A hydropower station self-supervised risk detection method and system
By using a self-supervised risk detection method and model, the problem of lagging fault monitoring in hydropower stations has been solved, enabling the prediction and detection of early fault risks.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHN ENERGY DADU RIVER REPAIR & INSTALLATION CO LTD
- Filing Date
- 2023-05-22
- Publication Date
- 2026-07-24
Smart Images
Figure CN116805034B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of hydropower station fault monitoring technology, specifically to a self-monitoring risk detection method and system for hydropower stations. Background Technology
[0002] Hydropower station fault monitoring primarily employs a method of deploying multiple sensors, each collecting data from specific mechanical components. An alarm is triggered when sensor data consistently exceeds a threshold. However, due to the high threshold setting, by the time sensor data consistently exceeds the threshold, the hydropower station has already suffered significant damage. Consequently, the fault goes unnoticed, leading to severe mechanical damage. Therefore, this threshold-based method for monitoring hydropower station fault risks suffers from a serious lag in monitoring fault risks. Summary of the Invention
[0003] To address the aforementioned shortcomings in existing technologies, this invention provides a self-monitoring risk detection method and system for hydropower stations, which solves the problem that existing methods for monitoring hydropower station fault risks by setting thresholds have a serious lag in monitoring fault risks.
[0004] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: a self-monitoring risk detection method for hydropower stations, comprising the following steps:
[0005] S1. Collect sensor data of the hydropower station under test during operation to obtain the sensor data to be tested;
[0006] S2. Construct a fault risk characteristic data sequence based on the sensor data to be detected;
[0007] S3. Based on the fault risk characteristic data sequence, a self-supervised fault detection model is used to detect electrical fault risks in water stations.
[0008] The beneficial effects of the method of this invention are as follows: This invention obtains the sensor data to be detected by collecting sensor data during the operation of the hydropower station under test, extracts data features from the sensor data to be detected, constructs a fault risk feature data sequence, and realizes the detection of hydropower station fault risk through a self-supervised fault detection model. Compared with the existing method of setting thresholds to monitor the fault risk of hydropower stations, this invention can capture the changes in data before the sensor data exceeds the threshold. These changes are reflected by data features, which amplify the changes in sensor data. Then, the fault risk is predicted by the self-supervised fault detection model, which solves the problem that the existing method of setting thresholds to monitor the fault risk of hydropower stations has a serious lag in monitoring fault risk.
[0009] Furthermore, the data types in the fault risk feature data sequence in step S2 include: distribution features of the sensor data to be detected, vibration features of the sensor data to be detected, and pulse features of the sensor data to be detected.
[0010] Furthermore, the formula for calculating the distribution characteristics of the sensor data to be detected is as follows:
[0011]
[0012] Where, γ′ i Let x be the distribution feature of the i-th sensor data to be detected. i For the i-th sensor data to be detected, Let I be the mean value of the i-th sensor data to be detected, and let I be the amount of data.
[0013] The beneficial effects of the above-mentioned further scheme are as follows: from the perspective of the distribution of the sensor data to be detected, the distribution characteristics of the sensor data to be detected are extracted, and the skewness of the distribution of the sensor data to be detected is characterized by the ratio of the third moment between the sensor data to be detected and the mean to the cube of the standard deviation.
[0014] Furthermore, the formula for calculating the vibration characteristics of the sensor data to be detected is as follows:
[0015]
[0016] Where, γ″ i Let x be the vibration feature of the i-th sensor data to be detected. i Let I be the i-th sensor data to be detected, and let I be the data volume. Let be the mean value of the i-th sensor data to be detected.
[0017] The beneficial effect of the above-mentioned further scheme is that the vibration of the sensor data to be detected can be characterized by the ratio of the fourth moment between the sensor data to be detected and the mean to the variance.
[0018] Furthermore, the formula for calculating the pulse characteristics of the sensor data to be detected is as follows:
[0019]
[0020] Where, γ″′ max For the pulse characteristics of the sensor data to be detected, x i Let I be the i-th sensor data to be detected, and let max(x) be the data volume. i Let x be I sensor data to be detected. i The maximum value in.
[0021] The beneficial effect of the above-mentioned further solution is that the size of the spike pulse data in the sensor data to be detected is measured by extracting the largest data in the sensor data to be detected.
[0022] Furthermore, the self-supervised fault detection model in step S3 is as follows:
[0023]
[0024] Where y is the prediction output of the self-supervised fault detection model, and ω′ i Let γ′ be the distribution feature of the i-th sensor data to be detected. i The weight, ω″ i For the vibration feature γ″ of the i-th sensor data to be detected i The weights, ω″′ are the pulse features γ″′ of the sensor data to be detected. max The weight, γ′ i For the distribution characteristics of the i-th sensor data to be detected, γ″ i Let γ″′ be the vibration feature of the i-th sensor data to be detected. max I represents the pulse characteristics of the sensor data to be detected, and I represents the data volume.
[0025] The beneficial effect of the above-mentioned further scheme is that it can predict the electrical fault situation of water station by using three aspects: the distribution characteristics of the sensor data to be detected, the vibration characteristics of the sensor data to be detected, and the pulse characteristics of the sensor data to be detected.
[0026] Furthermore, the weight training process in the self-supervised fault detection model is as follows:
[0027] A1. Construct a weight sequence from all weights;
[0028] A2. Assign initial values to the weights in the weight sequence;
[0029] A3. Based on the current weight sequence, calculate the predicted output of the self-supervised fault detection model;
[0030] A4. Calculate the model distance score based on the prediction output of the self-supervised fault detection model;
[0031] A5. Determine if the model distance score is less than the score threshold. If yes, training is complete. If no, proceed to step A6.
[0032] A6. Update the weights in the current weight sequence based on the model distance score, and then proceed to step A3.
[0033] The beneficial effects of the above further scheme are as follows: after assigning initial values to the weights in the weight sequence, the self-supervised fault detection model is input to obtain the prediction output. The model distance score is calculated based on the prediction output. The model distance score is used to measure whether the weight values in the current weight sequence meet the requirements. If yes, the training is complete. If no, the weights in the current weight sequence need to be updated until the weights meet the requirements.
[0034] Furthermore, the formula for calculating the model distance score in step A4 is as follows:
[0035]
[0036] in, Let λ be the model distance score, λ be the score weight, Y be the actual probability of electrical faults at the water station, and y be the predicted output of the self-supervised fault detection model.
[0037] The beneficial effects of the above-mentioned further scheme are as follows: the model distance score considers the distance between the predicted output of the self-supervised fault detection model and the actual fault probability of the water station, and corrects the distance by the ratio of the predicted output of the self-supervised fault detection model to the actual fault probability of the water station. Then, it considers the difference between the ratio of the predicted output of the self-supervised fault detection model and the actual fault probability of the water station, and comprehensively measures the accuracy of the model evaluation.
[0038] Furthermore, the formula for updating the weights in step A6 is as follows:
[0039]
[0040] Where, ω * Here, ω represents the updated weights, and ω represents the original weights. Scoring the model distance. For the maximum model distance score, is a partial derivative, and tan is a trigonometric function.
[0041] The beneficial effects of the above-mentioned further scheme are: the larger the model distance score, the faster the weight changes, and the smaller the model distance score, the slower the weight changes, so that the predicted output of the self-supervised fault detection model can closely approximate the actual fault probability of the water station.
[0042] A system for a self-monitoring risk detection method for hydropower stations includes: a data acquisition unit, a fault risk characteristic data sequence construction unit, and a hydropower station electrical fault risk detection unit;
[0043] The data acquisition unit is used to collect sensor data when the hydropower station under test is working, and obtain the sensor data to be tested;
[0044] The fault risk feature data sequence construction unit is used to construct a fault risk feature data sequence based on the sensor data to be detected;
[0045] The water station electrical fault risk detection unit is used to detect water station electrical fault risks using a self-supervised fault detection model based on fault risk characteristic data sequences.
[0046] The beneficial effects of this invention are as follows: This invention obtains the sensor data to be detected by collecting sensor data during the operation of the hydropower station under test, extracts data features from the sensor data to be detected, constructs a fault risk feature data sequence, and realizes the detection of hydropower station fault risk through a self-supervised fault detection model. Compared with the existing method of monitoring hydropower station fault risk by setting thresholds, this invention can capture data changes before the sensor data exceeds the threshold. This change is reflected by data features, which amplify the changes in sensor data. Then, the fault risk is predicted by the self-supervised fault detection model, which solves the problem that the existing method of monitoring hydropower station fault risk by setting thresholds has a serious lag in monitoring fault risk. Attached Figure Description
[0047] Figure 1 This is a flowchart of a self-monitoring risk detection method for hydropower stations.
[0048] Figure 2 This is a schematic diagram of the system structure of a self-monitoring risk detection method for hydropower stations. Detailed Implementation
[0049] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.
[0050] Example 1:
[0051] like Figure 1 As shown, a self-monitoring risk detection method for hydropower stations includes the following steps:
[0052] S1. Collect sensor data of the hydropower station under test during operation to obtain the sensor data to be tested;
[0053] S2. Construct a fault risk characteristic data sequence based on the sensor data to be detected;
[0054] S3. Based on the fault risk characteristic data sequence, a self-supervised fault detection model is used to detect electrical fault risks in water stations.
[0055] The beneficial effects of Example 1 are as follows: This invention obtains the sensor data to be detected by collecting sensor data during the operation of the hydropower station under test, extracts data features from the sensor data to be detected, constructs a fault risk feature data sequence, and realizes the detection of hydropower station fault risk through a self-supervised fault detection model. Compared with the existing method of setting thresholds to monitor the fault risk of hydropower stations, this invention can capture the changes in data before the sensor data exceeds the threshold. These changes are reflected by data features, which amplify the changes in sensor data. Then, the fault risk is predicted by the self-supervised fault detection model, which solves the problem that the existing method of setting thresholds to monitor the fault risk of hydropower stations has a serious lag in monitoring fault risk.
[0056] Example 2:
[0057] The data types in the fault risk characteristic data sequence in step S2 of the above embodiment 1 include: distribution characteristics of the sensor data to be detected, vibration characteristics of the sensor data to be detected, and pulse characteristics of the sensor data to be detected.
[0058] Example 3:
[0059] The calculation formula for the distribution characteristics of the sensor data to be detected in Embodiment 2 above is as follows:
[0060]
[0061] Where, γ′ i Let x be the distribution feature of the i-th sensor data to be detected. i For the i-th sensor data to be detected, Let I be the mean value of the i-th sensor data to be detected, and let I be the amount of data.
[0062] The beneficial effect of Example 3 is that, from the perspective of the distribution of the sensor data to be detected, the distribution characteristics of the sensor data to be detected are extracted, and the skewness of the distribution of the sensor data to be detected is characterized by the ratio of the third moment between the sensor data to be detected and the mean to the cube of the standard deviation.
[0063] Example 4:
[0064] The calculation formula for the vibration characteristics of the sensor data to be detected in Embodiment 2 above is as follows:
[0065]
[0066] Where, γ″ i Let x be the vibration feature of the i-th sensor data to be detected. i Let I be the i-th sensor data to be detected, and let I be the data volume. Let be the mean value of the i-th sensor data to be detected.
[0067] The beneficial effect of Example 4 is that the vibration of the sensor data to be detected can be characterized by the ratio of the fourth moment between the sensor data to be detected and the mean to the variance.
[0068] Example 5:
[0069] The calculation formula for the pulse characteristics of the sensor data to be detected in Embodiment 2 above is as follows:
[0070]
[0071] Where, γ″′ max For the pulse characteristics of the sensor data to be detected, x i Let I be the i-th sensor data to be detected, and let max(x) be the data volume. i Let x be I sensor data to be detected. i The maximum value in.
[0072] The beneficial effect of Example 5 is that the size of the spike pulse data in the sensor data to be detected is measured by extracting the largest data in the sensor data to be detected.
[0073] Example 6:
[0074] The self-supervised fault detection model in step S3 of embodiment 1 above is as follows:
[0075]
[0076] Where y is the prediction output of the self-supervised fault detection model, and ω′ i Let γ′ be the distribution feature of the i-th sensor data to be detected. i The weight, ω″ i For the vibration feature γ″ of the i-th sensor data to be detected i The weights, ω″′ are the pulse features γ″′ of the sensor data to be detected. max The weight, γ′ i For the distribution characteristics of the i-th sensor data to be detected, γ″ i Let γ″′ be the vibration feature of the i-th sensor data to be detected. max I represents the pulse characteristics of the sensor data to be detected, and I represents the data volume.
[0077] The beneficial effect of Example 6 is that it predicts the electrical faults of water stations by using three aspects: the distribution characteristics of the sensor data to be detected, the vibration characteristics of the sensor data to be detected, and the pulse characteristics of the sensor data to be detected.
[0078] Example 7:
[0079] The weight training process in the self-supervised fault detection model of Example 6 above is as follows:
[0080] A1. Construct a weight sequence from all weights;
[0081] A2. Assign initial values to the weights in the weight sequence;
[0082] A3. Based on the current weight sequence, calculate the predicted output of the self-supervised fault detection model;
[0083] A4. Calculate the model distance score based on the prediction output of the self-supervised fault detection model;
[0084] A5. Determine if the model distance score is less than the score threshold. If yes, training is complete. If no, proceed to step A6.
[0085] A6. Update the weights in the current weight sequence based on the model distance score, and then proceed to step A3.
[0086] The beneficial effect of Example 7 is as follows: After assigning initial values to the weights in the weight sequence, the self-supervised fault detection model is input to obtain the prediction output. The model distance score is calculated based on the prediction output. The model distance score is used to measure whether the weight values in the current weight sequence meet the requirements. If yes, the training is complete. If no, the weights in the current weight sequence need to be updated until the weights meet the requirements.
[0087] Example 8:
[0088] The formula for calculating the model distance score in step A4 of embodiment 7 above is as follows:
[0089]
[0090] in, Let λ be the model distance score, λ be the score weight, Y be the actual probability of electrical faults at the water station, and y be the predicted output of the self-supervised fault detection model.
[0091] The beneficial effects of Example 8 are as follows: The model distance score considers the distance between the predicted output of the self-supervised fault detection model and the actual fault probability of the water station. The distance is corrected by the ratio of the predicted output of the self-supervised fault detection model to the actual fault probability of the water station. The difference between the ratio of the predicted output of the self-supervised fault detection model and the actual fault probability of the water station is then considered to comprehensively measure the accuracy of the model evaluation.
[0092] Example 9:
[0093] The formula for updating the weights in step A6 of embodiment 7 above is:
[0094]
[0095] Where, ω *Here, ω represents the updated weights, and ω represents the original weights. Scoring the model distance. For the maximum model distance score, is a partial derivative, and tan is a trigonometric function.
[0096] The beneficial effect of Example 9 is that the larger the model distance score, the faster the weight changes, and the smaller the model distance score, the slower the weight changes, so that the predicted output of the self-supervised fault detection model can approximate the actual fault probability of the water station.
[0097] Example 10:
[0098] like Figure 2 As shown, the present invention provides a system for a self-monitoring risk detection method for hydropower stations, comprising: a data acquisition unit, a fault risk feature data sequence construction unit, and a hydropower station electrical fault risk detection unit;
[0099] The data acquisition unit is used to collect sensor data when the hydropower station under test is working, and obtain the sensor data to be tested;
[0100] The fault risk feature data sequence construction unit is used to construct a fault risk feature data sequence based on the sensor data to be detected;
[0101] The water station electrical fault risk detection unit is used to detect water station electrical fault risks using a self-supervised fault detection model based on fault risk characteristic data sequences.
[0102] The beneficial effects of Example 10 are as follows: This invention obtains the sensor data to be detected by collecting sensor data during the operation of the hydropower station under test, extracts data features from the sensor data to be detected, constructs a fault risk feature data sequence, and realizes the detection of hydropower station fault risk through a self-supervised fault detection model. Compared with the existing method of setting thresholds to monitor the fault risk of hydropower stations, this invention can capture the changes in data before the sensor data exceeds the threshold. These changes are reflected by data features, which amplify the changes in sensor data. Then, the fault risk is predicted by the self-supervised fault detection model, which solves the problem that the existing method of setting thresholds to monitor the fault risk of hydropower stations has a serious lag in monitoring fault risk.
[0103] Example 11:
[0104] The data types in the fault risk characteristic data sequence in the above embodiment 10 include: distribution characteristics of the sensor data to be detected, vibration characteristics of the sensor data to be detected, and pulse characteristics of the sensor data to be detected;
[0105] The formula for calculating the distribution characteristics of the sensor data to be detected is:
[0106]
[0107] Where, γ′ i Let x be the distribution feature of the i-th sensor data to be detected. i For the i-th sensor data to be detected, Let I be the mean value of the i-th sensor data to be detected, and let I be the data volume.
[0108] The formula for calculating the vibration characteristics of the sensor data to be detected is as follows:
[0109]
[0110] Where, γ″ i Let x be the vibration feature of the i-th sensor data to be detected. i Let I be the i-th sensor data to be detected, and let I be the data volume. Let be the mean value of the i-th sensor data to be detected;
[0111] The formula for calculating the pulse characteristics of the sensor data to be detected is:
[0112]
[0113] Where, γ″′ max For the pulse characteristics of the sensor data to be detected, x i Let I be the i-th sensor data to be detected, and let max(x) be the data volume. i Let x be I sensor data to be detected. i The maximum value in.
[0114] The beneficial effects of Example 11 are as follows: From the perspective of the distribution of the sensor data to be detected, the distribution characteristics of the sensor data to be detected are extracted; the skewness of the distribution of the sensor data to be detected is characterized by the ratio of the third moment between the sensor data to the mean to the cube of the standard deviation; the vibration of the sensor data to be detected is characterized by the ratio of the fourth moment between the sensor data to the mean to the variance; and the magnitude of the spike pulse data in the sensor data to be detected is measured by extracting the largest data in the sensor data to be detected.
[0115] Example 12
[0116] The self-supervised fault detection model in Example 10 above is as follows:
[0117]
[0118] Where y is the prediction output of the self-supervised fault detection model, and ω′ i Let γ′ be the distribution feature of the i-th sensor data to be detected. i The weight, ω″ i For the vibration feature γ″ of the i-th sensor data to be detectedi The weights, ω″′ are the pulse features γ″′ of the sensor data to be detected. max The weight, γ′ i For the distribution characteristics of the i-th sensor data to be detected, γ″ i Let γ″′ be the vibration feature of the i-th sensor data to be detected. max I represents the pulse characteristics of the sensor data to be detected, and I represents the data volume.
[0119] The beneficial effect of Example 12 is that it predicts the electrical faults of water stations by using three aspects: the distribution characteristics of the sensor data to be detected, the vibration characteristics of the sensor data to be detected, and the pulse characteristics of the sensor data to be detected.
[0120] Example 13
[0121] The weight training process in the self-supervised fault detection model in Example 12 above is as follows:
[0122] B1. Construct a weight sequence from all weights;
[0123] B2. Assign initial values to the weights in the weight sequence;
[0124] B3. Based on the current weight sequence, calculate the predicted output of the self-supervised fault detection model;
[0125] B4. Calculate the model distance score based on the prediction output of the self-supervised fault detection model;
[0126] B5. Determine if the model distance score is less than the score threshold. If yes, training is complete. If no, update the weights in the current weight sequence according to the model distance score and jump to step B3.
[0127] The beneficial effect of Example 13 is as follows: After assigning initial values to the weights in the weight sequence, the self-supervised fault detection model is input to obtain the prediction output. The model distance score is calculated based on the prediction output. The model distance score is used to measure whether the weight values in the current weight sequence meet the requirements. If yes, the training is complete. If no, the weights in the current weight sequence need to be updated until the weights meet the requirements.
[0128] Example 14
[0129] The formula for calculating the model distance score in Example 13 above is as follows:
[0130]
[0131] in, Let λ be the model distance score, λ be the score weight, Y be the actual probability of electrical faults at the water station, and y be the predicted output of the self-supervised fault detection model.
[0132] The beneficial effect of Example 14 is as follows: The model distance score considers the distance between the predicted output of the self-supervised fault detection model and the actual fault probability of the water station. The distance is corrected by the ratio of the predicted output of the self-supervised fault detection model to the actual fault probability of the water station. The difference between the ratio of the predicted output of the self-supervised fault detection model and the actual fault probability of the water station is then considered to comprehensively measure the accuracy of the model evaluation.
[0133] Example 15
[0134] The formula for updating the weights in Example 13 above is:
[0135]
[0136] Where, ω * Here, ω represents the updated weights, and ω represents the original weights. Scoring the model distance. Let θ be the maximum model distance score, θ be the partial derivative, and tan be a trigonometric function.
[0137] The beneficial effect of Example 15 is that the weight changes faster when the model distance score is larger, and the weight changes slower when the model distance score is smaller, so that the predicted output of the self-supervised fault detection model can approximate the actual fault probability of the water station.
[0138] Example 16
[0139] The present invention provides an electronic device, characterized in that it includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps of the self-monitoring risk detection method for hydropower stations as described in any of Embodiment 1.
[0140] In this embodiment, the electronic device may include: a processor, a memory, a bus, and a communication interface. The processor, the communication interface, and the memory are connected via the bus. The memory stores a computer program that can run on the processor. When the processor runs the computer program, it executes some or all of the steps of the self-monitoring risk detection method for hydropower stations provided in the aforementioned embodiment 1 of this application.
[0141] Example 17
[0142] The present invention provides a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed, it implements the steps of the self-monitoring risk detection method for hydropower stations as described in any one of Embodiment 1.
[0143] The aforementioned computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or combination thereof, such as static random access memory (SRAM), erasable and non-volatile read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer. The readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. The readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an application-specific integrated circuit (ASIC), or they can exist as discrete components in a point deregistration system.
[0144] The present invention can be provided as a method, apparatus, or computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) including computer-usable program code. Referring to the flowcharts and / or block diagrams of methods, apparatuses, and computer program products according to embodiments of the present invention, it should be understood that each flowchart and / or block diagram, and combinations thereof, can be implemented by computer program instructions. These computer program instructions can be provided to a computer-readable storage medium that operates in a particular manner on a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing device, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means implemented in the flowchart. Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes in the flowchart. These computer program instructions may also be accrued to a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing one or more processes and / or boxes in the flowchart. Figure 1 The steps of the function specified in one or more boxes.
[0145] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.
Claims
1. A self-monitoring risk detection method for hydropower stations, characterized in that, Includes the following steps: S1. Collect sensor data of the hydropower station under test during operation to obtain the sensor data to be tested; S2. Construct a fault risk characteristic data sequence based on the sensor data to be detected; S3. Based on the fault risk characteristic data sequence, a self-supervised fault detection model is used to detect electrical fault risks in water stations. The self-supervised fault detection model is: in, This is the predicted output of a self-supervised fault detection model. For the first Distribution characteristics of sensor data to be detected The weight, For the first Vibration characteristics of the sensor data to be detected The weight, Pulse characteristics of the sensor data to be detected The weight, For the first Distribution characteristics of sensor data to be detected For the first Vibration characteristics of the sensor data to be detected The pulse characteristics of the sensor data to be detected For data volume.
2. The self-monitoring risk detection method for hydropower stations according to claim 1, characterized in that, The data types in the fault risk characteristic data sequence in step S2 include: distribution characteristics of the sensor data to be detected, vibration characteristics of the sensor data to be detected, and pulse characteristics of the sensor data to be detected.
3. The self-monitoring risk detection method for hydropower stations according to claim 2, characterized in that, The formula for calculating the distribution characteristics of the sensor data to be detected is: in, For the first Distribution characteristics of sensor data to be detected For the first One sensor data to be detected, For the first The mean value of the data from each sensor to be detected For data volume.
4. The self-monitoring risk detection method for hydropower stations according to claim 2, characterized in that, The formula for calculating the vibration characteristics of the sensor data to be detected is as follows: in, For the first Vibration characteristics of the sensor data to be detected For the first One sensor data to be detected, For data volume, For the first The average value of the data from each sensor to be tested.
5. The self-monitoring risk detection method for hydropower stations according to claim 2, characterized in that, The formula for calculating the pulse characteristics of the sensor data to be detected is: in, The pulse characteristics of the sensor data to be detected For the first One sensor data to be detected, For data volume, for Data from the sensor to be detected The maximum value in.
6. The self-monitoring risk detection method for hydropower stations according to claim 1, characterized in that, The weight training process in the self-supervised fault detection model is as follows: A1. Construct a weight sequence from all weights; A2. Assign initial values to the weights in the weight sequence; A3. Based on the current weight sequence, calculate the predicted output of the self-supervised fault detection model; A4. Calculate the model distance score based on the prediction output of the self-supervised fault detection model; A5. Determine if the model distance score is less than the score threshold. If yes, training is complete. If no, proceed to step A6. A6. Update the weights in the current weight sequence based on the model distance score, and then proceed to step A3.
7. The self-monitoring risk detection method for hydropower stations according to claim 6, characterized in that, The formula for calculating the model distance score in step A4 is as follows: in, Scoring the model distance. As the scoring weight, This represents the actual probability of electrical failure at the water station. This is the predicted output of the self-supervised fault detection model.
8. The self-monitoring risk detection method for hydropower stations according to claim 6, characterized in that, The formula for updating the weights in step A6 is as follows: in, For the updated weights, The weights before the update. Scoring the model distance. For the maximum model distance score, For partial derivatives, These are trigonometric functions.
9. A system for a self-monitoring risk detection method for hydropower stations according to any one of claims 1 to 8, characterized in that, include: Data acquisition unit, fault risk characteristic data sequence construction unit, and water station electrical fault risk detection unit; The data acquisition unit is used to collect sensor data when the hydropower station under test is working, and obtain the sensor data to be tested; The fault risk feature data sequence construction unit is used to construct a fault risk feature data sequence based on the sensor data to be detected; The water station electrical fault risk detection unit is used to detect water station electrical fault risks using a self-supervised fault detection model based on fault risk characteristic data sequences. The self-supervised fault detection model is: in, This is the predicted output of a self-supervised fault detection model. For the first Distribution characteristics of sensor data to be detected The weight, For the first Vibration characteristics of the sensor data to be detected The weight, Pulse characteristics of the sensor data to be detected The weight, For the first Distribution characteristics of sensor data to be detected For the first Vibration characteristics of the sensor data to be detected The pulse characteristics of the sensor data to be detected For data volume.
10. The system of the self-monitoring risk detection method for hydropower stations according to claim 9, characterized in that, The data types in the fault risk characteristic data sequence include: distribution characteristics of the sensor data to be detected, vibration characteristics of the sensor data to be detected, and pulse characteristics of the sensor data to be detected; The formula for calculating the distribution characteristics of the sensor data to be detected is: in, For the first Distribution characteristics of sensor data to be detected For the first One sensor data to be detected, For the first The mean value of the data from each sensor to be detected For data volume; The formula for calculating the vibration characteristics of the sensor data to be detected is as follows: in, For the first Vibration characteristics of the sensor data to be detected For the first One sensor data to be detected, For data volume, For the first The average value of data from each sensor to be tested; The formula for calculating the pulse characteristics of the sensor data to be detected is: in, The pulse characteristics of the sensor data to be detected For the first One sensor data to be detected, For data volume, for Data from the sensor to be detected The maximum value in.
11. The system of the self-monitoring risk detection method for hydropower stations according to claim 9, characterized in that, The weight training process in the self-supervised fault detection model is as follows: B1. Construct a weight sequence from all weights; B2. Assign initial values to the weights in the weight sequence; B3. Based on the current weight sequence, calculate the predicted output of the self-supervised fault detection model; B4. Calculate the model distance score based on the prediction output of the self-supervised fault detection model; B5. Determine if the model distance score is less than the score threshold. If yes, training is complete. If no, update the weights in the current weight sequence according to the model distance score and jump to step B3.
12. The system of the self-monitoring risk detection method for hydropower stations according to claim 11, characterized in that, The formula for calculating the distance score in the calculation model is: in, Scoring the model distance. As the scoring weight, This represents the actual probability of electrical failure at the water station. This is the predicted output of the self-supervised fault detection model.
13. The system of the self-monitoring risk detection method for hydropower stations according to claim 11, characterized in that, The formula for updating the weights is: in, For the updated weights, The weights before the update. Scoring the model distance. For the maximum model distance score, For partial derivatives, These are trigonometric functions.
14. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps of the self-monitoring risk detection method for hydropower stations as described in any one of claims 1-8.
15. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed, the steps of the self-monitoring risk detection method for hydropower stations as described in any one of claims 1-8 are implemented.