Water conservancy electromechanical equipment fault prediction method and device, storage medium and electronic equipment

Abstract

The invention relates to a water conservancy electromechanical equipment fault prediction method and device, a storage medium and electronic equipment, and relates to the technical field of water conservancy projects, and the method comprises the steps: obtaining multi-source heterogeneous monitoring data corresponding to water conservancy electromechanical equipment; performing fault fingerprint matching on equipment state data in the multi-source heterogeneous monitoring data to obtain fault features corresponding to the equipment state data; fusing the fault features and working condition features corresponding to equipment working condition data in the multi-source heterogeneous monitoring data, and obtaining equipment monitoring features corresponding to the water conservancy electromechanical equipment; based on the fault fingerprint matching degree of the equipment monitoring characteristics, predicting an equipment health index and a fault evolution rate corresponding to the water conservancy electromechanical equipment; and determining the fault prediction time and the maintenance window period of the water conservancy electromechanical equipment according to the equipment health index, the fault evolution rate and the scheduling plan data corresponding to the water conservancy electromechanical equipment. Potential faults of the equipment are predicted in advance, and the condition of fault passive response is reduced.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of water conservancy engineering, in particular to a water conservancy electromechanical equipment fault prediction method and device, a storage medium and an electronic device. BACKGROUND

[0002] Water conservancy electromechanical equipment is the core of ensuring the safe and stable operation of projects such as hydropower stations and pump stations. State monitoring and fault troubleshooting of these devices are the key to realizing predictive maintenance, reducing unplanned downtime and improving operational efficiency.

[0003] In related technologies, the water conservancy electromechanical equipment is usually monitored and fault-troubled by using an artificial patrol mode. The operation and maintenance personnel regularly go to the site to judge the device state by visual observation and handheld instrument detection, and record and report the fault after finding obvious abnormalities, However, the patrol cycle of the artificial patrol mode in related technologies is long. If a device suddenly fails between two patrols, it cannot be discovered in time, which may lead to abnormal situations such as unit shutdown and reservoir scheduling interruption. The monitoring is highly lagging, resulting in passive response to faults. SUMMARY

[0004] Therefore, the application provides a water conservancy electromechanical equipment fault prediction method and device, a storage medium and an electronic device, which mainly aims to solve the technical problem that the patrol cycle of the artificial patrol mode in related technologies is long, and if a device suddenly fails between two patrols, it cannot be discovered in time, which may lead to abnormal situations such as unit shutdown and reservoir scheduling interruption. The monitoring is highly lagging, resulting in passive response to faults.

[0005] According to a first aspect of the application, a water conservancy electromechanical equipment fault prediction method is provided, which comprises: Obtaining multi-source heterogeneous monitoring data corresponding to the water conservancy electromechanical equipment; Performing fault fingerprint matching on the device state data in the multi-source heterogeneous monitoring data to obtain fault features corresponding to the device state data; Fusing the fault features and working condition features corresponding to the device working condition data in the multi-source heterogeneous monitoring data to obtain device monitoring features corresponding to the water conservancy electromechanical equipment; Based on the fault fingerprint matching degree of the device monitoring features, predicting the device health index and fault evolution rate corresponding to the water conservancy electromechanical equipment; According to the device health index, the fault evolution rate and the scheduling plan data corresponding to the water conservancy electromechanical equipment, determining the fault prediction time and the maintenance window period of the water conservancy electromechanical equipment.

[0006] According to a second aspect of the application, a water conservancy electromechanical equipment fault prediction device is provided, which comprises: The acquisition module is used to acquire multi-source heterogeneous monitoring data corresponding to water conservancy electromechanical equipment; perform fault fingerprint matching on the equipment status data in the multi-source heterogeneous monitoring data, and acquire the fault characteristics corresponding to the equipment status data. The fusion module is used to fuse fault characteristics and equipment operating condition data from multi-source heterogeneous monitoring data to obtain the equipment monitoring characteristics corresponding to the hydraulic electromechanical equipment. The prediction module is used to predict the equipment health index and fault evolution rate of water conservancy electromechanical equipment based on the fault fingerprint matching degree of equipment monitoring characteristics. The determination module is used to determine the fault prediction time and maintenance window period of water conservancy electromechanical equipment based on the equipment health index, fault evolution rate and the scheduling plan data corresponding to the water conservancy electromechanical equipment.

[0007] According to a third aspect of this application, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of the first aspect described above.

[0008] According to a fourth aspect of this application, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause a computer to perform the method of the first aspect described above.

[0009] Compared with related technologies, the method, device, storage medium, and electronic equipment for predicting faults in hydraulic electromechanical equipment provided in this application firstly acquire multi-source heterogeneous monitoring data corresponding to the hydraulic electromechanical equipment; performs fault fingerprint matching on the equipment status data in the multi-source heterogeneous monitoring data to obtain fault characteristics corresponding to the equipment status data; fuses the fault characteristics and the operating condition characteristics corresponding to the equipment operating condition data in the multi-source heterogeneous monitoring data to obtain equipment monitoring characteristics corresponding to the hydraulic electromechanical equipment; predicts the equipment health index and fault evolution rate corresponding to the hydraulic electromechanical equipment based on the fault fingerprint matching degree of the equipment monitoring characteristics; and determines the fault prediction time and maintenance window period of the hydraulic electromechanical equipment based on the equipment health index, fault evolution rate, and scheduling plan data corresponding to the hydraulic electromechanical equipment. In this way, this application can perform fault fingerprint matching on the equipment status data in the multi-source heterogeneous monitoring data corresponding to the water conservancy electromechanical equipment to obtain the fault characteristics corresponding to the equipment status data, so as to identify the fault type of the equipment. Then, the fault characteristics are fused with the operating condition characteristics corresponding to the equipment operating condition data to obtain the equipment monitoring characteristics, eliminating the differences in multi-source heterogeneous monitoring data. Based on the fault fingerprint matching degree of the equipment monitoring characteristics, the equipment health index and fault evolution rate of the water conservancy electromechanical equipment are initially predicted. Finally, combined with the scheduling plan data corresponding to the water conservancy electromechanical equipment, the fault prediction time and maintenance window period of the water conservancy electromechanical equipment are determined, thereby predicting potential equipment faults in advance, facilitating planned maintenance intervention by maintenance personnel, reducing the situation of passive fault response, and thus reducing the power generation loss and ineffective operation and maintenance costs caused by passive fault response. Attached Figure Description

[0010] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating a method for predicting faults in hydraulic electromechanical equipment provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a fault prediction device for hydraulic electromechanical equipment provided in an embodiment of this application. Detailed Implementation

[0013] The exemplary embodiments of the present application will be described below in conjunction with the accompanying drawings. Various details of the embodiments of the present application are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, descriptions of well-known functions and structures are omitted in the following description for clarity and conciseness. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0014] The following describes a method, device, storage medium, and electronic device for predicting faults of water conservancy electromechanical equipment according to embodiments of the present application with reference to the accompanying drawings.

[0015] The present application provides a method, device, storage medium, and electronic device for predicting faults of water conservancy electromechanical equipment, mainly aiming to solve the technical problems in the related art that the inspection cycle of the manual inspection mode is relatively long. If a device suddenly fails between two inspections, it cannot be discovered in time, easily leading to abnormal situations such as unit shutdown and reservoir dispatching interruption, and the monitoring hysteresis is relatively strong, resulting in passive response to faults.

[0016] As Figure 1 shown, an embodiment of the present application provides a method for predicting faults of water conservancy electromechanical equipment, including: Step 101: Obtain multi-source heterogeneous monitoring data corresponding to water conservancy electromechanical equipment.

[0017] In some embodiments, the water conservancy electromechanical equipment can be core power and control electromechanical devices serving functions such as reservoir flood control, power generation, irrigation, and ecological dispatching. It can include core equipment for power generation, such as water turbine generator units, governors, excitation systems, etc. The water turbine generator unit can include main shaft bearings, stator cores, rotor windings, etc.

[0018] Correspondingly, the multi-source heterogeneous monitoring data can refer to monitoring data obtained through different data sources and having different dimensional attributes. The data sources can include: sensing and monitoring devices installed at key fault positions of water conservancy electromechanical equipment, such as vibration sensors installed on the bearing seats of main shaft bearings, and sensing and monitoring devices of the water conservancy electromechanical equipment itself, such as speed sensors; the data sources can also include hydrological monitoring stations, meteorological monitoring stations, etc. supporting the water conservancy electromechanical equipment to monitor the influence of the external environment on the water conservancy electromechanical equipment. The data collected from these different data sources have differential attributes such as time series characteristics, numerical characteristics, and classification characteristics, and are heterogeneous in terms of data dimension, acquisition frequency, and structural form. Based on these data, the health status of water conservancy electromechanical equipment can be comprehensively evaluated from multiple dimensions, improving the accuracy of fault prediction of water conservancy electromechanical equipment.

[0019] Step 102: Perform fault fingerprint matching on the equipment status data in the multi-source heterogeneous monitoring data to obtain the fault characteristics corresponding to the equipment status data.

[0020] The equipment status data can be data collected by sensors and other devices installed at critical fault locations to detect the status of hydraulic electromechanical equipment. This data reflects the operating status of different parts and components of the equipment. For example, vibration sensors can be installed on the bearing housings and spindle ends of hydraulic electromechanical equipment to collect data such as effective values ​​of vibration acceleration, vibration velocity, vibration displacement, and vibration frequency domain characteristics. This data can be used to determine faults such as bearing wear, rotor imbalance, coupling misalignment, and abnormal gear meshing. Fault characteristics can be abnormal equipment data obtained through fault fingerprint matching, such as abnormal frequencies (e.g., bearing inner ring fault frequency, gear meshing abnormal frequency) and abnormal amplitudes (e.g., effective values ​​of vibration acceleration exceeding standard thresholds).

[0021] In some embodiments, based on a fault fingerprint database corresponding to different models and operating years of hydraulic electromechanical equipment, the real-time collected equipment status data can be compared with the fault fingerprints corresponding to various faults in the database to obtain the most similar fault fingerprint. The equipment status data is then fused based on the most similar fault fingerprint, such as by weighted fusion based on the matching degree between various status data in the equipment status data and various fault data in the fault fingerprint, to obtain the fault features corresponding to the equipment status data. Here, a fault fingerprint can refer to a unique and repeatable fault feature template presented by the equipment status data of hydraulic electromechanical equipment when a certain type of fault occurs. For example, the fault feature set corresponding to spindle bearing wear may include: a 150Hz vibration peak value and a continuous temperature increase of 5℃ / h.

[0022] Step 103: Integrate the fault characteristics and the equipment operating condition data from the multi-source heterogeneous monitoring data to obtain the equipment monitoring characteristics corresponding to the water conservancy electromechanical equipment.

[0023] Among them, equipment operating condition data can be the working attributes of the hydraulic electromechanical equipment itself, used to reflect the working status of the hydraulic electromechanical equipment, such as speed, output power, lubricating oil temperature, cooling water pressure, guide vane opening, etc. Operating condition characteristics can be the characteristics corresponding to the operating condition data of related equipment operating status selected from the equipment operating condition data, such as guide vane opening deviation, output power decrease, etc.

[0024] In some embodiments, the weights corresponding to fault features and operating condition features can be adaptively adjusted according to the operating characteristics corresponding to the equipment operating condition data. Weighted fusion is then performed based on the adjusted weights to obtain equipment monitoring features. Deep learning can also be used for fusion processing to generate equipment monitoring features that reflect the equipment's operating status. Feature-level fusion enhances the completeness and reliability of fault information, avoiding the bias of features from a single source. Correspondingly, the equipment monitoring features can be a set of features obtained through fault fingerprint matching and multi-source feature fusion processing, reflecting the direct correlation between the operating status and faults of hydraulic electromechanical equipment from multiple dimensions. This eliminates the differences between multi-source heterogeneous monitoring data, integrates multi-source heterogeneous monitoring data into equipment monitoring features, provides reliable input for subsequent equipment health assessment, and improves the accuracy of fault prediction for hydraulic electromechanical equipment.

[0025] Step 104: Based on the fault fingerprint matching degree of equipment monitoring characteristics, predict the equipment health index and fault evolution rate of the corresponding water conservancy electromechanical equipment.

[0026] In some embodiments, a fault prediction model can be constructed based on a dynamic weighted LSTM layer and an attention layer. This model is used to input equipment monitoring features, which are compared with a fault fingerprint database to determine the fault fingerprint matching degree of each equipment monitoring feature. Then, equipment fault prediction is performed, initially predicting the equipment health index and fault evolution rate of the hydraulic electromechanical equipment. The equipment health index can be a quantitative indicator for assessing the health status of hydraulic electromechanical equipment, and can be divided into multiple index intervals to assess different health states. Correspondingly, the fault evolution rate can be the rate of decrease of the equipment health index over time, with units such as equipment health index / day or equipment health index / hour. It can be used to reflect the changing trend of equipment health status. A higher fault evolution rate indicates a higher fault development rate and faster equipment degradation of the hydraulic electromechanical equipment.

[0027] Step 105: Based on the equipment health index, fault evolution rate, and scheduling plan data corresponding to the water conservancy electromechanical equipment, determine the fault prediction time and maintenance window period for the water conservancy electromechanical equipment.

[0028] Among them, the scheduling plan data can be the reservoir scheduling plan data synchronously collected by the reservoir scheduling system based on water conservancy electromechanical equipment, such as the flood season water level control range, power generation load curve, ecological flow guarantee requirements, etc. The reservoir scheduling system can be used to monitor and analyze reservoir water level, flow, rainfall and other data in real time, and control water conservancy electromechanical equipment such as gate hoists, pump units, and turbine generator units according to different scheduling tasks.

[0029] In some embodiments, the time node reaching the critical fault threshold can be determined based on the predicted equipment health index and fault evolution rate, serving as the initial predicted fault time for the hydraulic electromechanical equipment. This initial predicted fault time is then dynamically corrected based on the current scheduling plan data corresponding to the hydraulic electromechanical equipment, resulting in the corrected fault prediction time and the optimal maintenance window. This allows for early prediction of equipment failures, helping to reduce power generation losses caused by downtime after a fault occurs. Correspondingly, the maintenance window can be a time period selected by comprehensively considering the fault prediction time, fault evolution rate, and scheduling plan data, which meets the equipment maintenance time requirements and does not conflict with the core scheduling tasks of the reservoir. This facilitates maintenance personnel in developing maintenance plans in advance, reducing power generation losses caused by sudden faults, and also reducing the cost of aimless inspections.

[0030] For example, the time point at which the equipment will experience a substantial failure can be estimated by using a time-series prediction model based on the current equipment health index and failure evolution rate, and this can be used as a preliminary prediction of the failure time. Alternatively, a health decay curve can be plotted based on the equipment health index and failure evolution rate over a period of time, and the preliminary prediction of the failure time of the equipment can be estimated based on the curve.

[0031] Compared with related technologies, this embodiment can perform fault fingerprint matching on the equipment status data in the multi-source heterogeneous monitoring data corresponding to the hydraulic electromechanical equipment to obtain the fault characteristics corresponding to the equipment status data, so as to identify the fault type of the equipment. Then, the fault characteristics are fused with the operating characteristics corresponding to the equipment operating data to obtain the equipment monitoring characteristics, eliminating the differences in multi-source heterogeneous monitoring data. Based on the fault fingerprint matching degree of the equipment monitoring characteristics, the equipment health index and fault evolution rate of the hydraulic electromechanical equipment are initially predicted. Finally, combined with the scheduling plan data corresponding to the hydraulic electromechanical equipment, the fault prediction time and maintenance window period of the hydraulic electromechanical equipment are determined, thereby predicting potential equipment faults in advance, facilitating planned maintenance intervention by maintenance personnel, reducing the situation of passive fault response, and thus reducing the power generation loss and ineffective operation and maintenance costs caused by passive fault response.

[0032] Furthermore, as a refinement and extension of the specific implementation methods of the above embodiments, and to fully illustrate the implementation methods of this embodiment, the multi-source heterogeneous monitoring data may optionally include, but is not limited to, hydrological environmental data and meteorological forecast data. Hydrological environmental data can be used to reflect the dynamics of reservoir water volume and scheduling conditions, as well as the physicochemical properties of the water body, and is related to the operating load of the generating units; excessive sediment content will exacerbate turbine impeller wear. Meteorological forecast data may include precipitation forecast data, temperature forecast data, wind condition forecast data, etc., for the area where the hydraulic equipment is located, so as to combine meteorological conditions to assess and adjust the equipment health index and fault evolution rate, improve the accuracy of fault prediction time, and reduce ineffective maintenance costs.

[0033] Optionally, step 101 may specifically include: identifying key fault locations prone to failure through structural mechanical analysis of the hydraulic equipment, deploying sensors at the key fault locations to collect equipment status data in real time; obtaining hydrological environmental data corresponding to the hydraulic equipment through the reservoir monitoring station; and obtaining meteorological forecast data for the region where the hydraulic equipment is located through the meteorological forecast data interface.

[0034] In some embodiments, triaxial vibration sensors and high-precision temperature sensors (sampling frequency can be 1kHz) can be deployed at key fault locations of the hydraulic equipment, such as the turbine main shaft bearing housing, generator stator core, and guide vane housing. Acoustic and ultrasonic sensors can also be deployed within a 1m radius around the equipment to collect vibration signals, noise signals, and ultrasonic detection signals of surface defects as equipment status data. Simultaneously, equipment operating data, such as rotational speed, output power, lubricating oil temperature, cooling water pressure, and guide vane opening, can be collected. Real-time water level, inflow / outflow, water quality pH, and sediment content can be collected through a reservoir monitoring station. Weather forecast data for the next 15 days, such as rainfall, temperature, and wind speed, can be obtained by connecting to a meteorological forecast data interface with the meteorological department. The corresponding scheduling plan data for the hydraulic equipment can be synchronized from the reservoir scheduling system.

[0035] For example, one vibration sensor and one temperature sensor can be installed at locations such as the turbine main shaft bearing housing, generator stator core, water guide mechanism housing, and lubricating oil tank; three acoustic sensors can be evenly deployed within a 1m radius around the equipment; equipment operating condition data can be collected via a PLC system at a frequency of 10 seconds per acquisition; hydrological environment data can be collected via a reservoir monitoring station at a frequency of 10 seconds per acquisition; vibration / noise data from the equipment status data can be directly collected via sensors at a frequency of 1 second per acquisition; meteorological forecast data can be obtained by connecting to a meteorological platform API, collecting daily rainfall and temperature data for the next 7-15 days, and updating once a day; all data can be stored in an edge computing gateway and a cloud database, named according to "equipment number-collection time-data type", and the data format can be uniformly JSON for easy subsequent processing.

[0036] Optionally, data preprocessing can be performed on multi-source heterogeneous monitoring data and scheduling plan data, such as using a denoising algorithm based on empirical mode decomposition (EMD) to remove environmental noise such as water flow noise and electromagnetic interference from multi-source heterogeneous monitoring data; and using an interpolation method based on spatiotemporal correlation to supplement missing data in multi-source heterogeneous monitoring data.

[0037] Optionally, step 102 may specifically include: based on the fault fingerprint database corresponding to the water conservancy electromechanical equipment, performing fault fingerprint matching on the equipment status data to obtain the fault characteristics corresponding to the water conservancy electromechanical equipment. The fault fingerprint database includes a set of fault characteristics of different models of water conservancy electromechanical equipment during the aging process; fusing fault characteristics, operating condition characteristics, and the equipment aging coefficient corresponding to the water conservancy electromechanical equipment, as well as the hydrological environment characteristics corresponding to the hydrological environment data in the multi-source heterogeneous monitoring data, to obtain the equipment monitoring characteristics corresponding to the water conservancy electromechanical equipment. The equipment aging coefficient is determined based on the operating time and maintenance frequency of the water conservancy electromechanical equipment.

[0038] In some embodiments, a fault fingerprint database can be constructed based on historical fault data of hydraulic electromechanical equipment. This database may contain feature templates for over 50 typical faults, such as bearing wear, stator winding aging, and water guide mechanism jamming. Equipment status data can be compared with various feature templates in the fault fingerprint database. The fault characteristic is determined based on the most similar feature template. Then, combined with operating condition characteristics and hydrological environment characteristics, equipment detection characteristics are fused to achieve accurate early identification of subtle faults. Compared to methods that only collect partial equipment data, this embodiment reduces the bias of analysis due to a single data dimension, reduces errors in fault attribution, and lowers ineffective maintenance costs.

[0039] Among them, hydrological environmental characteristics can be environmental characteristics extracted from hydrological environmental data that are associated with fault characteristics. They can be used to reflect the impact of the external environment on hydraulic electromechanical equipment, such as reservoir water level, water flow sediment content, head difference, water temperature, etc. For example, when the water flow sediment content exceeds the preset sediment content threshold, it is easy to cause a sudden increase in impeller vibration amplitude and accelerated wear rate of flow components.

[0040] Correspondingly, the equipment aging coefficient for hydraulic electromechanical equipment can be used to reflect the degree of performance degradation caused by factors such as usage time and component wear. For example, it can be calculated according to the following formula: Equipment aging factor = (Actual operating time of equipment / Design life) × a + Number of maintenance operations × b; Where 'a' can represent the usage time weight (e.g., 0.7) and 'b' can represent the maintenance weight (e.g., 0.3). 'a' and 'b' can be adjusted according to different equipment to make the calculated equipment aging coefficient more closely match the actual aging state of the equipment.

[0041] Optionally, based on the fault fingerprint database corresponding to the hydraulic electromechanical equipment, fault fingerprint matching is performed on the equipment status data to obtain the fault features corresponding to the hydraulic electromechanical equipment. Specifically, this may include: performing wavelet packet decomposition on the vibration signal in the equipment status data to obtain the entropy features of the vibration signal in different frequency bands; performing joint time-domain and frequency-domain feature extraction on the vibration signal to obtain the time-domain and frequency-domain features corresponding to the vibration signal; generating vibration features corresponding to the vibration signal based on the entropy features, time-domain features, and frequency-domain features; and performing fault fingerprint matching between the vibration features and the fault fingerprint database to obtain the fault features corresponding to the hydraulic electromechanical equipment.

[0042] In some embodiments, the denoised vibration signal can first be decomposed into wavelet packets, with up to 6 decomposition levels and db6 as the wavelet basis function. Then, the vibration signal is decomposed into different frequency bands, and entropy features such as energy entropy, singular value entropy, and approximate entropy of each frequency band are extracted to capture the nonlinear characteristics of the vibration signal. Then, joint time-domain and frequency-domain feature extraction is performed. The decomposed signals are analyzed in the time domain, such as extracting multiple time-domain features such as peak value, root mean square value, and peak factor. The decomposed signals are also analyzed in the frequency domain, such as extracting multiple frequency-domain features such as main frequency, harmonic frequency, and sideband energy through Fast Fourier Transform (FFT). Based on the above features, a multidimensional basic feature vector is formed as the vibration feature corresponding to the vibration signal.

[0043] For example, the original vibration signal is processed using wavelet threshold denoising. The db4 wavelet basis can be selected, and the number of decomposition layers is 5. Water flow noise and environmental noise are removed with the default threshold to obtain a clean vibration signal. After removing environmental interference, the time domain signal is converted into a frequency domain signal by Fourier transform, and the time domain features and frequency domain features are extracted. The vibration features are then standardized according to the calculation method of "(eigenvalue - mean) / standard deviation" to eliminate the influence of dimensions and obtain a standardized vibration feature vector. Then, fault fingerprint matching is performed.

[0044] In this way, targeting the fault signal characteristics of hydraulic electromechanical equipment, a three-level feature extraction scheme of "wavelet packet decomposition - time-domain and frequency-domain joint features - fault fingerprint database matching" is adopted to separate the inherent fault characteristics of the equipment and environmental interference signals from the original signal, and construct a fingerprint database containing more than 50 types of typical faults to facilitate multi-scale feature fusion and fault fingerprint extraction, achieve accurate identification of early weak faults, and the early fault identification lead time is ≥15 days. The three-level feature extraction scheme can capture weak fault signals in the early stage of equipment health degradation, leaving sufficient time for maintenance.

[0045] Furthermore, fault fingerprint matching can be performed. The vibration feature vectors corresponding to the vibration characteristics are input into the constructed fault fingerprint database. Feature matching is achieved through distance metric algorithms (such as Euclidean distance and cosine similarity). The fault types and corresponding feature weights that are most similar to the vibration characteristics of the current equipment are then selected to form fault features. Distance metric algorithms can include Euclidean distance, cosine similarity, etc.

[0046] Optionally, step 104 may specifically include: obtaining a fault prediction model corresponding to the water conservancy electromechanical equipment, the fault prediction model being trained based on the full life cycle data of water conservancy electromechanical equipment of different models and different years of operation; using the fault prediction model, based on the matching degree between the equipment monitoring characteristics and the fault fingerprint database, predicting the equipment health index and fault evolution rate corresponding to the water conservancy electromechanical equipment.

[0047] For example, the fault prediction model structure may include: an input layer, a feature enhancement layer, a dynamic weight LSTM layer, an attention layer, a fully connected layer, and an output layer, realizing the quantification of equipment health status and modeling of the degradation process. Data from the entire lifecycle of 60 different models and operating years of hydraulic electromechanical equipment, totaling 150,000 samples, can be collected and divided into training, validation, and test sets in a 7:2:1 ratio. Mean squared error (MSE) is used as the loss function, and the AdamW optimizer is used for training. The learner rate of this optimizer can be 0.0008, and the weight decay can be 0.001. Training stops when the validation set loss changes less than 1e-6 for 30 consecutive iterations. The fault prediction model is obtained through this training method, and the health index prediction error of the trained fault prediction model on the test set is ≤0.015. Correspondingly, the fault prediction model can be used to correlate with a fault fingerprint database, matching equipment monitoring features with the fault fingerprint database to predict the equipment health index and fault evolution rate based on the fault feature set of highly matched fault fingerprints.

[0048] Optionally, a fault prediction model can be used to predict the equipment health index and fault evolution rate of hydraulic electromechanical equipment based on the matching degree between equipment monitoring features and fault fingerprints in the fault fingerprint database. Specifically, this may include: using the attention layer in the fault prediction model, adaptively adjusting the feature weights corresponding to the equipment monitoring features according to the matching degree between the equipment monitoring features and fault fingerprints of different fault types in the fault fingerprint database, and determining the target fault type corresponding to the equipment monitoring features; and predicting the equipment health index and fault evolution rate based on the adjusted feature weights and the target fault type.

[0049] Optionally, the method in this embodiment may further include: using the dynamic weight LSTM layer in the fault prediction model to adjust the vibration feature weights of the vibration features corresponding to the operating condition features in real time.

[0050] For example, the workflow corresponding to the model structure of the fault prediction model is as follows: Input layer: Input the feature vector corresponding to the 28-dimensional equipment monitoring features. The equipment monitoring features may include 18 core fault features, 6 operating condition features corresponding to equipment operating data, 3 hydrological environment features corresponding to hydrological environment data, and a feature set consisting of 1 equipment aging coefficient. The equipment aging coefficient is calculated as: (Actual operating time / Design life) × 0.7 + (Number of maintenance visits) × 0.3. Feature enhancement layer: The input device monitoring features are enhanced in dimensionality and redundancy are removed by using a residual network (ResNet) to strengthen the effective feature representation; Dynamic Weight LSTM Layer: Based on the current operating condition of the equipment corresponding to the operating condition characteristics, the weights of the neurons in the LSTM layer are adjusted in real time through the reinforcement learning algorithm (DQN). Higher weights are assigned to vibration features under high load conditions as vibration feature weights, while temperature and noise features under low load conditions are given special attention. Attention layer: Dynamically allocate attention weights to each feature dimension in the device monitoring features, and assign weight coefficients such as 1.2-1.5 to features with high fault fingerprint matching degree to strengthen the influence of key fault features; Output layer: Outputs device health index (HI, range 0-1) and fault evolution rate (v, unit: HI / day); specifically, the device health index and fault evolution rate can be predicted based on the adjusted feature weights of each layer and the identified target fault types.

[0051] In this way, this embodiment can introduce equipment aging coefficients and dynamic weights based on LSTM and attention mechanisms, and adjust model parameters in real time through reinforcement learning. The equipment aging coefficients can be dynamically updated based on equipment operating time and maintenance frequency, and feature weights can be adaptively allocated according to parameters such as rotational speed, load, and water level, solving the problem of poor adaptability of fixed models. The fault prediction accuracy is ≥97%. By adapting to equipment under different operating conditions and at different aging stages through the dynamic weight adaptive mechanism, the false alarm rate can be reduced by 60%, and the fault occurrence time prediction error is controlled within ±1 day.

[0052] Optionally, step 105 may specifically include: determining the initial fault prediction time of the hydraulic electromechanical equipment based on the equipment health index, fault evolution rate, and fault critical threshold corresponding to the equipment health index; determining the unrepairable period of the hydraulic electromechanical equipment based on the scheduling constraint nodes in the scheduling plan data and the meteorological forecast data of the hydraulic electromechanical equipment within the prediction period; and determining the fault prediction time and maintenance window period of the hydraulic electromechanical equipment based on the maintenance conflict coefficient corresponding to the unrepairable period and the initial fault prediction time.

[0053] For example, the initial fault prediction time can be calculated using the following formula: Δt0 = (HI - HI_threshold) / v; In the formula, Δt0 can represent the initial fault prediction time; HI can represent the equipment health index output by the fault prediction model; HI_threshold can be the preset fault threshold corresponding to the equipment health index, that is, the critical value of the equipment health index. When the equipment health index is lower than this threshold, it can be said that the equipment is prone to failure. It can be set to 0.3; v can represent the fault evolution rate output by the fault prediction model.

[0054] Furthermore, scheduling demand constraint analysis can be performed to extract key constraint nodes in the reservoir scheduling plan as scheduling constraint nodes, such as the flood season water level release period, peak power generation load period, and ecological flow guarantee period. Combined with the meteorological forecast data for the next 15 days, the impact of rainfall on the reservoir water level can be analyzed in detail, and the unrepairable period T can be determined based on the initial fault prediction time.

[0055] For example, an objective function can be constructed based on the maintenance conflict coefficient corresponding to the unrepairable period and the initial fault prediction time. The objective function can then be solved using a particle swarm optimization algorithm, dynamically correcting and co-optimizing the initial fault prediction time. The objective function can be expressed as: F=α×|Δt-Δt0|+β×σ(T,Δt); In the formula, α can represent the prediction accuracy weight, which can be 0.4, β can represent the scheduling matching weight, which can be 0.6, and σ(T,Δt) is the conflict coefficient between the maintenance period Δt and the non-maintainable period T. For example, σ=0 when there is no conflict, σ=0.5 when there is a partial conflict, and σ=1 when there is a complete conflict. Accordingly, the objective function F is solved by the particle swarm optimization algorithm to obtain the corrected fault prediction time Δt' and the optimal maintenance window [Δt'-3, Δt'-1].

[0056] In some embodiments, a fault prediction model can be trained based on full lifecycle sample data from 50+ identical devices, and a health decay curve can be fitted to quantify the equipment degradation stages. The health decay curve can trace the degradation trend of the equipment over the past year, helping to pinpoint key factors that accelerate degradation. Specifically, based on the slope of the health decay curve, the initial fault prediction time when the equipment health index reaches the fault critical threshold can be initially calculated, and then corrected based on reservoir scheduling data and meteorological forecast data. Furthermore, the predicted fault evolution rate can be adjusted based on multi-source heterogeneous monitoring data. For example, if the equipment load increases by 20% due to rising water levels, the fault evolution rate can be amplified by 1.15 times before calculating the fault prediction time. This avoids conflicts between maintenance plans and reservoir scheduling, provides real-time assessment of the operating status of hydropower units, and predicts the time of equipment failure based on monitoring data, providing maintenance personnel with accurate equipment information to facilitate the development of equipment maintenance plans.

[0057] In this way, a scheduling-prediction collaborative optimization mechanism is constructed, realizing a two-way feedback model for fault prediction and reservoir scheduling. By combining meteorological data for the next 7-15 days, reservoir water level control targets, power generation load plans, and other data, the fault prediction time is dynamically corrected, and the optimal maintenance window period that matches the scheduling needs is output, avoiding conflicts between maintenance and operation. The matching degree between maintenance and scheduling is ≥95%. The scheduling-prediction collaborative mechanism can ensure that maintenance plans avoid critical scheduling periods such as flood season and peak power generation period, improve equipment availability to over 99%, and reduce losses from ineffective downtime.

[0058] Optionally, the method in this embodiment may further include: generating multi-level early warning and maintenance strategies for water conservancy electromechanical equipment based on maintenance window periods, target fault types, and equipment health indices; adjusting the equipment operation plan corresponding to the water conservancy electromechanical equipment based on multi-level early warnings, and pushing the equipment operation plan and maintenance strategies to the reservoir scheduling system and / or the mobile terminal of maintenance personnel; obtaining maintenance feedback data of water conservancy electromechanical equipment, and using the maintenance feedback data to update the model parameters corresponding to the fault prediction model.

[0059] Specifically, by combining the optimal maintenance window, target fault type, and equipment health index, targeted maintenance strategies can be output. These strategies may include maintenance priorities (e.g., levels 1-5), key maintenance parts, recommended maintenance solutions (e.g., bearing replacement, stator winding overhaul, lubricant replacement), a list of required spare parts, and estimated maintenance time, which can be simultaneously pushed to the reservoir dispatch system and the mobile terminals of maintenance personnel.

[0060] Correspondingly, the water conservancy electromechanical equipment fault prediction system can monitor and issue early warnings for water conservancy electromechanical equipment in real time. Based on the steps of the above embodiments, the system can continuously collect real-time data of the equipment to achieve 24-hour uninterrupted health monitoring. When abnormal conditions are detected, multi-level warnings are generated. The multi-level warnings can be different levels of warnings generated according to the approaching time of the maintenance window, the severity of the target fault type, and the value of the equipment health index, such as a yellow warning indicating mild urgency, an orange warning indicating moderate urgency, and a red warning indicating moderate urgency.

[0061] For example, multiple fault thresholds corresponding to the equipment health index can be set to determine multi-level early warnings. For instance, when the equipment health index is below 0.7, it indicates mild degradation and a yellow warning is issued; when the equipment health index is below 0.5, it indicates moderate degradation and an orange warning is issued, along with a maintenance plan; when the equipment health index is below 0.3, it indicates severe degradation and a red warning is issued, along with the initiation of an emergency maintenance process. Simultaneously, the reservoir dispatching system is linked to adjust the operation plan, such as load, start and stop times, and the number of operating units. The adjusted equipment operation plan and specific maintenance strategies are then sent to the reservoir dispatching system and / or the mobile terminals of maintenance personnel, so that the reservoir dispatching system and maintenance personnel can make specific adjustments to avoid affecting core water conservancy tasks and reduce failure losses.

[0062] Correspondingly, maintenance feedback data can be collected, recording data such as fault confirmation results, repair locations, replaced parts, and maintenance effects during equipment maintenance, forming a maintenance feedback dataset. Using a Bayesian optimization algorithm, the maintenance feedback data is input into the fault prediction model to update the fault fingerprint database feature templates, the operating condition adaptation parameters of the dynamic weight LSTM layer, and the attention layer weight allocation rules, achieving iterative optimization of model parameters. This allows for the construction of a full lifecycle closed-loop optimization system, integrating monitoring data, prediction results, maintenance records, and fault verification data to build a closed-loop iterative mechanism. Through continuous updates to model parameters and the fault fingerprint database using the Bayesian optimization algorithm, a full-process evolution of "monitoring-prediction-maintenance-optimization" is achieved. The long-term accuracy and stability of the fault prediction model can be improved by 40%, and the closed-loop optimization mechanism ensures that the model's prediction accuracy decays by no more than 3% within one year as the equipment continues to evolve, far superior to existing fixed models.

[0063] Compared with related technologies, this embodiment can target the fault signal characteristics of water conservancy electromechanical equipment by using vibration signal wavelet packet decomposition and joint time-domain and frequency-domain feature extraction to separate the inherent fault characteristics of the equipment and environmental interference signals from the original vibration signal, capture weak fault signals in the early stage of equipment health degradation, and construct a fault fingerprint database for fault fingerprint matching to achieve accurate identification of early weak faults. Using a fault prediction model, based on the matching degree between the fused equipment monitoring features and the fault fingerprint database, the health index and fault evolution rate of the corresponding water conservancy electromechanical equipment are predicted. Combined with scheduling constraint nodes and meteorological forecast data, the fault prediction time and maintenance window period of the water conservancy electromechanical equipment are determined to ensure that the maintenance plan avoids key scheduling periods such as flood season and peak power generation period, reduce ineffective downtime losses, and reserve sufficient time for maintenance.

[0064] Based on the above Figure 1 The specific implementation of the method shown in this embodiment provides a fault prediction device for hydraulic electromechanical equipment, such as... Figure 2 As shown, the device includes: an acquisition module 31, a fusion module 32, a prediction module 33, and a determination module 34; The acquisition module 31 is used to acquire multi-source heterogeneous monitoring data corresponding to water conservancy electromechanical equipment; perform fault fingerprint matching on the equipment status data in the multi-source heterogeneous monitoring data to acquire the fault characteristics corresponding to the equipment status data; The fusion module 32 is used to fuse the fault characteristics and the equipment operating condition data in the multi-source heterogeneous monitoring data to obtain the equipment monitoring characteristics corresponding to the water conservancy electromechanical equipment. The prediction module 33 is used to predict the equipment health index and fault evolution rate of water conservancy electromechanical equipment based on the fault fingerprint matching degree of equipment monitoring characteristics. The determination module 34 is used to determine the fault prediction time and maintenance window period of the water conservancy electromechanical equipment based on the equipment health index, fault evolution rate and the scheduling plan data corresponding to the water conservancy electromechanical equipment.

[0065] In some examples of this embodiment, the acquisition module 31 is specifically configured to perform fault fingerprint matching on the equipment status data based on the fault fingerprint database corresponding to the water conservancy electromechanical equipment, and obtain the fault features corresponding to the water conservancy electromechanical equipment. The fault fingerprint database includes a set of fault features of different models of water conservancy electromechanical equipment during the aging process.

[0066] In some examples of this embodiment, the fusion module 32 is specifically configured to fuse fault features, operating condition features, and equipment aging coefficients corresponding to the hydraulic electromechanical equipment, as well as hydrological environmental features corresponding to hydrological environmental data in multi-source heterogeneous monitoring data, to obtain equipment monitoring features corresponding to the hydraulic electromechanical equipment. The equipment aging coefficient is determined based on the operating time and maintenance frequency of the hydraulic electromechanical equipment.

[0067] In some examples of this embodiment, the acquisition module 31 is specifically configured to perform wavelet packet decomposition on the vibration signal in the equipment status data to obtain the entropy features of the vibration signal in different frequency bands; perform joint time-domain and frequency-domain feature extraction on the vibration signal to obtain the time-domain features and frequency-domain features corresponding to the vibration signal; generate vibration features corresponding to the vibration signal based on the entropy features, time-domain features and frequency-domain features; and perform fault fingerprint matching with the fault fingerprint database to obtain the fault features corresponding to the hydraulic electromechanical equipment.

[0068] In some examples of this embodiment, the prediction module 33 is specifically configured to obtain the fault prediction model corresponding to the water conservancy electromechanical equipment. The fault prediction model is trained based on the full life cycle data of water conservancy electromechanical equipment of different models and different years of operation. Using the fault prediction model, based on the matching degree between the equipment monitoring features and the fault fingerprint database, the equipment health index and fault evolution rate corresponding to the water conservancy electromechanical equipment are predicted.

[0069] In some examples of this embodiment, the prediction module 33 is specifically configured to use the attention layer in the fault prediction model to adaptively adjust the feature weights corresponding to the device monitoring features based on the matching degree between the device monitoring features and the fault fingerprints of different fault types in the fault fingerprint database, and determine the target fault type corresponding to the device monitoring features; and predict the device health index and fault evolution rate based on the adjusted feature weights and the target fault type.

[0070] In some examples of this embodiment, the acquisition module 31 is further configured to generate multi-level early warning and maintenance strategies for water conservancy electromechanical equipment based on maintenance window period, target fault type and equipment health index; adjust the equipment operation plan corresponding to water conservancy electromechanical equipment according to multi-level early warning, and push the equipment operation plan and maintenance strategy to the reservoir scheduling system and / or the mobile terminal of operation and maintenance personnel; acquire maintenance feedback data of water conservancy electromechanical equipment, and use the maintenance feedback data to update the model parameters corresponding to the fault prediction model.

[0071] In some examples of this embodiment, the determining module 34 is specifically configured to determine the initial fault prediction time of the hydraulic electromechanical equipment based on the equipment health index, the fault evolution rate, and the fault critical threshold corresponding to the equipment health index; determine the unrepairable period of the hydraulic electromechanical equipment based on the scheduling constraint nodes in the scheduling plan data and the meteorological forecast data of the hydraulic electromechanical equipment within the prediction time period; and determine the fault prediction time and maintenance window period of the hydraulic electromechanical equipment based on the maintenance conflict coefficient corresponding to the unrepairable period and the initial fault prediction time.

[0072] Based on the above, Figure 1 Accordingly, this embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.Figure 1 The method shown.

[0073] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0074] Based on the above, Figure 1 The method shown, and Figure 2 To achieve the above objectives, the present application also provides an electronic device, comprising a storage medium and a processor; the storage medium for storing a computer program; and the processor for executing the computer program to implement the above-described virtual device embodiments. Figure 1 The method shown.

[0075] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0076] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0077] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0078] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms, or it can be implemented by hardware. The solution of this application can target the fault signal characteristics of hydraulic electromechanical equipment by employing vibration signal wavelet packet decomposition and joint time-domain and frequency-domain feature extraction to separate the inherent fault characteristics of the equipment and environmental interference signals from the original vibration signal, capture weak fault signals in the early stages of equipment health degradation, and construct a fault fingerprint database for fault fingerprint matching to achieve accurate identification of early weak faults. Using a fault prediction model, based on the fused equipment monitoring features and the fault fingerprint matching degree of the fault fingerprint database, the health index and fault evolution rate of the corresponding hydraulic electromechanical equipment are predicted. Combined with scheduling constraint nodes and meteorological forecast data, the fault prediction time and maintenance window period of the hydraulic electromechanical equipment are determined to ensure that maintenance plans avoid critical scheduling periods such as flood season and peak power generation periods, reduce ineffective downtime losses, and reserve sufficient time for maintenance.

[0079] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0080] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for predicting faults in hydraulic electromechanical equipment, characterized in that, include: Acquire multi-source heterogeneous monitoring data corresponding to water conservancy electromechanical equipment; Fault fingerprint matching is performed on the equipment status data in the multi-source heterogeneous monitoring data to obtain the fault characteristics corresponding to the equipment status data. By integrating the fault characteristics and the operating condition characteristics corresponding to the equipment operating condition data in the multi-source heterogeneous monitoring data, the equipment monitoring characteristics corresponding to the water conservancy electromechanical equipment are obtained. Based on the fault fingerprint matching degree of the equipment monitoring characteristics, the equipment health index and fault evolution rate of the water conservancy electromechanical equipment are predicted. Based on the equipment health index, the fault evolution rate, and the scheduling plan data corresponding to the water conservancy electromechanical equipment, the fault prediction time and maintenance window period of the water conservancy electromechanical equipment are determined.

2. The method for predicting faults in hydraulic electromechanical equipment according to claim 1, characterized in that, The step of performing fault fingerprint matching on the device status data in the multi-source heterogeneous monitoring data to obtain the fault features corresponding to the device status data includes: Based on the fault fingerprint database corresponding to the water conservancy electromechanical equipment, fault fingerprint matching is performed on the equipment status data to obtain the fault characteristics corresponding to the water conservancy electromechanical equipment. The fault fingerprint database includes a set of fault characteristics of different models of water conservancy electromechanical equipment during the aging process. The process of fusing the fault characteristics and the equipment operating condition data corresponding to the equipment operating condition data in the multi-source heterogeneous monitoring data to obtain the equipment monitoring characteristics corresponding to the hydraulic electromechanical equipment includes: By integrating the fault characteristics, the operating condition characteristics, and the equipment aging coefficient corresponding to the water conservancy electromechanical equipment, as well as the hydrological environment characteristics corresponding to the hydrological environment data in the multi-source heterogeneous monitoring data, the equipment monitoring characteristics corresponding to the water conservancy electromechanical equipment are obtained. The equipment aging coefficient is determined based on the operating time and maintenance frequency of the water conservancy electromechanical equipment.

3. The method for predicting faults in hydraulic electromechanical equipment according to claim 2, characterized in that, The step of matching fault fingerprints on the equipment status data based on the fault fingerprint database corresponding to the hydraulic electromechanical equipment to obtain the fault characteristics corresponding to the hydraulic electromechanical equipment includes: Wavelet packet decomposition is performed on the vibration signal in the device status data to obtain the entropy characteristics of the vibration signal in different frequency bands; The vibration signal is subjected to joint time-domain and frequency-domain feature extraction to obtain the time-domain and frequency-domain features corresponding to the vibration signal; Based on the entropy value characteristics, time domain characteristics, and frequency domain characteristics, the vibration characteristics corresponding to the vibration signal are generated; The vibration characteristics are matched with the fault fingerprint database to obtain the fault characteristics corresponding to the hydraulic electromechanical equipment.

4. The method for predicting faults in hydraulic electromechanical equipment according to claim 2, characterized in that, The method of predicting the equipment health index and fault evolution rate of the hydraulic electromechanical equipment based on the fault fingerprint matching degree of the equipment monitoring characteristics includes: Obtain the fault prediction model corresponding to the water conservancy electromechanical equipment. The fault prediction model is trained based on the full life cycle data of water conservancy electromechanical equipment of different models and different years of operation. Using the fault prediction model, based on the matching degree between the equipment monitoring features and the fault fingerprint database, the equipment health index and fault evolution rate of the water conservancy electromechanical equipment are predicted.

5. The method for predicting faults in hydraulic electromechanical equipment according to claim 4, characterized in that, The step of using the fault prediction model to predict the equipment health index and fault evolution rate of the hydraulic electromechanical equipment based on the matching degree between the equipment monitoring features and the fault fingerprint database includes: By utilizing the attention layer in the fault prediction model, the feature weights corresponding to the device monitoring features are adaptively adjusted based on the matching degree between the device monitoring features and the fault fingerprints of different fault types in the fault fingerprint database, and the target fault type corresponding to the device monitoring features is determined. Based on the adjusted feature weights and the target fault type, the equipment health index and the fault evolution rate are predicted.

6. The method for predicting faults in hydraulic electromechanical equipment according to claim 5, characterized in that, The method further includes: Based on the maintenance window period, the target fault type, and the equipment health index, generate multi-level early warning and maintenance strategies for the hydraulic electromechanical equipment; Based on the multi-level early warning, adjust the equipment operation plan corresponding to the water conservancy electromechanical equipment, and push the equipment operation plan and maintenance strategy to the reservoir dispatch system and / or the mobile terminal of operation and maintenance personnel; Obtain maintenance feedback data of the water conservancy electromechanical equipment, and use the maintenance feedback data to update the model parameters corresponding to the fault prediction model.

7. The method for predicting faults in hydraulic electromechanical equipment according to claim 1, characterized in that, The step of determining the fault prediction time and maintenance window period of the water conservancy electromechanical equipment based on the equipment health index, the fault evolution rate, and the scheduling plan data corresponding to the water conservancy electromechanical equipment includes: The initial fault prediction time of the hydraulic electromechanical equipment is determined based on the equipment health index, the fault evolution rate, and the fault critical threshold corresponding to the equipment health index. Based on the scheduling constraint nodes in the scheduling plan data and the meteorological forecast data of the water conservancy electromechanical equipment within the predicted time period, the unmaintainable period of the water conservancy electromechanical equipment is determined. Based on the maintenance conflict coefficient corresponding to the unrepairable period and the initial fault prediction time, the fault prediction time and maintenance window period of the water conservancy electromechanical equipment are determined.

8. A fault prediction device for hydraulic electromechanical equipment, characterized in that, include: The acquisition module is used to acquire multi-source heterogeneous monitoring data corresponding to water conservancy electromechanical equipment; Fault fingerprint matching is performed on the equipment status data in the multi-source heterogeneous monitoring data to obtain the fault characteristics corresponding to the equipment status data. The fusion module is used to fuse the fault characteristics and the operating condition characteristics corresponding to the equipment operating condition data in the multi-source heterogeneous monitoring data to obtain the equipment monitoring characteristics corresponding to the water conservancy electromechanical equipment. The prediction module is used to predict the equipment health index and fault evolution rate of the water conservancy electromechanical equipment based on the fault fingerprint matching degree of the equipment monitoring characteristics. The determination module is used to determine the fault prediction time and maintenance window period of the water conservancy electromechanical equipment based on the equipment health index, the fault evolution rate and the scheduling plan data corresponding to the water conservancy electromechanical equipment.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-7.

Images