Power transmission and distribution line node operation and maintenance method and system based on cloud computing

Abstract

The invention discloses a power transmission and distribution line node operation and maintenance method and system based on cloud computing, and the method comprises the steps: collecting operation data, such as a node voltage amplitude and a current effective value, based on a cloud computing platform, carrying out the vectorization, carrying out the modeling of a time sequence characteristic through an optimized long-short-term memory network, analyzing the spatial dependence and parameter relation through a Transformer model, generating a depth characteristic, and carrying out the calculation of the depth characteristic. The fault type is identified in combination with historical fault data, and a model is constructed according to edge node operation and maintenance parameters to output an optimal operation and maintenance strategy to drive a terminal to operate. The system is correspondingly provided with a data acquisition and feature extraction unit, an optimized LSTM processing unit, a Transform feature analysis unit, a fault identification unit, an operation and maintenance strategy decision unit and an operation and maintenance execution driving unit for cooperative work. According to the method, multiple technologies are deeply fused, line operation and operation and maintenance parameters are comprehensively considered, and efficient and accurate operation and maintenance of power transmission and distribution line nodes are realized.

Description

Technical Field

[0001] This invention relates to the field of power transmission and distribution line management, and in particular to a cloud computing-based method and system for the operation and maintenance of power transmission and distribution line nodes. Background Technology

[0002] With the development of smart grids, the scale and complexity of transmission and distribution lines are constantly increasing, placing higher demands on the efficiency and accuracy of line node operation and maintenance. Traditional transmission and distribution line node operation and maintenance largely rely on manual inspections and simple data monitoring, which struggles to quickly and accurately identify potential faults when faced with massive amounts of complex data. Furthermore, existing operation and maintenance systems lack the ability to deeply mine time-series data and spatial correlation information, failing to comprehensively grasp the operational status of line nodes. This leads to unscientific operation and maintenance decisions, easily resulting in over-maintenance or untimely maintenance, causing resource waste and failing to guarantee the safe and stable operation of transmission and distribution lines.

[0003] The rise of cloud computing technology has brought new ideas to the operation and maintenance of power transmission and distribution line nodes. However, current cloud-based operation and maintenance methods still have significant shortcomings in data processing and operation and maintenance strategy formulation. First, most solutions use a single algorithm to process data, which cannot fully utilize the temporal and spatial characteristics of power transmission and distribution line operation data, resulting in limited accuracy and timeliness of fault diagnosis. Second, the formulation of operation and maintenance strategies does not fully consider the actual operation and maintenance conditions of edge nodes, such as geographical location and accessibility of maintenance personnel, making the formulated strategies less feasible in actual implementation and unable to meet diverse operation and maintenance needs. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of existing technologies, this invention provides a cloud computing-based method and system for the operation and maintenance of power transmission and distribution line nodes.

[0005] The technical solution adopted in this invention is a cloud computing-based operation and maintenance method for power transmission and distribution line nodes, comprising the following steps:

[0006] S1: Through the cloud computing platform, feature extraction is performed on the real-time operating data collected from each node of the power transmission and distribution line. The real-time operating data includes node voltage amplitude, current effective value, active power, reactive power and ambient temperature parameters. The extracted multi-dimensional feature data is then vectorized.

[0007] S2: Construct an optimized Long Short-Term Memory (LSTM) network structure. Introduce an adaptive weight adjustment module between the input layer and hidden layer of the LSM network. Input the vectorized feature data into the optimized LSM network. Use this network to model the time series characteristics of the power transmission and distribution line node operation data and obtain a preliminary feature representation containing time dimension context information.

[0008] S3: Input the preliminary feature representation into the Transformer model. Through the multi-head attention mechanism in the Transformer model, perform correlation analysis on the features of transmission and distribution line nodes from different dimensions and scales, capture the spatial dependencies between nodes and the nonlinear relationships between parameters, and generate a deep feature representation that integrates temporal and spatial information.

[0009] S4: Based on the historical fault data and deep feature representation of transmission and distribution line nodes, construct a fault feature database, and use preset fault classification rules to perform fault mode recognition on the deep feature representation and classify fault types.

[0010] S5: Based on the fault type and various parameters of edge node operation and maintenance, including node geographical location, accessibility of maintenance personnel, and spare parts inventory, an operation and maintenance strategy decision model is established, and the optimal operation and maintenance strategy for transmission and distribution line nodes with different fault types is output through the model.

[0011] S6: Send the optimal operation and maintenance strategy to the corresponding transmission and distribution line node operation and maintenance terminal, and drive the operation and maintenance terminal to execute operation and maintenance operations.

[0012] Furthermore, in the optimized Long Short-Term Memory network, the weight update of the adaptive weight adjustment module adopts the following formula:

[0013]

[0014] Among them, W t Let W be the weight matrix of the adaptive weight adjustment module at time t. t+1 Here is the weight matrix updated at time t+1; n represents the number of dimensions of the input features; α i,t f is the importance coefficient of the i-th feature at time t, which is calculated based on the correlation between the voltage amplitude and effective current value of the transmission and distribution line node and historical fault data; i,t Let be the value of the i-th feature at time t after processing by the activation function; σ is the activation function.

[0015] Furthermore, in the calculation of the multi-head attention mechanism of the Transformer model, the distance parameter between transmission and distribution line nodes is introduced, and the formula for calculating the attention weight is as follows:

[0016]

[0017] Where Q, K, and V are the query matrix, key matrix, and value matrix, respectively; d kis the dimension of the key matrix; Dist is the distance matrix between nodes of the transmission and distribution line, and the element values ​​represent the physical distance between the corresponding nodes; β is the distance weight coefficient, which is dynamically adjusted according to the topology of the transmission and distribution line and the importance of the nodes; softmax is the normalized exponential function.

[0018] Furthermore, during the construction of the fault feature database, the deep feature representation is dimensionality-reduced using a dimensionality-reduction formula based on the voltage fluctuation coefficient and current distortion rate of transmission and distribution line nodes:

[0019]

[0020] Where X is the original depth feature matrix, X new The feature matrix after dimensionality reduction; PCA is the principal component analysis method; diag represents the operation of generating a diagonal matrix; λ1 and λ2 are weight coefficients; VF i CD represents the voltage fluctuation coefficient at the i-th transmission and distribution line node, reflecting the degree of change in voltage amplitude over time. i Let be the current distortion rate of the i-th transmission and distribution line node, which measures the degree to which the current waveform deviates from a sine wave.

[0021] Furthermore, in the operation and maintenance strategy decision model, considering the load rate (LR) and failure rate (FR) of the transmission and distribution line nodes, the optimal operation and maintenance strategy is determined using the following decision formula:

[0022]

[0023] Where OSM is the optimal operation and maintenance strategy; M is the set of all available operation and maintenance strategies; m represents a specific operation and maintenance strategy; LR m FR m These represent the load rate and failure rate of transmission and distribution line nodes when using operation and maintenance strategy m; Cost m γ1 represents the operation and maintenance cost of adopting operation and maintenance strategy m; γ2, γ3 are weighting coefficients set according to the importance of transmission and distribution lines and the allocation of operation and maintenance resources.

[0024] Furthermore, in step S2, the optimized Long Short-Term Memory (LSTM) network also includes a forgetting gate control unit, and the output calculation formula for the forgetting gate is:

[0025] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0026] Among them, f t W is the output of the forget gate at time t. f h is the weight matrix of the forget gate; t-1x is the output of the hidden layer at the previous time step; t b is the input feature vector at time t; f σ is the bias vector of the forget gate; σ is the activation function. The forget gate control unit controls the degree of retention of historical information based on the active power change trend and reactive power compensation requirements of the transmission and distribution line nodes.

[0027] Furthermore, in step S3, an information fusion mechanism based on the real-time temperature gradient and humidity change rate of the transmission and distribution line nodes is introduced between the output of the encoding layer and the input of the decoding layer of the Transformer model. The fusion formula is as follows:

[0028] I fused =δ1·O enc +δ2·I dec +δ3·Env

[0029] Among them, I fused For the merged information; O enc For the output of the coding layer; I dec δ1, δ2, and δ3 are the inputs to the decoding layer; Env is an environmental information vector containing the real-time temperature gradient and humidity change rate of the transmission and distribution line nodes; δ1, δ2, and δ3 are fusion weight coefficients set according to the degree of influence of environmental factors on the operation of transmission and distribution lines.

[0030] Furthermore, in step S4, during fault mode identification, a fault feature enhancement formula based on the phase difference and impedance change rate between transmission and distribution line nodes is used:

[0031]

[0032] Among them, F enhanced The enhanced fault feature vector; F original This represents the original fault feature vector; ΔZ represents the change in phase difference between nodes i and j of the power transmission and distribution line. ij τ is the impedance change rate between nodes i and j; τ is the adjustment coefficient, which is set according to the voltage level and line length of the transmission and distribution line; sigmoid is the activation function.

[0033] Furthermore, in step S5, the operation and maintenance strategy decision model also combines the insulation aging index IAI of the transmission and distribution line nodes and the service life SL of the equipment, and adopts an improved decision formula:

[0034]

[0035] Among them, OSM ' The improved optimal operation and maintenance strategy; IAI m SL mThese represent the insulation aging index and equipment service life of transmission and distribution line nodes when the operation and maintenance strategy m is adopted; Time m The estimated maintenance time required to adopt maintenance strategy m; γ ' 1. γ ' 2. γ ' 3. γ ' 4 represents the weighting coefficients set according to the operational safety level and maintenance urgency of power transmission and distribution lines.

[0036] A cloud-based power transmission and distribution line node operation and maintenance system includes:

[0037] The data acquisition and feature extraction unit is used to extract features and vectorize the real-time operating data collected from each node of the power transmission and distribution line through the cloud computing platform.

[0038] The LSTM processing unit is optimized and connected to the data acquisition and feature extraction unit to construct an optimized long short-term memory network structure, and to model the time series characteristics of the vectorized feature data.

[0039] The Transformer feature analysis unit, connected to the optimized LSTM processing unit, receives the initial feature representation and generates a deep feature representation that integrates temporal and spatial information through a multi-head attention mechanism.

[0040] The fault identification unit, connected to the Transformer feature analysis unit, constructs a fault feature database based on historical fault data and deep feature representation and performs fault mode identification.

[0041] The operation and maintenance strategy decision unit is connected to the fault identification unit. It establishes an operation and maintenance strategy decision model based on the fault type and edge node operation and maintenance parameters and outputs the optimal operation and maintenance strategy.

[0042] The operation and maintenance execution drive unit is connected to the operation and maintenance strategy decision unit. It receives the optimal operation and maintenance strategy and sends it to the corresponding transmission and distribution line node operation and maintenance terminal to drive the operation and maintenance operation.

[0043] Beneficial Effects: This invention proposes a cloud computing-based operation and maintenance method and system for transmission and distribution line nodes. In data processing, it combines an optimized Long Short-Term Memory (LSTM) network with a Transformer model. The former utilizes an adaptive weight adjustment module and a forget gate control unit to accurately model time-series characteristics based on parameters such as voltage and current of transmission and distribution line nodes. The latter introduces parameters such as inter-node distance to improve the multi-head attention mechanism, deeply mining node spatial dependencies and parameter nonlinear relationships. Compared to traditional single algorithms, this approach can process data more comprehensively and accurately, significantly improving the timeliness and accuracy of fault diagnosis. In terms of operation and maintenance strategy formulation, it fully considers operation and maintenance parameters such as the geographical location of edge nodes, the accessibility of maintenance personnel, and spare parts inventory. It constructs an operation and maintenance strategy decision model by combining node load rate, failure rate, and insulation aging index. Through multi-formula comprehensive decision-making, it outputs the optimal operation and maintenance strategy, changing the situation where traditional solutions fail to consider actual operation and maintenance conditions and have poor strategy feasibility. This avoids resource waste caused by excessive operation and maintenance, and prevents untimely operation and maintenance from affecting line safety, achieving scientific and efficient operation and maintenance management, ensuring the safe and stable operation of transmission and distribution lines, and improving the overall efficiency and intelligence level of power grid operation and maintenance. Attached Figure Description

[0044] Figure 1 This is a flowchart of the method steps of the present invention;

[0045] Figure 2 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0046] It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. The following describes the application in further detail with reference to the accompanying drawings and specific embodiments.

[0047] like Figure 1 As shown, a cloud computing-based operation and maintenance method for power transmission and distribution line nodes includes the following steps:

[0048] S1: Through the cloud computing platform, feature extraction is performed on the real-time operating data collected from each node of the power transmission and distribution line. The real-time operating data includes node voltage amplitude, current effective value, active power, reactive power and ambient temperature parameters. The extracted multi-dimensional feature data is then vectorized.

[0049] Specifically, in the operation and maintenance scenario of power transmission and distribution lines, step S1 integrates a multi-source data acquisition system through a cloud computing platform to acquire electrical parameters such as node voltage amplitude and effective current value, as well as external factors such as ambient temperature, in real time. The acquisition frequency of these parameters is dynamically adjusted according to the importance of the line, with millisecond-level sampling implemented for key nodes of the main line and second-level sampling for nodes of branch lines. After acquisition, the data is processed by feature extraction algorithms to separate steady-state features, transient features, and trend features. For example, sub-features such as fluctuation frequency and amplitude change rate are extracted from voltage amplitude. In the vectorization processing stage, high-dimensional tensors are used to represent multi-dimensional feature data. Each dimension corresponds to a specific physical quantity, such as voltage dimension and current dimension. Through normalization operations, parameters of different dimensions are mapped to a unified feature space, forming structured data suitable for deep learning model processing.

[0050] This step establishes the data foundation for operation and maintenance (O&M) decisions. Through multi-parameter fusion and feature vectorization, it solves the data fragmentation problem in traditional O&M. For example, after vectorizing the correlation between ambient temperature and RMS current, it can effectively identify the decrease in line current carrying capacity caused by rising temperature, providing multi-dimensional feature support for subsequent fault prediction. The dimensionality of the vectorized feature data is determined according to the line complexity, generally ranging from 50 to 200 dimensions, preserving data integrity while avoiding the curse of dimensionality.

[0051] S2: Construct an optimized Long Short-Term Memory (LSTM) network structure. Introduce an adaptive weight adjustment module between the input layer and hidden layer of the LSM network. Input the vectorized feature data into the optimized LSM network. Use this network to model the time series characteristics of the power transmission and distribution line node operation data and obtain a preliminary feature representation containing time dimension context information.

[0052] Specifically, the optimized LSTM network constructed in step S2 incorporates two improvements tailored to the temporal characteristics of power transmission and distribution data. The adaptive weight adjustment module dynamically adjusts the weights of each input feature by analyzing the correlation between historical fault data and current parameters. For example, when a voltage amplitude fluctuation is detected to be highly correlated with historical short-circuit faults, the weight coefficient of that feature is automatically increased. The forget gate control unit selectively retains or discards historical information based on active power change trends and reactive power compensation requirements. During peak load periods, this unit enhances the memory of recent load change patterns, weakens the impact of off-peak data, and improves the model's response to short-term load fluctuations. This improvement enables the LSTM to accurately capture the time lag effects and periodic changes between parameters when processing power transmission and distribution time-series data. For example, the prediction accuracy for voltage sag events is approximately 15% higher than that of traditional LSTM, especially excelling in identifying precursors to intermittent faults. The network training employs a stochastic gradient descent algorithm with a batch size of 64. The number of iterations is dynamically adjusted between 100 and 500 rounds based on the amount of data, ensuring a balance between convergence and generalization ability.

[0053] S3: Input the preliminary feature representation into the Transformer model. Through the multi-head attention mechanism in the Transformer model, perform correlation analysis on the features of transmission and distribution line nodes from different dimensions and scales, capture the spatial dependencies between nodes and the nonlinear relationships between parameters, and generate a deep feature representation that integrates temporal and spatial information.

[0054] Specifically, step S3 introduces the Transformer model to address the spatial dependency problem between nodes in transmission and distribution lines. A multi-head attention mechanism analyzes feature correlations from different dimensions, such as simultaneously considering the differences in the mutual influence between nodes based on electrical and physical distance. The introduction of the node distance parameter enables the model to distinguish the influence weights of adjacent nodes and key hub nodes; in the analysis of 500kV transmission lines, this improvement increases fault location accuracy by approximately 20%. The fusion of environmental information such as temperature gradient and humidity change rate further enhances the model's ability to identify faults caused by external factors; for example, in high-temperature and high-humidity environments, it can provide a 3-5 hour advance warning of insulator flashover risk. The information fusion mechanism between the encoding and decoding layers employs dynamic weight allocation, adjusting the influence factors of environmental parameters according to seasonal changes. During the high-temperature period in summer, the temperature gradient weight automatically increases to 0.4 (normally 0.2), while the humidity change rate weight decreases to 0.1 (normally 0.3). Model training uses a masked language model and a next-sentence prediction task, mining the inherent patterns in the data through self-supervised learning, achieving high-precision feature extraction without requiring a large number of labeled samples.

[0055] S4: Based on the historical fault data and deep feature representation of transmission and distribution line nodes, construct a fault feature database, and use preset fault classification rules to perform fault mode recognition on the deep feature representation and classify fault types.

[0056] Specifically, the fault feature database constructed in step S4 integrates multi-dimensional feature information. During the dimensionality reduction stage, feature importance is adjusted by voltage fluctuation coefficient and current distortion rate, retaining features strongly correlated with faults. This compresses the original feature matrix to 30%-50% of its original dimensions, significantly improving subsequent recognition efficiency. For example, when processing a dataset containing 200 features, dimensionality reduction retains 60-100 key features, increasing recognition speed by approximately 40% and reducing the false alarm rate by 12%. Fault pattern recognition employs a hierarchical classification architecture. First, coarse classification is performed based on phase difference and impedance change rate, categorizing faults into broad classes such as short circuit, overload, and insulation aging. Then, precise classification is performed using fine-grained features.

[0057] This step established a complete identification system, from feature enhancement to pattern matching. The introduction of phase difference and impedance change rate improved the model's accuracy in identifying high-resistance grounding faults from 78% to 92%. The database employs an incremental learning mechanism, automatically updating classification rules as new fault cases emerge, ensuring the system's adaptability to novel faults and enabling the identification of approximately 10 new rare fault patterns annually.

[0058] S5: Based on the fault type and various parameters of edge node operation and maintenance, including node geographical location, accessibility of maintenance personnel, and spare parts inventory, an operation and maintenance strategy decision model is established, and the optimal operation and maintenance strategy for transmission and distribution line nodes with different fault types is output through the model.

[0059] Specifically, the operation and maintenance strategy decision model in step S5 comprehensively considers multi-dimensional constraints such as equipment status, operation and maintenance resources, and environmental conditions. The basic decision formula uses load rate, failure rate, and operation and maintenance cost as core indicators, and reflects the differences in the importance of lines through dynamic weight coefficients. For lines in urban core areas, the reliability weight coefficient γ1 is set to 0.5 (normally 0.3), prioritizing the protection of power supply continuity; for lines in remote areas, the cost weight coefficient γ3 is increased to 0.4 (normally 0.2). The improved formula further incorporates the insulation aging index and the service life of the equipment, forming a four-dimensional decision system, making the strategy more closely aligned with the needs of equipment lifecycle management.

[0060] The model employs a mixed-integer programming algorithm, generating the optimal strategy within 100ms. In the operation and maintenance of a 500kV substation, this model reduced unplanned power outage time by 40% and operation and maintenance costs by 18%. The strategy output includes maintenance priority ranking, spare parts allocation plans, and personnel scheduling suggestions, forming a complete operation and maintenance execution plan that directly guides on-site operations.

[0061] S6: Send the optimal operation and maintenance strategy to the corresponding transmission and distribution line node operation and maintenance terminal, and drive the operation and maintenance terminal to execute operation and maintenance operations.

[0062] Specifically, step S6 executes the strategy through collaboration between the cloud computing platform and edge computing terminals. The optimal strategy is sent to the maintenance terminal in the form of encrypted data packets, including the three-dimensional coordinates of the fault location, recommended repair solutions, and safe operation guidelines. Upon receiving the data, the terminal automatically parses the task and generates an execution interface that includes functions such as GPS navigation and AR-assisted maintenance. For example, when handling cable faults, the terminal can display a real-time map of the underground cable distribution and overlay repair points and operation steps using AR technology to guide maintenance personnel to quickly locate the fault. The system establishes a two-way feedback mechanism, with maintenance results transmitted back to the cloud computing platform in real time via the terminal. If a strategy deviation is detected during execution, such as a discrepancy between spare parts inventory and actual demand, the system automatically triggers a strategy correction process to re-optimize the decision-making model. This closed-loop structure enables the system to continuously optimize during operation, improving maintenance accuracy by approximately 1%-2% per month, forming a data-driven intelligent maintenance ecosystem.

[0063] Preferably, in the optimized Long Short-Term Memory network, the weight update of the adaptive weight adjustment module adopts the following formula:

[0064]

[0065] Among them, W t Let W be the weight matrix of the adaptive weight adjustment module at time t. t+1 Here is the weight matrix updated at time t+1; n represents the number of dimensions of the input features; α i,t f is the importance coefficient of the i-th feature at time t, which is calculated based on the correlation between the voltage amplitude and effective current value of the transmission and distribution line node and historical fault data; i,t Let be the value of the i-th feature at time t after processing by the activation function; σ is the activation function.

[0066] Specifically, in optimizing the adaptive weight adjustment module of the Long Short-Term Memory (LSTM) network, this module dynamically determines the importance coefficient of each input feature at the current moment based on the correlation between the voltage amplitude, effective current value, and historical fault data of the transmission and distribution line nodes. As operational data is continuously input, the module updates the weight matrix for different features, taking into full account the values ​​of each feature after activation function processing and its importance coefficient. This adaptive adjustment mechanism enables the LSM network to automatically highlight features closely related to potential faults and weaken the influence of interference information when processing operational data of transmission and distribution line nodes. This effectively improves the network's ability to capture key information in time-series data, laying a precise feature extraction foundation for subsequent fault analysis and prediction.

[0067] Preferably, in the calculation of the multi-head attention mechanism of the Transformer model, the distance parameter between transmission and distribution line nodes is introduced, and the formula for calculating the attention weight is as follows:

[0068]

[0069] Where Q, K, and V are the query matrix, key matrix, and value matrix, respectively; d k is the dimension of the key matrix; Dist is the distance matrix between nodes of the transmission and distribution line, and the element values ​​represent the physical distance between the corresponding nodes; β is the distance weight coefficient, which is dynamically adjusted according to the topology of the transmission and distribution line and the importance of the nodes; softmax is the normalized exponential function.

[0070] Specifically, the multi-head attention mechanism of the Transformer model is improved by introducing a distance parameter between transmission and distribution line nodes. During actual operation, the distance weight coefficient is dynamically adjusted based on the topology of the transmission and distribution line and the importance of the nodes. When calculating the attention weights, the physical distance information between nodes is fused with data such as the query matrix, key matrix, and value matrix. This allows the Transformer model to fully consider the spatial relationships between nodes when analyzing the characteristics of transmission and distribution line nodes, prioritizing the features of nodes that are spatially close and have a significant impact on the overall line operation. It accurately uncovers the spatial dependencies between nodes and the nonlinear relationships between various operating parameters, thereby generating a more spatially relevant and practically meaningful deep feature representation.

[0071] Preferably, during the construction of the fault feature database, the deep feature representation is dimensionality-reduced using a dimensionality-reduction formula based on the voltage fluctuation coefficient and current distortion rate of transmission and distribution line nodes:

[0072]

[0073] Where X is the original depth feature matrix, X new The feature matrix after dimensionality reduction; PCA is the principal component analysis method; diag represents the operation of generating a diagonal matrix; λ1 and λ2 are weight coefficients; VF i CD represents the voltage fluctuation coefficient at the i-th transmission and distribution line node, reflecting the degree of change in voltage amplitude over time. i Let be the current distortion rate of the i-th transmission and distribution line node, which measures the degree to which the current waveform deviates from a sine wave.

[0074] Specifically, the dimensionality reduction process in the construction of the fault feature database is carried out using a method based on the voltage fluctuation coefficient and current distortion rate of transmission and distribution line nodes. In implementation, the voltage fluctuation coefficient of each node is first calculated, reflecting the degree of voltage amplitude change over time. Then, the current distortion rate is calculated to measure the degree to which the current waveform deviates from a sine wave. Combining these two parameters, principal component analysis is used to reduce the dimensionality of the original deep feature matrix. While retaining the dimensions of features strongly correlated with faults, redundant information is removed, reducing data dimensionality and improving the storage efficiency of the fault feature database and the computational speed of subsequent fault mode recognition. This enables the system to quickly and accurately extract key fault features from massive amounts of data.

[0075] Preferably, in the operation and maintenance strategy decision model, considering the load rate (LR) and failure rate (FR) of the transmission and distribution line nodes, the optimal operation and maintenance strategy is determined using the following decision formula:

[0076]

[0077] Where OSM is the optimal operation and maintenance strategy; M is the set of all available operation and maintenance strategies; m represents a specific operation and maintenance strategy; LR m FR m These represent the load rate and failure rate of transmission and distribution line nodes when using operation and maintenance strategy m; Cost m γ1 represents the operation and maintenance cost of adopting operation and maintenance strategy m; γ2, γ3 are weighting coefficients set according to the importance of transmission and distribution lines and the allocation of operation and maintenance resources.

[0078] Specifically, the operation and maintenance (O&M) strategy decision-making model comprehensively considers factors such as the load rate and failure rate of transmission and distribution line nodes to determine the optimal O&M strategy. In practice, first, a set of all available O&M strategies is determined. For each specific O&M strategy, the load rate, failure rate, and O&M cost of the line nodes after adopting that strategy are calculated. Based on the importance of the transmission and distribution lines and the allocation of O&M resources, corresponding weighting coefficients are set. Through a specific calculation method, each O&M strategy is evaluated, and finally, the strategy that minimizes the overall evaluation value is selected as the optimal O&M strategy. This ensures that the O&M strategy meets the line operation requirements while rationally allocating O&M resources, achieving efficient O&M.

[0079] Preferably, in step S2, the optimized long short-term memory network further includes a forgetting gate control unit, and the output calculation formula of the forgetting gate is:

[0080] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0081] Among them, ft W is the output of the forget gate at time t. f h is the weight matrix of the forget gate; t-1 x is the output of the hidden layer at the previous time step; t b is the input feature vector at time t; f σ is the bias vector of the forget gate; σ is the activation function. The forget gate control unit controls the degree of retention of historical information based on the active power change trend and reactive power compensation requirements of the transmission and distribution line nodes.

[0082] Specifically, a forget gate control unit is added to the optimized Long Short-Term Memory (LSTM) network in step S2. This unit precisely controls the degree of retention of historical information in the network based on the active power variation trend and reactive power compensation requirements of the transmission and distribution line nodes. When calculating the forget gate output, the output of the hidden layer at the previous time step, the input feature vector at the current time step, and the weight matrix and bias vector of the forget gate are comprehensively considered. Through this mechanism, the LSM network can selectively retain historical information valuable for current fault prediction and analysis while forgetting irrelevant or interfering historical data, thus enhancing the network's adaptability to dynamic changes in the operating status of transmission and distribution lines and improving prediction accuracy.

[0083] Preferably, in step S3, an information fusion mechanism based on the real-time temperature gradient and humidity change rate of the transmission and distribution line nodes is introduced between the output of the encoding layer and the input of the decoding layer of the Transformer model. The fusion formula is as follows:

[0084] I fused =δ1·O enc +δ2·I dec +δ3·Env

[0085] Among them, I fused For the merged information; O enc For the output of the coding layer; I dec δ1, δ2, and δ3 are the inputs to the decoding layer; Env is an environmental information vector containing the real-time temperature gradient and humidity change rate of the transmission and distribution line nodes; δ1, δ2, and δ3 are fusion weight coefficients set according to the degree of influence of environmental factors on the operation of transmission and distribution lines.

[0086] Specifically, an information fusion mechanism based on real-time temperature gradients and humidity change rates at transmission and distribution line nodes is introduced between the encoding and decoding layers of the Transformer model. During actual operation, the temperature gradients and humidity change rates at the line nodes are monitored in real time, forming a vector containing environmental information. Based on the degree of influence of environmental factors on the operation of the transmission and distribution lines, corresponding fusion weight coefficients are set. The output of the encoding layer, the input of the decoding layer, and the environmental information vector are then calculated according to a specific fusion method to obtain the fused information. This mechanism allows the Transformer model to fully incorporate the influence of environmental factors when analyzing line node characteristics, avoiding missed or false fault detections caused by environmental changes, and improving the model's ability to identify line faults in complex operating environments.

[0087] Preferably, in step S4, during fault mode identification, a fault feature enhancement formula based on the phase difference and impedance change rate between transmission and distribution line nodes is used:

[0088]

[0089] Among them, F enhanced The enhanced fault feature vector; F original This represents the original fault feature vector; ΔZ represents the change in phase difference between nodes i and j of the power transmission and distribution line. ij τ is the impedance change rate between nodes i and j; τ is the adjustment coefficient, which is set according to the voltage level and line length of the transmission and distribution line; sigmoid is the activation function.

[0090] Specifically, in the fault mode identification process of step S4, a fault feature enhancement method based on the phase difference and impedance change rate between transmission and distribution line nodes is adopted. In implementation, the phase difference change and impedance change rate between line nodes are first calculated, and then the original fault feature vector is processed in conjunction with adjustment coefficients set according to the voltage level and length of the transmission and distribution lines. Through specific calculation methods, features closely related to the fault are enhanced, making the fault features more prominent. This helps to more clearly distinguish different types of faults during fault mode identification, improving the accuracy and reliability of fault diagnosis, especially the ability to identify faults with inconspicuous features.

[0091] Preferably, in step S5, the operation and maintenance strategy decision model also combines the insulation aging index IAI of the transmission and distribution line nodes and the service life SL of the equipment, and adopts an improved decision formula:

[0092]

[0093] Among them, OSM ' The improved optimal operation and maintenance strategy; IAI m SLm These represent the insulation aging index and equipment service life of transmission and distribution line nodes when the operation and maintenance strategy m is adopted; Time m The estimated maintenance time required to adopt maintenance strategy m; γ ' 1. γ ' 2. γ ' 3. γ ' 4 represents the weighting coefficients set according to the operational safety level and maintenance urgency of power transmission and distribution lines.

[0094] Specifically, the operation and maintenance strategy decision model in step S5 is further optimized by considering the insulation aging index and equipment service life of transmission and distribution line nodes. When determining the optimal operation and maintenance strategy, in addition to calculating the load rate, failure rate, and operation and maintenance cost, the insulation aging index and equipment service life of line nodes after adopting each operation and maintenance strategy are also calculated. New weighting coefficients are set according to the operational safety level and urgency of the transmission and distribution lines, and these factors are incorporated into the comprehensive evaluation formula. Through this improvement, the operation and maintenance strategy decision model can more comprehensively consider the actual condition of the line equipment, formulate operation and maintenance strategies that better meet the needs of the entire equipment lifecycle management, and ensure the long-term stable operation of the lines.

[0095] like Figure 2 As shown, a cloud-based power transmission and distribution line node operation and maintenance system includes:

[0096] The data acquisition and feature extraction unit is used to extract features and vectorize the real-time operating data collected from each node of the power transmission and distribution line through the cloud computing platform.

[0097] The LSTM processing unit is optimized and connected to the data acquisition and feature extraction unit to construct an optimized long short-term memory network structure, and to model the time series characteristics of the vectorized feature data.

[0098] The Transformer feature analysis unit, connected to the optimized LSTM processing unit, receives the initial feature representation and generates a deep feature representation that integrates temporal and spatial information through a multi-head attention mechanism.

[0099] The fault identification unit, connected to the Transformer feature analysis unit, constructs a fault feature database based on historical fault data and deep feature representation and performs fault mode identification.

[0100] The operation and maintenance strategy decision unit is connected to the fault identification unit. It establishes an operation and maintenance strategy decision model based on the fault type and edge node operation and maintenance parameters and outputs the optimal operation and maintenance strategy.

[0101] The operation and maintenance execution drive unit is connected to the operation and maintenance strategy decision unit. It receives the optimal operation and maintenance strategy and sends it to the corresponding transmission and distribution line node operation and maintenance terminal to drive the operation and maintenance operation.

[0102] This invention proposes a cloud computing-based operation and maintenance method and system for power transmission and distribution line nodes. Through technological integration and innovative design, it effectively overcomes the shortcomings of the background technology and possesses significant advantages. At the data processing level, traditional methods often employ single algorithms, making it difficult to fully mine data features. This solution constructs an architecture combining an optimized Long Short-Term Memory (LSTM) network with a Transformer model. The optimized LSM network, with its adaptive weight adjustment module and forget gate control unit, can accurately capture time-series characteristics based on parameters such as voltage and current of power transmission and distribution line nodes. The Transformer model introduces parameters such as inter-node distance to improve the multi-head attention mechanism, deeply mining node spatial dependencies and parameter nonlinear relationships. This solves the problems of incomplete and low-accuracy traditional data processing, making fault diagnosis more timely and accurate.

[0103] In terms of operation and maintenance strategy formulation, existing solutions do not fully consider the actual operation and maintenance conditions of edge nodes, resulting in poor strategy feasibility. This system's operation and maintenance strategy decision model fully incorporates operation and maintenance parameters such as the geographical location of edge nodes, the accessibility of maintenance personnel, and spare parts inventory. It also combines indicators such as node load rate, failure rate, and insulation aging index to construct decision formulas. Both the basic decision formula and the improved formula can set weight coefficients based on the importance of the line, the allocation of operation and maintenance resources, and the security level. Through calculation and evaluation, it outputs the optimal operation and maintenance strategy, avoiding over-maintenance or untimely maintenance. This ensures both rational resource utilization and the safe and stable operation of the line, significantly improving the scientific nature and feasibility of the operation and maintenance strategy.

[0104] From a system-wide perspective, the cloud computing platform enables end-to-end management of data acquisition, feature extraction, model computation, and strategy execution, with each unit working closely together. The data acquisition and feature extraction unit provides the foundation for subsequent analysis; the optimized LSTM processing unit and Transformer feature analysis unit perform in-depth data processing; the fault identification unit accurately classifies fault types; the operation and maintenance strategy decision-making unit generates the optimal strategy; and the operation and maintenance execution drive unit ensures the strategy is implemented. This integrated design, compared to traditional distributed operation and maintenance methods, forms an efficient and intelligent operation and maintenance closed loop, significantly improving the overall efficiency and intelligence level of transmission and distribution line node operation and maintenance.

[0105] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A cloud computing-based operation and maintenance method for power transmission and distribution line nodes, characterized in that, Includes the following steps: S1: Through the cloud computing platform, feature extraction is performed on the real-time operating data collected from each node of the power transmission and distribution line. The real-time operating data includes node voltage amplitude, current effective value, active power, reactive power and ambient temperature parameters. The extracted multi-dimensional feature data is then vectorized. S2: Construct an optimized Long Short-Term Memory (LSTM) network structure. Introduce an adaptive weight adjustment module between the input layer and hidden layer of the LSM network. Input the vectorized feature data into the optimized LSM network. Use this network to model the time series characteristics of the power transmission and distribution line node operation data and obtain a preliminary feature representation containing time dimension context information. S3: Input the preliminary feature representation into the Transformer model. Through the multi-head attention mechanism in the Transformer model, perform correlation analysis on the features of transmission and distribution line nodes from different dimensions and scales, capture the spatial dependencies between nodes and the nonlinear relationships between parameters, and generate a deep feature representation that integrates temporal and spatial information. S4: Based on the historical fault data and deep feature representation of transmission and distribution line nodes, construct a fault feature database, and use preset fault classification rules to perform fault mode recognition on the deep feature representation and classify fault types. S5: Based on the fault type and various parameters of edge node operation and maintenance, including node geographical location, accessibility of maintenance personnel, and spare parts inventory, an operation and maintenance strategy decision model is established, and the optimal operation and maintenance strategy for transmission and distribution line nodes with different fault types is output through the model. S6: Send the optimal operation and maintenance strategy to the corresponding transmission and distribution line node operation and maintenance terminal, and drive the operation and maintenance terminal to execute operation and maintenance operations.

2. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, The optimized Long Short-Term Memory network uses the following formula for weight updates in its adaptive weight adjustment module: Among them, W t Let W be the weight matrix of the adaptive weight adjustment module at time t. t+1 Here is the weight matrix updated at time t+1; n represents the number of dimensions of the input features; α i,t f is the importance coefficient of the i-th feature at time t, which is calculated based on the correlation between the voltage amplitude and effective current value of the transmission and distribution line node and historical fault data; i,t Let be the value of the i-th feature at time t after processing by the activation function; σ is the activation function.

3. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, The multi-head attention mechanism of the Transformer model incorporates the distance parameter between transmission and distribution line nodes, and the formula for calculating the attention weights is as follows: Where Q, K, and V are the query matrix, key matrix, and value matrix, respectively; d k is the dimension of the key matrix; Dist is the distance matrix between nodes of the transmission and distribution line, and the element values ​​represent the physical distance between the corresponding nodes; β is the distance weight coefficient, which is dynamically adjusted according to the topology of the transmission and distribution line and the importance of the nodes; softmax is the normalized exponential function.

4. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, The construction process of the fault feature database involves dimensionality reduction of the deep feature representation, using a dimensionality reduction formula based on the voltage fluctuation coefficient and current distortion rate of transmission and distribution line nodes: Where X is the original depth feature matrix, X new The feature matrix after dimensionality reduction; PCA is the principal component analysis method; diag represents the operation of generating a diagonal matrix; λ1 and λ2 are weight coefficients; VF i CD represents the voltage fluctuation coefficient at the i-th transmission and distribution line node, reflecting the degree of change in voltage amplitude over time. i Let be the current distortion rate of the i-th transmission and distribution line node, which measures the degree to which the current waveform deviates from a sine wave.

5. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, The operation and maintenance strategy decision model considers the load rate (LR) and failure rate (FR) of transmission and distribution line nodes, and uses the following decision formula to determine the optimal operation and maintenance strategy: Where OSM is the optimal operation and maintenance strategy; M is the set of all available operation and maintenance strategies; m represents a specific operation and maintenance strategy; LR m FR m These represent the load rate and failure rate of transmission and distribution line nodes when using operation and maintenance strategy m; Cost m γ1 represents the operation and maintenance cost of adopting operation and maintenance strategy m; γ2, γ3 are weighting coefficients set according to the importance of transmission and distribution lines and the allocation of operation and maintenance resources.

6. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, In step S2, the optimized Long Short-Term Memory (LSTM) network is further equipped with a forgetting gate control unit, and the output calculation formula of the forgetting gate is: f t =σ(W f ·[h t-1 ,x t ]+b f ) Among them, f t W is the output of the forget gate at time t. f h is the weight matrix of the forget gate; t-1 x is the output of the hidden layer at the previous time step; t Let b be the input feature vector at time t; f σ is the bias vector of the forget gate; σ is the activation function. The forget gate control unit controls the degree of retention of historical information based on the active power change trend and reactive power compensation requirements of the transmission and distribution line nodes.

7. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, In step S3, an information fusion mechanism based on the real-time temperature gradient and humidity change rate of power transmission and distribution line nodes is introduced between the output of the encoding layer and the input of the decoding layer of the Transformer model. The fusion formula is as follows: I fused =δ1·O enc +δ2·I dec +δ3·Env Among them, I fused For the merged information; O enc For the output of the coding layer; I dec δ1, δ2, and δ3 are the inputs to the decoding layer; Env is an environmental information vector containing the real-time temperature gradient and humidity change rate of the transmission and distribution line nodes; δ1, δ2, and δ3 are fusion weight coefficients set according to the degree of influence of environmental factors on the operation of transmission and distribution lines.

8. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, In step S4, during fault mode identification, a fault feature enhancement formula based on the phase difference and impedance change rate between transmission and distribution line nodes is used: Among them, F enhanced The enhanced fault feature vector; F original This represents the original fault feature vector; ΔZ represents the change in phase difference between nodes i and j of the power transmission and distribution line. ij τ is the impedance change rate between nodes i and j; τ is the adjustment coefficient, which is set according to the voltage level and line length of the transmission and distribution line; sigmoid is the activation function.

9. The cloud computing-based operation and maintenance method for power transmission and distribution line nodes according to claim 1, characterized in that, In step S5, the operation and maintenance strategy decision model also combines the insulation aging index IAI of transmission and distribution line nodes and the service life SL of equipment, and adopts an improved decision formula: Among them, OSM ' The improved optimal operation and maintenance strategy; IAI m SL m These represent the insulation aging index and equipment service life of transmission and distribution line nodes when the operation and maintenance strategy m is adopted; Time m The estimated maintenance time required to adopt maintenance strategy m; γ ' 1. γ ' 2. γ ' 3. γ ' 4 represents the weighting coefficients set according to the operational safety level and maintenance urgency of power transmission and distribution lines.

10. A cloud computing-based operation and maintenance system for power transmission and distribution line nodes, characterized in that, include: The data acquisition and feature extraction unit is used to extract features and vectorize the real-time operating data collected from each node of the power transmission and distribution line through the cloud computing platform. The LSTM processing unit is optimized and connected to the data acquisition and feature extraction unit to construct an optimized long short-term memory network structure, and to model the time series characteristics of the vectorized feature data. The Transformer feature analysis unit, connected to the optimized LSTM processing unit, receives the initial feature representation and generates a deep feature representation that integrates temporal and spatial information through a multi-head attention mechanism. The fault identification unit, connected to the Transformer feature analysis unit, constructs a fault feature database based on historical fault data and deep feature representation and performs fault mode identification. The operation and maintenance strategy decision unit is connected to the fault identification unit. It establishes an operation and maintenance strategy decision model based on the fault type and edge node operation and maintenance parameters and outputs the optimal operation and maintenance strategy. The operation and maintenance execution drive unit is connected to the operation and maintenance strategy decision unit. It receives the optimal operation and maintenance strategy and sends it to the corresponding transmission and distribution line node operation and maintenance terminal to drive the operation and maintenance operation.

Images