Generator fault detection method, generator fault detection device, medium and equipment

Abstract

The invention discloses a generator fault detection method, a generator fault detection device, a medium and equipment, and relates to the field of power equipment. The accuracy of power equipment fault detection can be improved. The generator fault detection method comprises the following steps: acquiring fault data of different data sources of a generator; extracting a first feature of the fault data through a self-variational self-encoding network; generating an initial weight of each first feature, and optimizing the initial weight based on a simulated annealing algorithm to obtain a target weight; fusing the first features through the target weight to obtain a second feature; and inputting the second feature into a classification model to obtain the fault type of the generator.

Description

Technical Field

[0001] The present application relates to the field of electric power equipment, and in particular to a generator fault detection method, a generator fault detection device, a medium and equipment. Background Art

[0002] As generators are widely used in modern industry, fault diagnosis becomes the key to ensure their normal operation. Traditional fault diagnosis methods rely on the analysis of single types of signals such as vibration, temperature, and current, but the effectiveness and diagnostic accuracy of these signals under complex working conditions are limited. In recent years, deep learning models, especially autoencoder models, have demonstrated excellent performance in automatic feature extraction and pattern recognition. Variational autoencoders (VAEs) can extract deep features from complex data by probabilistically modeling latent spaces. However, most of the current fusion of multi-source data adopts a simple weighting strategy, and the weights are usually set manually, which makes it difficult to adapt to the differences in data sources under different working conditions. Summary of the invention

[0003] The present application provides a generator fault detection method, a generator fault detection device, a medium and a device, which can fully utilize the characteristics of different data sources and improve the accuracy of fault detection.

[0004] In a first aspect, the present application provides a generator fault detection method, comprising:

[0005] Obtain fault data from different data sources of the generator;

[0006] Extracting a first feature of the fault data by using a variational autoencoder network;

[0007] generating initial weights of the first features, and optimizing the initial weights based on a simulated annealing algorithm to obtain target weights;

[0008] The first features are fused by the target weight to obtain a second feature;

[0009] The second feature is input into a classification model to obtain a fault classification result of the generator.

[0010] According to the generator fault detection method provided by the embodiment of the present application, by acquiring fault data from different data sources, the operating status of the generator can be more comprehensively understood, and the characteristics of the generator fault can be extracted from multiple angles, which is helpful to identify complex faults. By extracting the first feature of the fault data through a self-encoding network, and then determining the weight of each first feature through a simulated annealing algorithm, the weight can be optimized, the local optimal solution can be avoided, and the optimality of the weight configuration can be improved, so as to more accurately determine the fault type.

[0011] In an exemplary embodiment, the optimizing the initial weights based on a simulated annealing algorithm to obtain target weights includes:

[0012] Get the initial temperature of the simulated annealing algorithm;

[0013] Perturbing the initial weights to obtain candidate solutions;

[0014] Calculating the acceptance probability of the candidate solution at the initial temperature, and when the acceptance probability meets a preset condition, updating the initial weight to the candidate solution; when the acceptance probability does not meet the preset condition, keeping the initial weight unchanged;

[0015] Updating the initial temperature based on a preset cooling rate;

[0016] The above steps are iterated until the initial temperature is less than the minimum value or the change in the objective function is less than a threshold value, thereby obtaining the target weight.

[0017] In an exemplary embodiment, the initial temperature includes a global temperature and a local temperature; and the method further includes:

[0018] Performing a global perturbation on the initial weights to obtain a global candidate solution, and performing a local perturbation on the initial weights to obtain a local candidate solution;

[0019] Calculating the acceptance probability of the global candidate solution by the global temperature, and calculating the acceptance probability of the local candidate solution by the local temperature;

[0020] The global temperature is updated by a global cooling rate corresponding to the global temperature, and the local temperature is updated by a local cooling rate corresponding to the local temperature.

[0021] In an exemplary embodiment, updating the global temperature by a global cooling rate corresponding to the global temperature, and updating the local temperature by a local cooling rate corresponding to the local temperature, comprises:

[0022] Acquire a global dynamic decay algorithm, and update the global temperature by using the global dynamic decay algorithm;

[0023] Acquire a local dynamic attenuation algorithm, and update the local temperature by using the local dynamic attenuation algorithm;

[0024] The global dynamic attenuation algorithm is:

[0025]

[0026] The local dynamic attenuation algorithm is:

[0027]

[0028] t is the number of iterations, is the global temperature of the tth iteration, is the local temperature of the tth iteration, α0 is the global cooling rate, β0 is the local cooling rate, 0<α, β<1, and λ is the adjustment factor.

[0029] In an exemplary embodiment, the acceptance probability of the global candidate solution is:

[0030]

[0031] The acceptance probability of the local candidate solution is:

[0032]

[0033] in, represents the change in the objective function of the global candidate solution, It represents the change of the objective function of the local candidate solution, and E(ω) represents the objective function value.

[0034] In an exemplary embodiment, it further includes:

[0035] In each round of iteration, a random number is generated, and when the acceptance probability is greater than the generated random number, it is determined that the acceptance probability meets a preset condition.

[0036] In an exemplary embodiment, the classification model is a gradient boosting decision tree.

[0037] In a second aspect, the present application provides a generator fault detection device, comprising:

[0038] A fault data acquisition module, used to acquire fault data from different data sources of the generator;

[0039] A feature extraction module, used for extracting a first feature of the fault data through a variational autoencoder network;

[0040] A weight optimization module, used for generating an initial weight of each of the first features, and optimizing the initial weight based on a simulated annealing algorithm to obtain a target weight;

[0041] A feature fusion module, used for fusing the first features through the target weight to obtain a second feature;

[0042] The classification module is used to input the second feature into a classification model to obtain the fault type of the generator.

[0043] In a third aspect, the present application provides an electronic device, the electronic device comprising a memory and one or more processors, wherein the memory stores one or more computer programs, the computer programs comprising instructions, and when the instructions are executed by the processor, the electronic device can execute the generator fault detection method in the first aspect.

[0044] In a fourth aspect, the present application provides a computer-readable storage medium having instructions stored therein. When the instructions are executed on an electronic device, the electronic device executes the generator fault detection method in the first aspect.

[0045] In a fifth aspect, the present application provides a computer program product. When the computer program product is run on an electronic device, the electronic device executes the generator fault detection method as described in the first aspect.

[0046] It can be understood that the beneficial effects that can be achieved by the generator fault detection device, electronic device, computer-readable storage medium, and computer program product provided above can refer to the beneficial effects in the first aspect and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a flow chart of a generator fault detection method provided in an embodiment of the present application;

[0048] Figure 2 A schematic diagram of the structure of a generator fault detection device provided in an embodiment of the present application;

[0049] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the words "first", "second" and the like are used to distinguish the same items or similar items with substantially the same functions and effects. For example, the first chip and the second chip are only used to distinguish different chips, and their order is not limited. It can be understood by those skilled in the art that the words "first", "second" and the like do not limit the quantity and execution order, and the words "first", "second" and the like do not necessarily limit the difference. It should be noted that in the embodiments of the present application, the words "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the present application should not be interpreted as being more preferred or more advantageous than other embodiments or design solutions. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way. In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more.

[0051] It should be noted that the “at…” in the embodiments of the present application may be the instant when a certain situation occurs, or may be a period of time after the occurrence of a certain situation, and the embodiments of the present application do not specifically limit this.

[0052] The implementation of this embodiment will be described in detail below with reference to the accompanying drawings.

[0053] This embodiment provides a generator fault detection method. By way of example, the generator fault detection method can be applied to various electronic devices such as computers (PCs), tablet computers, virtual reality / augmented reality devices, wearable devices, industrial computers, and vehicle computers; it can also be applied to servers, cloud computing, server clusters, etc., and this embodiment does not make any special limitations on this.

[0054] Figure 1 A flow chart of a generator fault detection method provided in an embodiment of the present application is shown.

[0055] like Figure 1 As shown, the generator fault detection method may include the following steps:

[0056] Step 101: Acquire fault data from different data sources of the generator.

[0057] Obtain fault data x1, x2, ..., x from multiple sensors of the generator n ,Each data source contains multiple features, and the original data is normalized so that the value of each feature is in the range of [0,1].

[0058] Step 102: extracting the first feature of the fault data through a variational autoencoder network.

[0059] The Variational Autoencoder (VAE) network consists of an encoder and a decoder. The encoder maps the input data x to the distribution q(z|x) of the latent variable z, and the decoder reconstructs the latent variable z into the reconstructed data through the distribution p(x|z) The optimization goal of VAE is to minimize the loss function L. The loss function of VAE consists of reconstruction error and KL divergence, as follows:

[0060] L=E q(z|x) [-logp(x|z)]+KL(q(z|x)||p(z)) (1)

[0061] Among them, the first term E q(z|x) [-logp(x|z)] is the reconstruction error. If the input data is continuous, the mean square error is used. Calculation; if the input data is discrete, binary cross entropy is used Calculation. The second term KL(q(z|x)||p(z)) is the Kullback-Leibler divergence, which constrains the potential distribution q(z|x) and the prior distribution p(z). By minimizing the loss function, the low-dimensional potential features z of each data source x are effectively extracted.

[0062] Step 103: Generate initial weights for each of the first features, and optimize the initial weights based on a simulated annealing algorithm to obtain target weights.

[0063] In this embodiment, the weights can be adjusted dynamically through the simulated annealing algorithm, and can adapt to different working conditions and failure modes.

[0064] Specifically, the initial weight is optimized based on a simulated annealing algorithm to obtain a target weight, including: obtaining an initial temperature of the simulated annealing algorithm; perturbing the initial weight to obtain a candidate solution; calculating the acceptance probability of the candidate solution at the initial temperature, and when the acceptance probability meets a preset condition, updating the initial weight to a candidate solution; when the acceptance probability does not meet the preset condition, keeping the initial weight unchanged; updating the initial temperature based on a preset cooling rate; iterating the above steps until the initial temperature is less than a minimum value or the change in the objective function is less than a threshold value, to obtain a target weight.

[0065] In this embodiment, the optimization of weights is an iterative process. First, the weight of each first feature is randomly determined to obtain a set of weight combinations as initial weights, and the initial temperature and cooling rate are initialized. Then in each round of iteration, the initial weights are perturbed to obtain a new set of weights as candidate solutions. The acceptance probability of the candidate solution can be calculated by the initial temperature, initial weights and candidate solutions. When the acceptance probability does not meet the preset conditions, the group of candidate solutions is not accepted, that is, the initial weights are kept unchanged. When the acceptance probability meets the preset conditions, the initial weights are updated to the candidate solutions as the initial weights for the next round of iterations. Then the initial temperature is updated, and the updated temperature is used as the initial temperature for the next round of iterations, and repeated multiple times until the initial temperature is less than the minimum value, or the change in the objective function is less than the threshold value. The minimum value and the threshold value can be set according to actual conditions, and this embodiment does not make any special restrictions on this.

[0066] The objective function change is the difference between the objective function value of the candidate solution and the objective function value of the initial weight.

[0067] In this embodiment, the initial temperature includes a global temperature and a local temperature; during the iteration process, the initial weights are globally perturbed to obtain a global candidate solution, and the initial weights are locally perturbed to obtain a local candidate solution; the acceptance probability of the global candidate solution is calculated by the global temperature, and the acceptance probability of the local candidate solution is calculated by the local temperature; the global temperature is updated by the global cooling rate corresponding to the global temperature, and the local temperature is updated by the local cooling rate corresponding to the local temperature.

[0068] That is to say, in each round of iteration, the initial weights are first perturbed globally and locally to obtain global candidate solutions and local candidate solutions. The acceptance probability of the global candidate solution and the acceptance probability of the local candidate solution are calculated respectively through the global temperature and local temperature of this round of iteration, and the initial weight of the next round of iteration is determined by the acceptance probability. If the acceptance probability of the global candidate solution meets the preset conditions, the global candidate solution is used as the initial weight of the next round of iteration. If the acceptance probability of the local candidate solution meets the preset conditions, the local candidate solution is used as the initial weight of the next round of iteration. If both the global candidate solution and the local candidate solution meet the preset conditions, both the global candidate solution and the local candidate solution are updated to the initial weight of the next iteration.

[0069] After each round of iteration, the global temperature and local temperature of the current round of iteration are updated as the global temperature and local temperature of the next round of iteration. The global temperature is updated by the global cooling rate, and the local temperature is updated by the local cooling rate. Specifically, a global dynamic attenuation algorithm is obtained, and the global temperature is updated by the global dynamic attenuation algorithm; a local dynamic attenuation algorithm is obtained, and the local temperature is updated by the local dynamic attenuation algorithm. In each round of iteration, a random number is generated, and when the acceptance probability is greater than the generated random number, it is determined that the acceptance probability meets the preset condition.

[0070] The global dynamic attenuation algorithm is:

[0071]

[0072] The local dynamic attenuation algorithm is:

[0073]

[0074] t is the number of iterations, is the global temperature of the tth iteration, is the local temperature of the tth iteration, α0 is the global cooling rate, β0 is the local cooling rate, 0<α, β<1, and λ is the adjustment factor.

[0075] The cooling rate is controlled by adjusting the factor λ, which is usually small to ensure a gradual temperature reduction. The temperature is updated dynamically to ensure a broad global search in the early stage and a fine local search in the later convergence stage.

[0076] Among them, the acceptance probability of the global candidate solution is:

[0077]

[0078] The acceptance probability of the local candidate solution is:

[0079]

[0080] in, represents the change in the objective function of the global candidate solution, It represents the change of the objective function of the local candidate solution, and E(ω) represents the objective function value.

[0081] The performance of the model is usually optimized by minimizing the objective function E(ω). Here, the objective function E(ω) is set to be the negative value of the accuracy. In this way, the accuracy can be maximized when E(ω) is minimized.

[0082] The final output target weight combination is the result of selecting the best solution between the global candidate solution and the local candidate solution. That is, if both are accepted, the objective function value of the global candidate solution is compared with the objective function value of the local candidate solution, and the solution with a better objective function value is selected as the target weight.

[0083] In general, the weight update process is as follows:

[0084] Step 1: Initialization

[0085] Initialize global temperature and local temperature As well as the cooling rate α0 and cooling rate β0 (0<α, β<1), and randomly generate the initial weight combination ω (0) ={ω1 (0) ,ω2 (0) ,...,ω n (0)}, the initial weights satisfy the weight constraints:

[0086]

[0087] Step 2: Generate candidate solutions

[0088] At the current weight ω (t) Based on this, a large random perturbation is performed to generate a global candidate solution At the current weight ω (t) Perform small random perturbations near the local region to generate local candidate solutions

[0089] Step 3: Accept the candidate solution

[0090] According to the current global temperature and the local temperature Use formula (4) and formula (5) to calculate the acceptance probabilities of the global candidate solution and the local candidate solution respectively.

[0091] In each iteration, generate a random number r ∈ [0, 1] to determine whether the acceptance probabilities of the global candidate solution and the local candidate solution meet the preset conditions. If the acceptance probability of any one of the global candidate solution or the local candidate solution is greater than the random number of this iteration, then this candidate solution can be accepted. That is, if r < P, then accept the candidate solution as the current solution; otherwise, keep the original solution.

[0092] Step 4: Dynamically update the temperature

[0093] After each iteration, update the global and local temperatures according to the corresponding dynamic decay algorithms (i.e., formula (2) and formula (3)).

[0094] Step 5: When the temperature drops to the minimum value or the change of the objective function is insufficient after several consecutive iterations, the algorithm stops. Finally, output the optimized weight combination ω * ={ω1 * , ω2 * ,..., ω n *}.

[0095] Step 104: Fuse the first feature through the target weight to obtain the second feature.

[0096] Weightedly fuse the first features of multiple data sources according to their respective target weights. The specific fusion formula is as follows:

[0097]

[0098] where n is the total number of data sources, Z fusion is the second feature obtained after weighted fusion of multiple data sources, ω i is the weight of each data source, satisfying z i represents the first feature extracted from the i-th data source.

[0099] In this implementation, in order to dynamically optimize the weighted weights of the data sources, an improved simulated annealing algorithm with a dual temperature mechanism is proposed. The balance between global search and local fine optimization is achieved through the coordinated control of global and local temperatures. The global temperature controls the exploration of the global range, allowing larger disturbances, and is used to explore more possible solutions in the early stage. The local temperature controls the fine search of the local range, and is used to make small-scale adjustments near the optimal solution in the later stage.

[0100] Step 105: Input the second feature into a classification model to obtain the fault type of the generator.

[0101] The fused feature vector obtained by optimizing the weights is input into the gradient boosting decision tree for fault classification. This application combines the deep feature extraction capability of the variational autoencoder network and the weighted fusion strategy of multi-source data, making full use of the feature information of different data sources to improve the accuracy of fault diagnosis; introducing the simulated annealing algorithm with a dual temperature mechanism to adaptively optimize the fusion weights of multi-source data, which can better balance global search and local search, improve optimization efficiency and diagnostic performance; and performing fault classification through the gradient boosting decision tree further improves the accuracy and robustness of the model.

[0102] Furthermore, this embodiment also provides a generator fault detection device, which can be used to execute the above generator fault detection method. Figure 2 As shown, the generator fault detection device 200 specifically includes: a fault data acquisition module 201, used to obtain fault data from different data sources of the generator; a feature extraction module 202, used to extract the first feature of the fault data through a variational autoencoder network; a weight optimization module 203, used to generate initial weights of each of the first features, and optimize the initial weights based on a simulated annealing algorithm to obtain a target weight; a feature fusion module 204, used to fuse the first features through the target weight to obtain a second feature; a classification module 205, used to input the second feature into a classification model to obtain the fault type of the generator.

[0103] In an exemplary embodiment, the weight optimization module 203 is specifically used to: obtain the initial temperature of the simulated annealing algorithm; perturb the initial weight to obtain a candidate solution; calculate the acceptance probability of the candidate solution at the initial temperature, and when the acceptance probability meets the preset conditions, update the initial weight to the candidate solution; when the acceptance probability does not meet the preset conditions, keep the initial weight unchanged; update the initial temperature based on a preset cooling rate; iterate the above steps until the initial temperature is less than a minimum value or the objective function change is less than a threshold value, and obtain a target weight; wherein the objective function change is the difference between the objective function value of the candidate solution and the objective function value of the initial weight.

[0104] In an exemplary embodiment, the initial temperature includes a global temperature and a local temperature; the weight optimization module 203 also includes: a dual-temperature mechanism, specifically used to globally perturb the initial weights to obtain a global candidate solution, and locally perturb the initial weights to obtain a local candidate solution; calculate the acceptance probability of the global candidate solution by the global temperature, and calculate the acceptance probability of the local candidate solution by the local temperature; update the global temperature by the global cooling rate corresponding to the global temperature, and update the local temperature by the local cooling rate corresponding to the local temperature.

[0105] In an exemplary embodiment, the dual temperature mechanism is further used to: obtain a global dynamic attenuation algorithm, and update the global temperature by the global dynamic attenuation algorithm; obtain a local dynamic attenuation algorithm, and update the local temperature by the local dynamic attenuation algorithm; the global dynamic attenuation algorithm is:

[0106]

[0107] The local dynamic attenuation algorithm is:

[0108]

[0109] t is the number of iterations, is the global temperature of the tth iteration, is the local temperature of the tth iteration, α0 is the global cooling rate, β0 is the local cooling rate, 0<α, β<1, and λ is the adjustment factor.

[0110] In an exemplary implementation, the acceptance probability of the global candidate solution is:

[0111]

[0112] The acceptance probability of the local candidate solution is:

[0113]

[0114] in, represents the change in the objective function of the global candidate solution, It represents the change of the objective function of the local candidate solution, and E(ω) represents the objective function value.

[0115] In an exemplary implementation, the weight optimization module 203 is further configured to: generate a random number in each round of iteration, and when the acceptance probability is greater than the generated random number, determine that the acceptance probability meets a preset condition.

[0116] In an exemplary embodiment, the classification model is a gradient boosting decision tree.

[0117] The specific details of each module or unit in the above generator fault detection device have been described in detail in the corresponding generator fault detection method, so they will not be repeated here.

[0118] The embodiment of the present application also provides an electronic device, Figure 3 A schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present disclosure is shown. Figure 3 The electronic device 600 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0119] like Figure 3 As shown, electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In RAM 603, various programs and data required for system operation are also stored. CPU 601, ROM 602 and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to bus 604.

[0120] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

[0121] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the embodiment of the present application are executed.

[0122] For example, when the computer program is executed by the central processing unit (CPU) 601, the following can be performed: obtaining fault data from different data sources of the generator; extracting the first feature of the fault data through a variational autoencoder network; generating initial weights for each of the first features, optimizing the initial weights based on a simulated annealing algorithm to obtain target weights; fusing the first features through the target weights to obtain second features; and inputting the second features into a classification model to obtain the fault type of the generator.

[0123] It should be noted that the computer-readable medium shown in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0124] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0125] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware, and the units described may also be arranged in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0126] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and the one or more programs include instructions, and when the instructions are executed by the electronic device, the electronic device implements the method described in the above embodiment.

[0127] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0128] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application shall be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A generator fault detection method, characterized in that: include: Obtain fault data from different data sources of the generator; Extracting a first feature of the fault data by using a variational autoencoder network; generating initial weights of the first features, and optimizing the initial weights based on a simulated annealing algorithm to obtain target weights; The first features are fused by the target weight to obtain a second feature; The second feature is input into a classification model to obtain the fault type of the generator.

2. The generator fault detection method according to claim 1, characterized in that: The step of optimizing the initial weight based on a simulated annealing algorithm to obtain a target weight includes: Get the initial temperature of the simulated annealing algorithm; Perturbing the initial weights to obtain candidate solutions; Calculating the acceptance probability of the candidate solution at the initial temperature, and when the acceptance probability meets a preset condition, updating the initial weight to the candidate solution; when the acceptance probability does not meet the preset condition, keeping the initial weight unchanged; Updating the initial temperature based on a preset cooling rate; Iterate the above steps until the initial temperature is less than the minimum value or the change of the objective function is less than the threshold value, and obtain the target weight; The objective function variation is the difference between the objective function value of the candidate solution and the objective function value of the initial weight.

3. The generator fault detection method according to claim 2, characterized in that: The initial temperature includes a global temperature and a local temperature; the method further includes: Performing a global perturbation on the initial weights to obtain a global candidate solution, and performing a local perturbation on the initial weights to obtain a local candidate solution; Calculating the acceptance probability of the global candidate solution by the global temperature, and calculating the acceptance probability of the local candidate solution by the local temperature; The global temperature is updated by a global cooling rate corresponding to the global temperature, and the local temperature is updated by a local cooling rate corresponding to the local temperature.

4. The generator fault detection method according to claim 3, characterized in that: The updating of the global temperature by a global cooling rate corresponding to the global temperature, and the updating of the local temperature by a local cooling rate corresponding to the local temperature, comprises: Acquire a global dynamic decay algorithm, and update the global temperature by using the global dynamic decay algorithm; Acquire a local dynamic attenuation algorithm, and update the local temperature by using the local dynamic attenuation algorithm; The global dynamic attenuation algorithm is: The local dynamic attenuation algorithm is: t is the number of iterations, is the global temperature of the tth iteration, is the local temperature of the tth iteration, α0 is the global cooling rate, β0 is the local cooling rate, 0<α, β<1, and λ is the adjustment factor.

5. The generator fault detection method according to claim 3, characterized in that: The acceptance probability of the global candidate solution is: The acceptance probability of the local candidate solution is: in, represents the change in the objective function of the global candidate solution, It represents the change of the objective function of the local candidate solution, and E(ω) represents the objective function value.

6. The generator fault detection method according to claim 2, characterized in that: Also includes: In each round of iteration, a random number is generated, and when the acceptance probability is greater than the generated random number, it is determined that the acceptance probability meets a preset condition.

7. The generator fault detection method according to claim 1, characterized in that: The classification model is a gradient boosting decision tree.

8. A generator fault detection device, characterized in that: include: A fault data acquisition module, used to acquire fault data from different data sources of the generator; A feature extraction module, used for extracting a first feature of the fault data through a variational autoencoder network; A weight optimization module, used for generating an initial weight of each of the first features, and optimizing the initial weight based on a simulated annealing algorithm to obtain a target weight; A feature fusion module, used for fusing the first features through the target weight to obtain a second feature; The classification module is used to input the second feature into a classification model to obtain the fault type of the generator. 9 . A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the generator fault detection method according to claim 1 .

10. An electronic device, characterized in that: It comprises a processor and a memory, wherein one or more computer programs are stored in the memory, and the one or more computer programs include instructions. When the instructions are executed by the electronic device, the electronic device executes the generator fault detection method according to any one of claims 1 to 7.

Images