status information; defining a corresponding severity for each fault type and assigning a corresponding health level to each severity level; and matching each health level with an output state corresponding to the initial generator status detection model to obtain the initial training samples. The initial training samples are processed through data cleaning, feature extraction, and data classification and fusion to obtain intermediate training samples. This includes: if some sample intervals in the initial training samples have missing data, the data in those intervals is supplemented or deleted; if some sample intervals in the initial training samples have outlier data, some of the outlier data is deleted using the three Sigma criterion to obtain cleaned initial training samples; the cleaned initial training samples are sampled to preserve the data format for subsequent modeling calls; the cleaned initial training samples are reconstructed by defining the sliding window length, sliding window step size, number of data sequence segments, signal slices for cooling water inlet temperature, signal slices for cooling water outlet temperature, and original pressure signal slices of the cooling system, to obtain the intermediate training samples. The intermediate training samples are represented as follows: Where L is the sliding window length; s is the sliding window step size; and k represents the k-th data sequence segment. {T w_in，1 T w_in，2 ,…,T w_in，k } represents a signal slice for the cooling water inlet temperature; A signal slice for the cooling water outlet temperature; {P1, P2, ..., P k This is a slice of the original pressure signal from the cooling system. The intermediate training samples are classified according to three different operating conditions of the generator: no-load power, partial load power, and full load power. The cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature, and average pressure within the sliding window step size are obtained under different operating conditions. The intermediate training samples after classification are further classified by a preset classification code to obtain the target training samples consisting of the training set and the test set. Construct an initial generator state detection model for detecting the state information, and train the initial generator state detection model based on the target training samples to obtain a trained generator state detection model. This includes: inputting the cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature, and average pressure within the sliding window step size into the initial generator state detection model, outputting the health status of the generator's cooling system corresponding to the input data, and training the initial generator state detection model by combining the labels of the input data. The generator condition detection model is used to predict whether there is a fault in the generator's cooling system.

2. The generator state detection method based on convolutional neural networks according to claim 1, characterized in that, The method further includes: When the accuracy of the initial generator state detection model in predicting the health of the generator's cooling system is greater than or equal to a preset threshold, the training is complete, and the trained generator state detection model is obtained.

3. A generator state detection system based on a convolutional neural network, characterized in that, A generator applied to a target device, the system comprising: The initial training sample acquisition module is used to collect the status information of the generator's cooling system and the corresponding health level of the status information to obtain the initial training sample. It is also used to obtain the fault type of the generator's cooling system based on the status information; define a corresponding severity for each fault type; assign a corresponding health level to each severity level; and match each health level with an output state corresponding to the initial generator status detection model to obtain the initial training sample. The sample processing module is used to perform data cleaning, feature extraction, and data classification and fusion on the initial training samples to obtain intermediate training samples. It is also used to supplement or delete data in sample intervals where some data is missing, and to delete outlier data using the three Sigma criterion if outlier data exists in some sample intervals, thus obtaining cleaned initial training samples. The cleaned initial training samples are then sampled to save the data format for subsequent modeling calls. Finally, the cleaned initial training samples are reconstructed by defining the sliding window length, sliding window step size, number of data sequence segments, signal slices for cooling water inlet temperature, cooling water outlet temperature, and the original pressure signal slice of the cooling system, to obtain the intermediate training samples, which are represented as follows: Where L is the sliding window length; s is the sliding window step size; and k represents the k-th data sequence segment. {T w_in，1 T w_in，2 ,…,T w_in，k } represents a signal slice for the cooling water inlet temperature; A signal slice for the cooling water outlet temperature; {P1, P2, ..., P k This is a slice of the original pressure signal from the cooling system. The sample classification module is used to classify the intermediate training samples according to three different operating conditions of the generator: no-load power, partial load power, and full load power. It obtains the cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature, and average pressure within the sliding window step size under different operating conditions. The intermediate training samples are then classified a second time using a preset classification code to obtain the target training samples consisting of the training set and the test set. The model training module is used to construct an initial generator state detection model for detecting the state information, and to train the initial generator state detection model based on the target training samples to obtain a trained generator state detection model. It is also used to input the cooling water inlet temperature, cooling water outlet temperature, average cooling water inlet temperature, average cooling water outlet temperature and average pressure within the sliding window step size into the initial generator state detection model, and output the health of the generator's cooling system corresponding to the input data. Combined with the labels of the input data, the initial generator state detection model is trained. The status detection module predicts whether there is a fault in the generator's cooling system based on the generator status detection model.

4. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the steps of the method as described in any one of claims 1 to 2.

5. A computer device comprising a processor, a memory, and a computer program stored in the memory, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 2.

Images