Method and device for detecting health state of power system

Abstract

The invention provides a health state detection method and device for a power system, and the method comprises the steps: obtaining a monitoring signal of target equipment in the power system, and enabling the monitoring signal to be used for monitoring the health state of the target equipment; recombining and denoising the monitoring signal to obtain a recombined signal; the recombined signal is input to a trained health state evaluation model, a prediction state label output by the health state evaluation model is obtained, and the health state evaluation model is constructed based on an attention convolution module and a deformation long-short-term memory network module; based on the prediction state label, determining a health state evaluation result of the target equipment; according to the invention, the operation health state of the target equipment can be accurately predicted.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring technology, and in particular to a method and device for detecting the health status of a power system. Background Art

[0002] During the operation of the equipment, the utilization rate of the equipment will inevitably increase. Due to the changing operating environment, the aging of various components of the system will be accelerated, resulting in a shortened remaining service life.

[0003] In addition, the sub-healthy state of the equipment is also worthy of attention. If the equipment often operates in a sub-healthy state, it will cause an increase in system energy consumption. Among them, the energy loss in the sub-healthy state accounts for about 10%-40% of the total energy consumption of the cooling system. Complex and changeable working conditions will also affect the system environmental quality and equipment reliability.

[0004] The equipment monitoring method in the related art has a technical problem of being unable to detect the decline in equipment health in a timely manner. Summary of the invention

[0005] The present invention provides a method and device for detecting the health status of a power system, which are used to solve the defect in the prior art that the decline in the health level of equipment cannot be detected in time, and to accurately predict the operating health status of the target equipment.

[0006] The present invention provides a method for detecting the health status of a power system, comprising the following steps: obtaining a monitoring signal of a target device in the power system, wherein the monitoring signal is used to monitor the health status of the target device; reorganizing and denoising the monitoring signal to obtain a reorganized signal; inputting the reorganized signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; and determining a health status assessment result of the target device based on the predicted status label.

[0007] According to a method for detecting the health status of a power system provided by the present invention, before obtaining the monitoring signal of a target device in the power system, the method also includes: obtaining a training monitoring signal in a training set of the power system and a true status label corresponding to the training monitoring signal; performing segmented reorganization and denoising based on each training monitoring signal to obtain a reorganized signal sample; inputting the reorganized signal sample into a preset health status assessment model to obtain a training predicted status label output by the health status assessment model; determining a prediction error of the health status assessment model based on the training predicted status label and the true status label; performing iterative parameter update on the health status assessment model based on the prediction error according to an error back propagation algorithm; and obtaining a trained health status assessment model when the number of iterations of the iterative parameter update is greater than a preset iteration threshold.

[0008] According to a health status detection method for a power system provided by the present invention, the parameter iterative update of the health status assessment model based on the prediction error according to the error back propagation algorithm includes: iteratively executing the following steps: determining the log-likelihood function of the health status assessment model based on the prediction error; determining the gradient of each parameter in the health status assessment model based on the log-likelihood function according to the error back propagation algorithm; updating each parameter based on the gradient of each parameter according to the Nadam algorithm to obtain an updated health status assessment model; inputting the recombined signal sample into the updated health status assessment model to obtain a training prediction state label output by the updated health status assessment model; and determining the prediction error of the updated health status assessment model based on the training prediction state label and the true state label.

[0009] According to a method for detecting the health status of a power system provided by the present invention, the method also includes: obtaining a test monitoring signal in a test set of the power system and a true status label corresponding to the test monitoring signal; performing segmented reorganization and denoising on each test monitoring signal to obtain a test signal sample; inputting the test signal sample into the trained health status assessment model to obtain a test prediction status label output by the trained health status assessment model; and determining a test evaluation result of the trained health status assessment model based on the test prediction status label and the true status label.

[0010] According to a method for detecting the health status of a power system provided by the present invention, the method also includes: obtaining a target signal of the target device, wherein the target signal is used to indicate that the target device is in a target device health state; determining the cosine similarity between the monitoring signal and the target signal; and based on the cosine similarity, determining a similar abnormal result of the monitoring signal.

[0011] According to a method for detecting the health status of a power system provided by the present invention, the method further includes: determining a Pearson correlation coefficient between the monitoring signal and the target signal; and determining a related abnormal result of the monitoring signal based on the Pearson correlation coefficient.

[0012] According to a method for detecting the health status of a power system provided by the present invention, the method further includes: determining a Pearson correlation coefficient between the monitoring signal and the target signal; and determining a related abnormal result of the monitoring signal based on the Pearson correlation coefficient.

[0013] The present invention also provides a device for detecting the health status of a power system, comprising the following modules: An acquisition module is used to acquire a monitoring signal of a target device in a power system, wherein the monitoring signal is used to monitor the health status of the target device; a reorganization module is used to reorganize and denoise the monitoring signal to obtain a reorganized signal; a prediction module is used to input the reorganized signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; a determination module is used to determine a health status assessment result of the target device based on the predicted status label.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for detecting the health status of a power system as described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for detecting the health status of a power system as described in any one of the above is implemented.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned methods for detecting the health status of a power system.

[0017] The health status detection method and device of the power system provided by the present invention can monitor the operating status of the equipment in real time by acquiring the monitoring signal of the target equipment; reorganizing and denoising the monitoring signal can effectively improve the signal quality, reduce noise interference, and make subsequent analysis more accurate and reliable. The reorganized signal can better reflect the actual operating status of the equipment. The health status assessment model constructed based on the attention convolution module and the deformation long short-term memory network module is used to output the predicted status index based on the reorganized signal, which can make full use of the information in the monitoring signal and capture the dynamic characteristics and fault characteristics of the equipment; the health status assessment result of the target equipment is determined based on the predicted status label, thereby realizing efficient and accurate monitoring and assessment of the health status of the target equipment in the power system, thereby solving the technical problem that the equipment monitoring method in the related technology cannot timely detect the decline in the health level of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of the method for detecting the health status of a power system provided by the present invention.

[0020] Figure 2 It is a structural schematic diagram of the deformation long short-term memory network model provided by the present invention.

[0021] Figure 3 It is a schematic diagram of the cosine distance provided by the present invention.

[0022] Figure 4 It is a schematic diagram of the overall flow of the method for detecting the health status of a power system provided by the present invention.

[0023] Figure 5 It is a structural schematic diagram of the health status detection device of the power system provided by the present invention.

[0024] Figure 6 It is a schematic diagram of the physical structure of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] During the operation of the equipment, the utilization rate of the equipment will inevitably increase. Due to the changing operating environment, the aging of various system components will be accelerated, resulting in a shortened remaining service life. In addition, the sub-healthy state of the equipment is also worthy of attention. If the equipment is often operated in a sub-healthy state, it will cause an increase in system energy consumption. Among them, the energy loss in the sub-healthy state accounts for about 10-40% of the total energy consumption of the cooling system. Complex and changeable working conditions will also affect the system environment quality and equipment reliability.

[0027] If the decline in equipment health is not discovered in advance, once a failure occurs during the mission, not only will the equipment be unable to be repaired in time due to lack of maintenance conditions and professionals, but it will also affect the execution of the mission and cause a large waste of manpower and financial resources.

[0028] At the same time, the system operates under adverse working conditions such as pressure and corrosion for a long time. The causes of its failures are complex and the types of failures are diverse, which has a great impact on the overall health of the system and greatly reduces the system's working life and efficiency.

[0029] If not handled in time, it will endanger the safe operation of the entire device, causing the device to stop production, resulting in significant economic losses and even affecting the safety of the entire equipment.

[0030] In order to reduce maintenance costs, improve the overall reliability and safety of equipment, accurately understand the use and operation status of equipment in real time, and solve the problems of daily maintenance and fault diagnosis and repair of the system, the present invention analyzes the system and its performance parameter testing method, and adds a quantitative assessment of the system health status to achieve the function of timely grasping various equipment information and discovering the hidden dangers of equipment failure as early as possible.

[0031] Optionally, the method for detecting the health status of the power system in the embodiment of the present application can be executed by a server, or by a terminal device, or jointly by a server and a terminal device, taking the method for detecting the health status of the power system in the present embodiment executed by a server as an example.

[0032] Figure 1 is a flow chart of the method for detecting the health status of a power system provided by the present invention, such as Figure 1As shown, the method includes the following steps.

[0033] Step 101: Acquire a monitoring signal of a target device in a power system, wherein the monitoring signal is used to monitor a health status of the target device.

[0034] By continuously acquiring the monitoring signals of the target equipment, real-time monitoring of the equipment status can be achieved; this real-time performance is crucial for timely detection of potential faults and prevention of unexpected accidents; the monitoring signals are used to reflect the actual operating status of the equipment, including key parameters such as vibration, temperature, pressure, and current. The accuracy and reliability of the monitoring signals are directly related to the accuracy of subsequent health status assessments.

[0035] By acquiring multiple types of monitoring signals, you can get a comprehensive understanding of the equipment's operating status. For example, vibration signals can reflect the mechanical health of the equipment, while temperature signals can reflect the thermal performance of the equipment. This comprehensiveness helps to more fully assess the health status of the equipment.

[0036] By analyzing the monitoring signals, abnormalities can be detected before equipment failure occurs, so that preventive maintenance can be carried out in time. This can not only reduce the downtime caused by equipment failure, but also extend the service life of the equipment.

[0037] Step 102, recombining and denoising the monitoring signal to obtain a recombined signal.

[0038] In the embodiment of the present invention, each monitoring signal is reassembled and denoised in sections to obtain a reassembled signal corresponding to each monitoring signal.

[0039] According to the characteristics of the monitoring signal and the analysis requirements, select the target segmentation method (such as time window, frequency window, etc.) to divide the monitoring signal into multiple sub-segments; each sub-segment should contain sufficient information to reflect a certain operating status or characteristic of the equipment.

[0040] Feature extraction is performed on each sub-segment to obtain sub-segment signal features, where sub-segment signal features include key parameters such as amplitude, frequency, and phase. Sub-segment signal features will be used for subsequent signal reconstruction and health status assessment.

[0041] The sub-segment signal features are reorganized according to the target reorganization strategy to obtain the reorganized signal; the target reorganization strategy includes rearranging the sub-segments in chronological order, clustering and reorganizing according to feature similarity, etc. The purpose of reorganization is to make the signal clearer and easier to analyze, and to reduce noise and redundant information.

[0042] It should be noted that the reassembled sub-segments are re-synthesized into a complete signal. This step needs to ensure the continuity and consistency of the signal and avoid introducing additional noise or distortion during the reassembly process.

[0043] Denoising is the process of removing noise and interference from monitoring signals, aiming to improve the signal-to-noise ratio and clarity of the signal.

[0044] For example, a filter (such as a high-pass filter, a low-pass filter, a band-pass filter, etc.) is used to filter the signal to remove noise within a specific frequency range. The signal is decomposed into multiple frequency components using wavelet transform, and then denoising is performed based on the characteristics of noise and signal at different frequency components.

[0045] Step 103, input the recombined signal into the trained health status assessment model to obtain the predicted state label output by the health status assessment model, wherein the health status assessment model is constructed based on the attention convolution module and the deformation long short-term memory network module.

[0046] In order to accurately evaluate the health status of the system, a system health status assessment model based on Attention convolution Mogrifier LSTM (ACMLSTM) has been derived. The ACMLSTM model mainly consists of two parts: the attention convolution module and the deformation long short-term memory network module.

[0047] Attention convolution module: Convolutional neural networks consist of a series of alternating convolutional layers and pooling layers. Its convolution operation is performed on the local perception domain and has a strong local information perception ability. However, since the size of the convolution kernel is much smaller than the size of the input feature map, its global perception ability is weak. Similar to the attention mechanism of human vision, the attention mechanism helps to increase the model's attention to task-related information and suppress attention to task-irrelevant information. Therefore, for one-dimensional and two-dimensional standard convolutions, an attention one-dimensional convolution module and an attention two-dimensional convolution module are proposed respectively to enhance the global perception ability of standard convolution.

[0048] refer to Figure 2 , Figure 2 is a schematic diagram of the structure of the deformable long short-term memory network model provided by the present invention, wherein the deformable long short-term memory network model (Mogrifier LSTM, MLSTM) includes an input , multiple hidden layers , hidden layer output 、 , intermediate output , , and LSTM units.

[0049] Deformed Long Short-Term Memory Network: MLSTM is a RNN structure with relatively excellent performance. By adding a "gate" structure, the input and the hidden layer output of the previous moment can interact with each other. At the same time, the "gate" structure in LSTM is continued, avoiding problems such as gradient disappearance and gradient explosion, and further improving the ability to learn long-term dependent information.

[0050] Step 104: Determine a health status assessment result of the target device based on the predicted status tag.

[0051] In an embodiment of the present invention, the health status assessment model has been trained and is capable of processing the input recombinant signal and outputting a predicted status label. The health status assessment model combines an attention convolution module and a deformable long short-term memory network module to capture key features in the signal and process long-term dependencies in sequence data.

[0052] The recombinant signal is input into a trained health state assessment model, and the trained health state assessment model outputs one or more predicted state labels.

[0053] The predicted state labels typically correspond to possible health states of the device, for example, "normal", "minor failure", "major failure", etc. The predicted state labels are determined based on the classification or regression task learned during model training.

[0054] Based on the predicted status label output by the model, the health status assessment result of the target device can be directly determined. For example, if the predicted status label is "normal", the health status of the device is considered normal; if the label is "sub-healthy", the device needs to be repaired or replaced.

[0055] Through the above steps of the embodiment of the present invention, a monitoring signal of a target device in a power system is obtained, wherein the monitoring signal is used to monitor the health status of the target device; the monitoring signal is reorganized and denoised to obtain a reorganized signal; the reorganized signal is input into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; based on the predicted status label, a health status assessment result of the target device is determined, thereby achieving efficient and accurate monitoring and assessment of the health status of the target device in the power system, thereby solving the technical problem that the equipment monitoring method in the related technology cannot timely detect the decline in the health level of the equipment.

[0056] According to a method for detecting the health status of a power system provided by the present invention, before obtaining a monitoring signal of a target device in the power system, the method further includes: Obtaining a training monitoring signal in a training set of a power system and a true state label corresponding to the training monitoring signal; Based on each training monitoring signal, segmented reorganization and denoising are performed to obtain reorganized signal samples; Inputting the recombinant signal sample into a preset health status assessment model to obtain a training prediction status label output by the health status assessment model; Determine the prediction error of the health status assessment model based on the training prediction status label and the true status label; Iteratively update the parameters of the health status assessment model based on the prediction error according to the error back propagation algorithm; When the number of iterations of the iterative update of the parameters is greater than a preset iteration threshold, a trained health status assessment model is obtained.

[0057] In an embodiment of the present invention, training monitoring signals are collected from the power system. The training monitoring signals may be data measured by various sensors, such as vibration, temperature, pressure, etc. A real state label corresponding to each training monitoring signal is obtained. The real state label is usually obtained through expert judgment, equipment history records, or other reliable sources, and is used to represent the real health state of the equipment when the monitoring signal is collected.

[0058] Each training monitoring signal is segmented and reorganized, for example, a long signal is cut into shorter segments to better fit the input requirements of the model. De-noising is used to remove noise and interference in the signal and improve the quality of the signal. For example, this is achieved through filtering, smoothing or other signal processing techniques.

[0059] Inputting the recombinant signal sample into a preset health status assessment model to obtain a training prediction status label output by the health status assessment model; Compare the training prediction state label output by the model with the true state label and calculate the prediction error. For example, this can be achieved through various loss functions, such as cross entropy loss, mean square error (MSE), etc.

[0060] Using the error back-propagation algorithm, the model’s parameters are iteratively updated based on the prediction error. This process involves calculating the gradients, which are the derivatives of the loss function with respect to the model parameters, and using these gradients to update the parameters in order to reduce the prediction error.

[0061] Repeat the above steps until the number of iterations is greater than the preset iteration threshold. This threshold is usually preset to ensure that the model has enough time to learn and optimize its parameters. During the iteration process, the change of the loss function can be monitored to evaluate the training progress and performance of the model.

[0062] When the iteration threshold is reached, the training process ends and a trained health status assessment model is obtained; the trained health status assessment model is used to perform health status assessment on the power system monitoring signal obtained in real time.

[0063] According to a method for detecting the health status of a power system provided by the present invention, a parameter of a health status assessment model is iteratively updated based on a prediction error according to an error back propagation algorithm, comprising: Iterate through the following steps: Determine the log-likelihood function of the health status assessment model based on the prediction error; Determine the gradient of each parameter in the health status assessment model based on the log-likelihood function according to the error back-propagation algorithm; Each parameter is updated based on the gradient of each parameter according to the Nadam algorithm to obtain an updated health status assessment model; Inputting the recombinant signal sample into the updated health status assessment model to obtain the training prediction state label output by the updated health status assessment model; Based on the training predicted state labels and the true state labels, the prediction error of the updated health state assessment model is determined.

[0064] At the beginning of each iteration, the log-likelihood function is first calculated based on the prediction error of the current health status assessment model. The log-likelihood function is a mathematical representation of the degree of match between the model's predicted data and the real data. In classification problems, this usually involves cross-entropy loss; in regression problems, it is the logarithmic form of the mean squared error (MSE) or other loss functions.

[0065] Using the error back propagation algorithm, the gradient of each parameter in the health state assessment model is calculated according to the log-likelihood function. These gradients represent the derivatives of the loss function with respect to the model parameters and are used to indicate how to adjust the parameters to reduce the prediction error.

[0066] According to the Nadam algorithm, the calculated gradient is used to update each parameter in the model. Nadam is a variant of the Adam optimization algorithm that combines the characteristics of momentum and Nesterov accelerated gradient (NAG) and uses adaptive learning rate adjustment. By updating the parameters, the model gradually approaches the optimal solution.

[0067] The recombinant signal samples are input into the updated health status assessment model to obtain new training prediction status labels. These labels reflect the prediction ability of the model under the updated parameters.

[0068] Based on the new training prediction state label and the true state label, the prediction error of the updated model is calculated. This error will be used for the log-likelihood function calculation and gradient calculation of the next iteration.

[0069] Repeat the above steps until the preset number of iterations is reached or the prediction error falls below a certain threshold. During the iteration process, as the model parameters are continuously optimized, the prediction error usually decreases and the performance of the model gradually improves.

[0070] According to a method for detecting the health status of a power system provided by the present invention, the method further includes: Acquire a test monitoring signal in a test set of a power system and a real state label corresponding to the test monitoring signal; Based on each test monitoring signal, segmented reorganization and denoising are performed to obtain a test signal sample; Inputting the test signal sample into the trained health status assessment model to obtain the test prediction state label output by the trained health status assessment model; Based on the test predicted state label and the true state label, a test evaluation result of the trained health state evaluation model is determined.

[0071] In an embodiment of the present invention, test monitoring signals are collected from the power system. These signals should come from the same or similar operating conditions as the data in the training set, but should be new data that has not participated in model training. Corresponding to each test monitoring signal, a real state label is obtained. These labels are used to compare with the prediction results of the model to evaluate the performance of the model.

[0072] Reassemble each test monitoring signal in segments to ensure they have the same format and length as the signal samples used during training. Apply denoising to reduce noise and interference in the signal and improve the quality of the signal. This step should be consistent with the denoising method used during the training phase.

[0073] The preprocessed test signal samples are input into the trained health status assessment model. The model will process these input signals and output the test prediction status label.

[0074] Compare the test prediction status label output by the model with the actual status label to determine the test evaluation result of the model, for example, calculate various performance indicators of the trained health status evaluation model, such as accuracy, recall, F1 score, confusion matrix, etc.

[0075] Here, precision measures the proportion of correct predictions made by the model, recall measures the model’s ability to identify the positive class (e.g., fault status), and the F1 score is the harmonic average of precision and recall, providing a comprehensive assessment of model performance.

[0076] According to a method for detecting the health status of a power system provided by the present invention, the method further includes: Acquire a target signal of a target device, wherein the target signal is used to indicate that the target device is in a target device health state; Determine the cosine similarity between the monitoring signal and the target signal; Based on the cosine similarity, similar abnormal results of the monitoring signals are determined.

[0077] Cosine distance is also called cosine similarity. In geometry, the cosine of the angle can be used to measure the difference in the directions of two vectors. Machine learning borrows this concept to measure the difference between sample vectors. Compared with distance measurement, cosine similarity pays more attention to the difference in direction between two vectors rather than the distance or length. The formula for calculating cosine distance (1) is: (1) in, is the cosine distance, are the values ​​of two data (i.e. monitoring signal and target signal), is the number of data to be calculated.

[0078] refer to Figure 3 , Figure 3 : is a schematic diagram of the cosine distance provided by the present invention, which includes angle A, angle B, angle C, side a, side b, side c, and the angle of angle A , the angle of angle B , and the angle of angle C .

[0079] The cosine value range is [−1, 1]. The angle between the two vectors is obtained, and the cosine value corresponding to the angle is obtained. This cosine value can be used to represent the similarity of the two vectors. The smaller the angle is, the closer it is to 0 degrees, and the closer the cosine value is to 1, the more consistent their directions are, and the more similar they are; when the directions of the two vectors are completely opposite, the cosine of the angle takes the minimum value of -1; when the cosine value is 0, the two vectors are orthogonal, and the angle is 90 degrees. Therefore, it can be seen that cosine similarity has nothing to do with the magnitude of the vector, but only with the direction of the vector.

[0080] According to a method for detecting the health status of a power system provided by the present invention, the method further includes: Determining a Pearson correlation coefficient between the monitoring signal and the target signal; Based on the Pearson correlation coefficient, the relevant abnormal results of the monitoring signal are determined.

[0081] In the embodiment of the present invention, the correlation refers to the percentage of mutual connection between two things. The most common indicator reflecting the correlation is the correlation coefficient designed by statistician Karl Pearson, which is generally represented by the letter r. Due to different research objects, there are many ways to define the correlation coefficient, and the more commonly used one is the Pearson correlation coefficient. The Pearson correlation coefficient calculation formula (2) is: (2) in, represents the correlation coefficient, , Respectively represent the values ​​of two data (i.e. monitoring signal and target signal), and represents the average of two data. is the number of data to be calculated.

[0082] By analyzing the relevant data characteristics, it is found that both the cosine distance model and the correlation model have certain defects. Both cannot completely and effectively determine whether the measured signal is normal. Therefore, the present invention will respectively calculate the cosine distance between the target signal and the monitoring signal and the correlation between the normal signal and the measured signal, and combine the results of the two to determine whether the measured signal is abnormal, so as to obtain the health status of the system and the fault diagnosis prediction.

[0083] The following describes an example of a method for detecting the health status of a power system provided by the present invention in a practical application scenario.

[0084] refer to Figure 4 , Figure 4 It is a schematic diagram of the overall flow of the method for detecting the health status of a power system provided by the present invention.

[0085] The basic process of applying the ACMLSTM-based health status assessment model in system health status assessment is as follows: In the first step, the dataset consisting of the experimental monitoring signal and the true state label is divided into a training set and a test set. For each signal sample in the training set and the test set, segmentation reorganization and denoising are performed respectively.

[0086] In the second step, the recombinant signal samples in the training set are sent to the constructed health status assessment ACMLSTM model to obtain the corresponding predicted status labels.

[0087] The third step is to calculate the prediction error of the health status assessment model based on the predicted state label and the true state label. Based on the prediction error, the error back propagation algorithm is used to obtain the gradient of the log-likelihood function for each parameter, and the Nadam algorithm is used to update each parameter.

[0088] Step 4: When the number of iterations is less than the set maximum number of iterations, go to step 3; when the number of iterations is greater than the set maximum number of iterations, go to step 5.

[0089] The fifth step is to establish the health status assessment ACMLSTM model based on the optimal parameter values, and use the model to identify the status of the samples in the test set.

[0090] Through the embodiment of the present invention, by constructing the ACMLSTM health status assessment model, it is helpful to increase the model's attention to task-related information and suppress attention to task-irrelevant information. At the same time, the cosine distance model is used for calculation and analysis; finally, different upper limit values ​​are set to distinguish normal, sub-healthy, abnormal and fault signals. The analysis shows that the cosine distance model has the advantages of fast calculation speed, stable results, accurate prediction results, etc., and can be used as a basis for judging the health status of equipment.

[0091] The healthy state detection device of the power system provided by the present invention is described below. The healthy state detection device of the power system described below and the healthy state detection method of the power system described above can be referred to each other.

[0092] refer to Figure 5 , Figure 5 It is a structural schematic diagram of the health status detection device of the power system provided by the present invention.

[0093] An acquisition module 501 is used to acquire a monitoring signal of a target device in a power system, wherein the monitoring signal is used to monitor a device health status of the target device; A recombining module 502, used for recombining and denoising the monitoring signal to obtain a recombined signal; A prediction module 503, used for inputting the recombined signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; The determination module 504 is used to determine the health status assessment result of the target device based on the predicted status tag.

[0094] Specifically, the health status detection device for the above-mentioned power system provided by the present invention can implement all the method steps implemented by the health status detection method embodiment of the above-mentioned power system, and can achieve the same technical effect. The parts and beneficial effects that are the same as the method embodiment in this embodiment will not be described in detail here.

[0095] Figure 6 is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the health status detection method of the power system, the method comprising: obtaining a monitoring signal of a target device in the power system, wherein the monitoring signal is used to monitor the health status of the target device; reorganizing and denoising the monitoring signal to obtain a reorganized signal; inputting the reorganized signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; and determining the health status assessment result of the target device based on the predicted status label.

[0096] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0097] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the health status detection method of the power system provided by the above methods, and the method includes: obtaining a monitoring signal of a target device in the power system, wherein the monitoring signal is used to monitor the device health status of the target device; reorganizing and denoising the monitoring signal to obtain a reorganized signal; inputting the reorganized signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; based on the predicted status label, determining the health status assessment result of the target device.

[0098] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the health status detection method of the power system provided by the above-mentioned methods, the method comprising: obtaining a monitoring signal of a target device in the power system, wherein the monitoring signal is used to monitor the device health status of the target device; reorganizing and denoising the monitoring signal to obtain a reorganized signal; inputting the reorganized signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; based on the predicted status label, determining the health status assessment result of the target device.

[0099] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0100] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the health status of a power system, characterized in that: include: Acquire a monitoring signal of a target device in a power system, wherein the monitoring signal is used to monitor a device health status of the target device; Recombining and denoising the monitoring signal to obtain a recombined signal; Inputting the recombined signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; Based on the predicted status tag, a health status assessment result of the target device is determined.

2. The method for detecting the health status of a power system according to claim 1, characterized in that: Before acquiring the monitoring signal of the target device in the power system, the method further includes: Acquire a training monitoring signal in a training set of a power system and a real state label corresponding to the training monitoring signal; Based on each training monitoring signal, segmented reorganization and denoising are performed to obtain reorganized signal samples; Inputting the recombinant signal sample into a preset health status assessment model to obtain a training prediction status label output by the health status assessment model; Determining a prediction error of the health status assessment model based on the training predicted state label and the true state label; Iteratively updating parameters of the health status assessment model based on the prediction error according to an error back propagation algorithm; When the number of iterations of the iterative update of the parameters is greater than a preset iteration threshold, a trained health status assessment model is obtained.

3. The method for detecting the health status of a power system according to claim 2, characterized in that: The iterative updating of parameters of the health status assessment model based on the prediction error according to the error back propagation algorithm includes: Iterate through the following steps: determining a log-likelihood function of the health status assessment model based on the prediction error; Determining the gradient of each parameter in the health status assessment model based on the log-likelihood function according to an error back-propagation algorithm; updating each parameter based on the gradient of each parameter according to the Nadam algorithm to obtain an updated health status assessment model; Inputting the recombinant signal sample into the updated health status assessment model to obtain a training prediction state label output by the updated health status assessment model; Based on the training predicted state label and the true state label, a prediction error of the updated health state assessment model is determined.

4. The method for detecting the health status of a power system according to claim 2, characterized in that: The method further comprises: Acquire a test monitoring signal in a test set of a power system and a real state label corresponding to the test monitoring signal; Based on each test monitoring signal, segmented reorganization and denoising are performed to obtain a test signal sample; Inputting the test signal sample into the trained health status assessment model to obtain a test prediction state label output by the trained health status assessment model; Based on the test predicted state label and the true state label, a test evaluation result of the trained health state evaluation model is determined.

5. The method for detecting the health status of a power system according to claim 1, characterized in that: The method further comprises: Acquire a target signal of the target device, wherein the target signal is used to indicate that the target device is in a target device health state; Determining the cosine similarity between the monitoring signal and the target signal; Based on the cosine similarity, similar abnormal results of the monitoring signal are determined.

6. The method for detecting the health status of a power system according to claim 5, characterized in that: The method further comprises: Determining a Pearson correlation coefficient between the monitoring signal and the target signal; Based on the Pearson correlation coefficient, the relevant abnormal results of the monitoring signal are determined.

7. A device for detecting the health status of a power system, characterized in that: include: An acquisition module, used to acquire a monitoring signal of a target device in a power system, wherein the monitoring signal is used to monitor a device health status of the target device; A recombining module, used for recombining and denoising the monitoring signal to obtain a recombined signal; A prediction module, used for inputting the recombined signal into a trained health status assessment model to obtain a predicted status label output by the health status assessment model, wherein the health status assessment model is constructed based on an attention convolution module and a deformation long short-term memory network module; A determination module is used to determine a health status assessment result of the target device based on the predicted status tag.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for detecting the health status of a power system as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for detecting the health status of a power system according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for detecting the health status of a power system according to any one of claims 1 to 6 is implemented.

Images