Intelligent operation and maintenance method, device and equipment for power equipment and storage medium

Abstract

The invention discloses an intelligent operation and maintenance method, device and equipment for power equipment and a storage medium, and the method comprises the steps: S101, collecting multi-modal original data of the power equipment through a plurality of different types of sensors, and carrying out the time-space alignment and preprocessing of the multi-modal original data, and obtaining a time series data set; s102, after feature extraction is carried out on the time sequence data set, a multi-dimensional time sequence feature vector is constructed according to extracted features; s103, inputting the multi-dimensional time sequence feature vector into a pre-trained LSTM-GRU attention network to obtain a health score and a fault probability of the power equipment; and S104, based on an HR-Tree algorithm and the health score, calculating a risk value of each power device, and according to the risk value and the fault probability, generating an automatic operation and maintenance scheme. According to the method, the fault missed judgment rate in a composite scene can be effectively reduced, and the transient abnormal response speed and the operation and maintenance efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, and in particular to a method, apparatus, equipment and storage medium for intelligent operation and maintenance of power equipment. Background Technology

[0002] The safe operation of power equipment is crucial to the smart grid, and its fault prediction relies on in-depth analysis of multi-source data (such as electrical parameters and historical operation and maintenance records).

[0003] Existing prediction methods mainly include monitoring schemes based on single-parameter thresholds, classification schemes based on traditional machine learning, and manual inspection and data visualization schemes. Among these, the single-parameter threshold-based monitoring scheme collects single-phase current data and sets a fixed threshold; an alarm is triggered when the current exceeds the threshold. However, this method cannot identify hidden risks, with a false negative rate exceeding 25%. The classification scheme based on traditional machine learning uses support vector machines to analyze three-dimensional electrical data, relying on manual feature extraction, resulting in a false positive rate exceeding 40% in scenarios with sudden load changes. The manual inspection and data visualization scheme transmits data to a cloud platform for display, lacking an automatic early warning mechanism, resulting in a fault detection delay of several hours and an inability to quantify data correlations.

[0004] In summary, existing technologies generally suffer from drawbacks such as limited data dimensions, insufficient dynamic modeling capabilities, and reliance on manual operation and maintenance decisions. These shortcomings result in high rates of missed detection of complex faults, delayed responses to transient anomalies, and low operational efficiency, failing to meet actual application needs. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method, apparatus, equipment and storage medium for intelligent operation and maintenance of power equipment, so as to improve the above-mentioned problems.

[0006] This invention provides an intelligent operation and maintenance method for power equipment, comprising: S101, collect multimodal raw data of power equipment through multiple different types of sensors, and perform spatiotemporal alignment and preprocessing on the multimodal raw data to obtain a time series dataset; S102, after extracting features from the time series dataset, a multi-dimensional time series feature vector is constructed based on the extracted features; S103, the multidimensional temporal feature vector is input into a pre-trained LSTM-GRU attention network to obtain the health score and failure probability of the power equipment; S104. Based on the HR-Tree algorithm and the health score, calculate the risk value of each power device, and generate an automated operation and maintenance plan based on the risk value and the failure probability.

[0007] Preferably, step S101 specifically includes: Multiple electrical parameters of electrical equipment and environmental data are acquired simultaneously by multiple different types of sensors to obtain multimodal raw data; Kalman filtering algorithm is used to suppress noise in electrical parameters; The filtered multimodal raw data is segmented using the sliding window technique to generate data frames of fixed length. For data frames of fixed length, a time-series dataset with a unified timeline is generated through a timestamp synchronization mechanism.

[0008] Preferably, in step S102, the extracted features include time-domain features, frequency-domain features, time-series correlation features, and trend features; wherein: The time-domain features include mean, variance, and peak-to-peak value; The frequency domain features include harmonic content and spectral entropy obtained through FFT transformation; The time-series correlation features are obtained by constructing an autocorrelation matrix with a lag order of 12 and analyzing the delay correlation between parameters. The trend features are obtained by using Savitzky-Golay filtering to extract data trend terms and identifying abnormal trends that are continuously rising or falling.

[0009] Preferably, in step S103, the LSTM-GRU attention network is trained using historical data containing fault samples; during training: the loss function is the cross-entropy loss function; the optimizer is the Adam optimizer with β1=0.9, β2=0.999, and a learning rate of 0.001; regularization uses Dropout with a dropout rate of 0.2 and an early stopping mechanism to prevent overfitting.

[0010] Preferably, in step S104, the risk value RI of the electrical equipment is DI × DH, where: DI stands for Equipment Importance, which is determined using the Analytic Hierarchy Process (AHP) based on static attributes in existing ledger data. DH represents the equipment vulnerability level, calculated by combining the health score predicted by the LSTM-GRU attention network and the historical failure frequency.

[0011] Preferably, when RI < 0.3, it indicates that the electrical equipment is of low risk; When 0.3 ≤ RI < 0.7, it indicates that the electrical equipment is of low risk; When RI ≥ 0.7, it indicates that the electrical equipment is of high risk; When RI≥0.7 and the output failure probability>0.8, a red work order is automatically generated, which includes: fault type, location information, priority level and maintenance recommendations.

[0012] Preferably, it further includes: Inspection paths are generated based on risk values. These paths are generated using a path optimization algorithm, where the optimization objective of the algorithm is to minimize the total travel time and maximize the processing priority of high-risk work orders. The optimal strategy π(s) is generated through iterative training.

[0013] This invention also provides an intelligent operation and maintenance device for power equipment, comprising: The data acquisition unit is used to acquire multimodal raw data of power equipment through multiple different types of sensors, and to perform spatiotemporal alignment and preprocessing on the multimodal raw data to obtain a time-series dataset; The feature extraction unit is used to extract features from the time series dataset and then construct a multi-dimensional time series feature vector based on the extracted features. An attention network unit is used to input the multidimensional temporal feature vector into a pre-trained LSTM-GRU attention network to obtain the health score and fault probability of the power equipment. The decision-making unit is used to calculate the risk value of the power equipment based on the HR-Tree algorithm and the health score, and generate an automated operation and maintenance plan based on the risk value and the failure probability.

[0014] This invention also provides an intelligent operation and maintenance device for power equipment, which includes a memory and a processor. The memory stores a computer program, which can be executed by the processor to implement the intelligent operation and maintenance method for power equipment as described above.

[0015] This invention also provides a computer-readable storage medium storing a computer program that can be executed by a processor of the device in which the computer-readable storage medium is located, so as to implement the intelligent operation and maintenance method for power equipment as described above.

[0016] In summary, compared with the prior art, this embodiment has the following advantages: 1. This embodiment integrates the time domain, frequency domain, and time series correlation features (such as autocorrelation matrix and trend term) of electrical parameters (voltage, current, harmonics). It can construct a 9-dimensional feature vector based on existing data without adding new hardware. Compared with the existing static statistics (such as mean current) that only calculate a single parameter, it can effectively reduce the false negative rate in complex scenarios. 2. This embodiment constructs a neural network model that includes an LSTM layer (capturing long-term dependencies with a 48-point memory window), a GRU layer (handling short-term fluctuations), and an attention layer (dynamically allocating feature weights). It automatically extracts temporal features through deep learning, replacing the existing technology of "static threshold + manual analysis". The early warning time is increased from "none" to more than 30 minutes, and the false judgment rate is reduced from 40% to 8%. 3. This embodiment adopts a closed-loop system of "feature analysis - risk assessment - work order generation - path optimization". Compared with the existing technology of "manual analysis + experience decision-making", the fault handling cycle is shortened from 4 hours to 45 minutes, which not only improves the fault handling efficiency, but also reduces the operation and maintenance cost by more than 40%. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the intelligent operation and maintenance method for power equipment provided in the first embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the structure of the intelligent operation and maintenance device for power equipment provided in the second embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To better understand the technical solution of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0022] It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0023] Please see Figure 1The first embodiment of the present invention provides a method for intelligent operation and maintenance of power equipment, which can be executed by a power equipment intelligent operation and maintenance device (hereinafter referred to as the intelligent operation and maintenance device), specifically, by one or more processors within the intelligent operation and maintenance device, to achieve the following steps: S101: Collect multimodal raw data of power equipment through multiple different types of sensors, and perform spatiotemporal alignment and preprocessing on the multimodal raw data to obtain a time-series dataset.

[0024] In this embodiment, the intelligent operation and maintenance equipment can be a smart mobile terminal, tablet computer, desktop computer, laptop computer, workstation or server, or other device with data storage and data processing capabilities. This invention does not impose any specific limitations.

[0025] In this embodiment, the multimodal raw data may include electrical parameters (such as voltage RMS, current peak factor CF, etc.) and environmental data (such as ambient temperature, humidity, etc.), which can be set according to actual needs, and the present invention does not make specific limitations.

[0026] In this embodiment, after obtaining the original multimodal data, it is necessary to perform spatiotemporal alignment and preprocessing, specifically including: First, the Kalman filter algorithm is used to suppress electrical parameter noise to obtain time series data; Next, the time-series data is segmented using the sliding window technique to generate data frames of fixed length. The sliding window can be set to a window size of 1024 points and a step size of 256 points. However, in other embodiments of the present invention, the window size and step size can also be set according to actual needs. All of these solutions are within the protection scope of the present invention.

[0027] Finally, a time-series dataset with a unified timeline is generated through a timestamp synchronization mechanism.

[0028] In this embodiment, the IEEE 1588 protocol can be used to ensure that the data timestamp error is less than 100μs, thereby solving the problem of multi-source data asynchrony.

[0029] S102, after extracting features from the time series dataset, a multidimensional time series feature vector is constructed based on the extracted features.

[0030] In this embodiment, the extracted features include time-domain features, frequency-domain features, time-series correlation features, and trend features; wherein: The time-domain features include mean, variance, and peak-to-peak value.

[0031] For example, the effective value of voltage (V) can be calculated. RMS ): Where n is the number of sampling points within the window, such as 1024 points corresponding to a 102.4ms window.

[0032] Peak current factor (CF): It is used to identify transient anomalies caused by impact loads.

[0033] The frequency domain features include harmonic content and spectral entropy obtained through FFT transformation.

[0034] Specifically: Perform an FFT transform on the voltage signal and calculate the 5th harmonic content:

[0035] in, f 0 is the fundamental frequency. X ( f 0) represents the frequency domain amplitude.

[0036] The voltage signal is converted from the time domain to the frequency domain by FFT transformation, and the ratio of the amplitude of the 5th harmonic to the amplitude of the fundamental frequency is calculated (the 5th harmonic is the most common non-fundamental component in the power system, mainly from nonlinear loads such as frequency converters and rectifiers).

[0037] Spectral entropy (SE):

[0038] Based on the information entropy theory, the SE value measures the uniformity of energy distribution in the frequency domain. The larger the SE value, the more complex the harmonic components in the signal (such as containing multiple higher harmonics).

[0039] The time-series correlation features are obtained by constructing an autocorrelation matrix with a lag order of 12 and analyzing the delay correlation between parameters.

[0040] First, construct the autocorrelation coefficient matrix (lag order p=12):

[0041] Used to analyze the correlation of current parameters at different lag times, such as ρ6>0.6 indicating that the current current is strongly correlated with the current 6 sampling points ago.

[0042] The trend features are obtained by using Savitzky-Golay filtering to extract data trend terms and identifying abnormal trends that are continuously rising or falling.

[0043] The Savitzky-Golay filter fitting formula is:

[0044] The window width is 2m+1=11. A second-order polynomial is fitted to extract the slope of the harmonic distortion rate increase over a continuous hour.

[0045] In this embodiment, after obtaining the characteristics of multiple domains, cross-domain correlation analysis can also be performed, for example, by calculating the correlation degree MI between load (L) and harmonics (H) using a mutual information algorithm:

[0046] When MI>0.7, it is determined to be a strong correlation, and a composite fault warning can be triggered at this time.

[0047] S103, the multidimensional temporal feature vector is input into a pre-trained LSTM-GRU attention network to obtain the health score and failure probability of the power equipment.

[0048] In this embodiment, the LSTM-GRU attention network includes: LSTM layers, which include: Memory window: 48 sampling points (corresponding to 4.8 seconds, sampling frequency 10kHz), covering typical transient processes in power systems (such as voltage fluctuations before a short-circuit fault, which typically last from tens of milliseconds to several seconds).

[0049] Neuron configuration: 128 neurons, which selectively retain historical information through gating mechanisms (forget gate, input gate, output gate), such as capturing the slow drift of parameters caused by equipment aging (such as the trend of oil temperature rising by 0.5℃ per year).

[0050] Output characteristics: The output dimension is R 48×128 Long-term characteristic sequences The hidden state at each time step contains global dependency information from point 1 to point t.

[0051] The GRU layer, used for short-term volatility analysis, has the following characteristics: Simplified structure: GRU is a lightweight variant of LSTM, which combines the forget gate and input gate into an update gate, improving computational efficiency by 30% and making it suitable for processing high-frequency sampled data (such as 10kHz current fluctuations).

[0052] Short-term focus: 64 neurons focus on local features within the current window (such as a voltage drop in the last second), with an output dimension of R. 64 short-term eigenvectors ht GRU It can detect transient anomalies (such as current surges lasting 2 seconds).

[0053] Attention layer: Dynamic feature weighting mechanism The core formula is:

[0054] Its calculation process includes: Feature fusion: through the weight matrix W h ∈R 32×128 and W g ∈R 32×64 The long-term features of LSTM and the short-term features of GRU are projected onto a 32-dimensional space, and nonlinearity is introduced through the hyperbolic tangent activation function (tanh).

[0055] Weight generation: trainable parameters w∈R 32 The fused features are weighted and summed to obtain a scalar attention score, which is then normalized using Softmax to generate time-step weights. ,and .

[0056] Here, weight This indicates the level of attention the model pays to the t-th sampling point. For example, when the input sequence contains the event "voltage drop for 2 seconds" (corresponding to time step t=10-20), the attention layer will automatically increase the weight of these time steps (e.g., α). 10-20 The sum increased from an average of 2% to 30%, enhancing the extraction of abnormal features.

[0057] In this embodiment, the LSTM-GRU attention network needs to be trained before use. Training is performed using historical data containing fault samples; the training configuration is as follows: The loss function used is the cross-entropy loss function; the optimizer used is the Adam optimizer with β1=0.9, β2=0.999 and a learning rate of 0.001; regularization uses Dropout with a dropout rate of 0.2 and an early stopping mechanism to prevent overfitting.

[0058] In this embodiment, after the LSTM-GRU attention network is trained, a multi-dimensional feature vector, such as a 9-dimensional feature vector [V], is input to it. RMS I RMS [PF, H5, SE, ρ1, ρ6, Trend, MI], and then the LSTM-GRU attention network can output the health score and failure probability corresponding to the current electrical equipment based on the input feature vector.

[0059] S104. Based on the HR-Tree algorithm and the health score, calculate the risk value of each power device, and generate an automated operation and maintenance plan based on the risk value and the failure probability.

[0060] In this embodiment, the risk assessment model can employ the HR-Tree algorithm. In the HR-Tree algorithm, the risk value RI of the electrical equipment can be expressed as RI = DI × DH, where: DI stands for Equipment Importance, which is determined by the Analytic Hierarchy Process (AHP) based on static attributes in existing ledger data.

[0061] Specifically, for the i-th electrical device, its importance can be defined as:

[0062] Where, x i,1 The voltage level (with a weight of w1, which can be set to 0.3361) reflects the hierarchical importance of the equipment in the power grid (e.g., 35kV→1, 110kV→2, with higher voltage levels corresponding to larger standardized values).

[0063] x i,2 Indicates the power supply area (with a weight of w2, which can be set to 0.2301): measures the importance of the power supply range (e.g., assigning higher values ​​to core urban areas, industrial parks, etc.).

[0064] x i,2 The number of users (with a weight of w3, which can be set to 0.4338) represents the scale of users directly affected by the device; the more users there are, the greater the weight percentage.

[0065] In this embodiment, before calculation, these static attributes need to be normalized (e.g., Min-Max standardization) to ensure that indices with different dimensions can be directly linearly weighted. For example:

[0066] Thus, the final DI is guaranteed i The value ranges from [0,1], and the larger the value, the more important the electrical equipment is.

[0067] DH represents the equipment vulnerability level, which is calculated by combining the health score predicted by the LSTM-GRU attention network with the historical failure frequency.

[0068] Specifically, it can be defined as:

[0069] Among them, S i The health score is output by the LSTM-GRU attention network mentioned above, with a value range of [0,1]. The smaller the value, the worse the health status.

[0070] F i Failure count: Represents the number of failures occurring within a recent period (e.g., the past year), and needs to be normalized to the [0,1] interval (e.g., F). i= (Historical maximum number of failures / Actual number of failures).

[0071] In this embodiment, the weight of the health score can be... w s Set the weight to be higher than the number of failures. w F This reflects a priority focus on real-time potential hazards; for example, it can... w s Set to 0.6, w F It's set to 0.4, but you can also adjust the value according to your needs. For example, if you want to show more focus on historical faults, you can set a different weight. w s Weight less than the number of failures w F These solutions are all within the protection scope of this invention.

[0072] In this embodiment, after calculating the risk value RI, risk classification can be performed based on the risk value RI. For example, the following can be defined: Green (Low Risk): RI < 0.3 Yellow (Medium Risk): 0.3 ≤ RI < 0.7 Red (High Risk): RIi ≥ 0.7 Then, by combining the risk level and the probability of failure, early warnings or decisions can be made. For example: When RI≥0.7 and the model outputs a "winding overheat" probability>0.8, a red work order is automatically generated, containing the following: Fault location: based on feature association rules (such as ">>" mapping to the B-phase winding of the main transformer); Maintenance recommendation: "Check the operating status of the cooling system fan and test the DC resistance of the winding."

[0073] Furthermore, this embodiment can also generate the optimal inspection path based on the risk level. Specifically, this embodiment can generate the optimal inspection path through a path optimization algorithm (Q-learning). In the path optimization algorithm: State space: S = {Device location, risk level, and skill matching of maintenance personnel} Action space: A = {Go to electrical equipment 1, Go to electrical equipment 2, ..., End inspection} Reward function: R= ttravel+a•δ (Process red work orders) 50•δ(timeout) where ttravel is the travel time, δ is the indicator function, and a is the reward points for processing red work orders, which can be set to 100 points.

[0074] The optimization objective of this embodiment is to minimize the total travel time and maximize the processing priority of high-risk work orders. Through iterative training using the above configuration parameters, the optimal strategy π(s), i.e., the optimal inspection path, can be generated.

[0075] In summary, compared with the prior art, this embodiment has the following advantages: 4. This embodiment integrates the time domain, frequency domain, and time series correlation features (such as autocorrelation matrix and trend term) of electrical parameters (voltage, current, harmonics). It can construct a 9-dimensional feature vector based on existing data without adding new hardware. Compared with the existing static statistics (such as the mean current) that only calculate a single parameter, it can effectively reduce the false negative rate in complex scenarios. 5. This embodiment constructs a neural network model that includes an LSTM layer (capturing long-term dependencies with a 48-point memory window), a GRU layer (handling short-term fluctuations), and an attention layer (dynamically allocating feature weights). It automatically extracts time-series features through deep learning, replacing the existing technology of "static threshold + manual analysis". The early warning time is increased from "none" to more than 30 minutes, and the false judgment rate is reduced from 40% to 8%.

[0076] 6. This embodiment adopts a closed-loop system of "feature analysis - risk assessment - work order generation - path optimization". Compared with the existing technology of "manual analysis + experience decision-making", the fault handling cycle is shortened from 4 hours to 45 minutes, which not only improves the fault handling efficiency, but also reduces the operation and maintenance cost by more than 40%.

[0077] Please see Figure 2 The second embodiment of the present invention also provides an intelligent operation and maintenance device for power equipment, which includes: The data acquisition unit 210 is used to acquire multimodal raw data of power equipment through multiple different types of sensors, and to perform spatiotemporal alignment and preprocessing on the multimodal raw data to obtain a time-series dataset. Feature extraction unit 220 is used to extract features from the time series dataset and then construct a multi-dimensional time series feature vector based on the extracted features. Attention network unit 230 is used to input the multidimensional temporal feature vector into a pre-trained LSTM-GRU attention network to obtain the health score and fault probability of the power equipment. Decision unit 240 is used to calculate the risk value of the power equipment based on the HR-Tree algorithm and the health score, and generate an automated operation and maintenance plan based on the risk value and the failure probability.

[0078] The third embodiment of the present invention also provides an intelligent operation and maintenance device for power equipment, which includes a memory and a processor. The memory stores a computer program, which can be executed by the processor to realize the intelligent operation and maintenance method for power equipment as described above.

[0079] The fourth embodiment of the present invention also provides a computer-readable storage medium storing a computer program, which can be executed by the processor of the device where the computer-readable storage medium is located, so as to realize the intelligent operation and maintenance method of power equipment as described above.

[0080] For example, the various devices and process steps described above can be implemented by a computer program, which can be divided into one or more units, which are stored in the memory and executed by the processor to complete the present invention.

[0081] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0082] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the present invention by running or executing the computer programs and / or modules stored in the memory and by calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0083] If the unit integrated into the electronic device or printer is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0084] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0085] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for intelligent operation and maintenance of power equipment, characterized in that, include: S101, collect multimodal raw data of power equipment through multiple different types of sensors, and perform spatiotemporal alignment and preprocessing on the multimodal raw data to obtain a time series dataset; S102, after extracting features from the time series dataset, a multi-dimensional time series feature vector is constructed based on the extracted features; S103, the multidimensional temporal feature vector is input into a pre-trained LSTM-GRU attention network to obtain the health score and failure probability of the power equipment; S104. Based on the HR-Tree algorithm and the health score, calculate the risk value of each power device, and generate an automated operation and maintenance plan based on the risk value and the failure probability.

2. The intelligent operation and maintenance method for power equipment according to claim 1, characterized in that, Step S101 specifically includes: Multiple electrical parameters of electrical equipment and environmental data are acquired simultaneously by multiple different types of sensors to obtain multimodal raw data; Kalman filtering algorithm is used to suppress noise in electrical parameters; The filtered multimodal raw data is segmented using the sliding window technique to generate data frames of fixed length. For data frames of fixed length, a time-series dataset with a unified timeline is generated through a timestamp synchronization mechanism.

3. The intelligent operation and maintenance method for power equipment according to claim 1, characterized in that, In step S102, the extracted features include time-domain features, frequency-domain features, time-series correlation features, and trend features; among which: The time-domain features include mean, variance, and peak-to-peak value; The frequency domain features include harmonic content and spectral entropy obtained through FFT transformation; The time-series correlation features are obtained by constructing an autocorrelation matrix with a lag order of 12 and analyzing the delay correlation between parameters. The trend features are obtained by using Savitzky-Golay filtering to extract data trend terms and identifying abnormal trends that are continuously rising or falling.

4. The intelligent operation and maintenance method for power equipment according to claim 1, characterized in that, In step S103, the LSTM-GRU attention network is trained using historical data containing fault samples; During training: the cross-entropy loss function is used; the Adam optimizer with β1=0.9, β2=0.999 and a learning rate of 0.001 is used; regularization uses Dropout with a dropout rate of 0.2 and an early stopping mechanism to prevent overfitting.

5. The intelligent operation and maintenance method for power equipment according to claim 1, characterized in that, In step S104, the risk value RI of the electrical equipment is DI × DH, where: DI stands for Equipment Importance, which is determined using the Analytic Hierarchy Process (AHP) based on static attributes in existing ledger data. DH represents the equipment vulnerability level, calculated by combining the health score predicted by the LSTM-GRU attention network and the historical failure frequency.

6. The intelligent operation and maintenance method for power equipment according to claim 5, characterized in that, When RI < 0.3, it indicates that the electrical equipment is of low risk; When 0.3 ≤ RI < 0.7, it indicates that the electrical equipment is of low risk; When RI ≥ 0.7, it indicates that the electrical equipment is of high risk; When RI≥0.7 and the output failure probability>0.8, a red work order is automatically generated, which includes: fault type, location information, priority level and maintenance recommendations.

7. The intelligent operation and maintenance method for power equipment according to claim 6, characterized in that, Also includes: Inspection paths are generated based on risk values. These paths are generated using a path optimization algorithm, where the optimization objective of the algorithm is to minimize the total travel time and maximize the processing priority of high-risk work orders. The optimal strategy π(s) is generated through iterative training.

8. An intelligent operation and maintenance device for power equipment, characterized in that, include: The data acquisition unit is used to acquire multimodal raw data of power equipment through multiple different types of sensors, and to perform spatiotemporal alignment and preprocessing on the multimodal raw data to obtain a time-series dataset; The feature extraction unit is used to extract features from the time series dataset and then construct a multi-dimensional time series feature vector based on the extracted features. An attention network unit is used to input the multidimensional temporal feature vector into a pre-trained LSTM-GRU attention network to obtain the health score and fault probability of the power equipment. The decision-making unit is used to calculate the risk value of the power equipment based on the HR-Tree algorithm and the health score, and generate an automated operation and maintenance plan based on the risk value and the failure probability.

9. An intelligent operation and maintenance device for power equipment, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be executed by the processor to implement the intelligent operation and maintenance method for power equipment as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The device contains a computer program that can be executed by a processor of the device where the computer-readable storage medium is located, to implement the intelligent operation and maintenance method for power equipment as described in any one of claims 1 to 7.

Images