Reliability analysis method and system for power system

Abstract

The invention discloses a reliability analysis method and system for a power system, and relates to the technical field of intelligent operation and maintenance of the power system, and the method comprises the steps: collecting data, and constructing a dynamic space-time atlas network of the power system, which comprises time dimension information and space dimension information; feature extraction is carried out based on a spatial-temporal feature learning framework constructed by a dual-channel graph attention network and a gated causal convolution network, and spatial-temporal fusion features containing complete spatial-temporal information are generated by using gated causal convolution and a time attention capture time sequence evolution mode; and converting the space-time fusion features into a local risk state and a global system state through a state feature mapping network, calculating a local component risk index and a system overall reliability index, generating a collaborative decision strategy, and realizing reliability analysis of the power system. According to the method, the intelligent operation and maintenance level of the power system is remarkably improved, the shutdown time and the maintenance cost of the system are reduced, and meanwhile, the system performance is continuously improved through preventive maintenance and an active optimization strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent operation and maintenance of power systems, and in particular to a reliability analysis method and system for power systems. Background Art

[0002] As the scale of the power system continues to expand and the proportion of new energy grid connection continues to increase, the system structure is becoming increasingly complex, and various types of faults and disturbance events occur frequently, posing severe challenges to the safe and stable operation of the system. Especially in the context of large-scale new energy access, the uncertainty of the system has increased significantly, and traditional static analysis methods are difficult to accurately characterize the dynamic characteristics of the system. At the same time, there are complex physical coupling relationships between power system components, and a single fault may evolve into a large-scale fault through multiple propagation paths, causing serious threats to system stability. Traditional power system fault analysis and processing methods mainly rely on expert experience and preset rules, which are difficult to cope with the fault evolution laws in complex dynamic scenarios, and lack the ability to perceive the overall state of the system globally. This experience-based method often shows great limitations when facing new fault modes, and is unable to identify potential risks in a timely manner and take effective preventive measures.

[0003] Although existing methods have made some progress in fault diagnosis and decision support, they generally have problems such as insufficient feature extraction, shallow spatiotemporal correlation mining, and separation of risk assessment and decision optimization. These methods often separate fault diagnosis, risk assessment, and decision optimization, lack systematic overall consideration, and are difficult to achieve intelligent fault handling throughout the entire process. Especially in the face of new fault modes and complex disturbance scenarios, existing methods are difficult to effectively capture the propagation characteristics and evolution laws of faults, and cannot achieve accurate assessment and timely intervention of system risks. In addition, existing methods have low computational efficiency when processing high-dimensional and multi-scale power system data, and are difficult to meet the needs of real-time monitoring and rapid decision-making. In the context of the intelligent transformation of power grids, it is urgent to develop new intelligent methods to improve the system's fault diagnosis capabilities, risk assessment accuracy, and decision optimization efficiency. Summary of the invention

[0004] In view of the problems existing in the existing reliability analysis method of the power system, the present invention proposes a reliability analysis method and system for the power system.

[0005] Therefore, the problem to be solved by the present invention is how to solve the technical problems of the prior art such as insufficient fault feature extraction, shallow spatiotemporal correlation mining, and separation of risk assessment and decision optimization.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides a reliability analysis method for an electric power system, which includes collecting data to construct a dynamic spatiotemporal graph network of the electric power system, wherein the spatiotemporal graph network includes time dimension information of component operating status, maintenance records, and fault history, as well as spatial dimension information of physical connection, electrical association, and geographic location; extracting features from the spatiotemporal graph network based on a spatiotemporal feature learning framework constructed based on a dual-channel graph attention network and a gated causal convolutional network, extracting spatial domain features through adaptive fusion of physical association and fault features, capturing time series evolution patterns through gated causal convolution and time attention, and generating spatiotemporal fusion features containing complete spatiotemporal information; converting the spatiotemporal fusion features into local risk states and global system states through a state feature mapping network, calculating local component risk indicators and overall system reliability indicators, generating collaborative decision-making strategies through a risk weight adaptive fusion mechanism, and realizing reliability analysis of the electric power system.

[0007] As a preferred solution of the reliability analysis method of the power system described in the present invention, the power system operation data is collected, including component operation parameters, historical fault data and geographical location information, and the collected data is stored in a spatiotemporal database; wherein the component operation parameters include voltage values , Current value , Active Power , reactive power And power factor ; The historical fault data includes the time of fault occurrence, fault type, fault duration and fault recovery time; The geographic location information includes the longitude, latitude and altitude coordinates of the component.

[0008] As a preferred solution of the reliability analysis method of the power system of the present invention, wherein: based on the component operation parameters, a component node is established, and based on the component node, a component association matrix is ​​generated. , the incidence matrix The rows and columns correspond to different component nodes, and the matrix elements Characterization Components With components The physical connection strength between them is expressed as: , in, , , is the weight coefficient, and Components and The active power, and Components and The reactive power, and are the maximum values ​​of active power and reactive power in the system, respectively. and Components and The power factor of adjacent components is evaluated by comparing the power parameter differences between them. Based on the historical fault data, it is divided into multiple time windows according to the time series , the length of each time window is , calculate the fault feature vector of the component node in each time window , expressed as: , , , , , in, is the fault frequency, i.e. the number of faults occurring per unit time, is the total number of failures, is the length of the time window, is the mean time to repair, For the The repair time of each failure, is the fault duration, For the The duration of the fault, is the fault impact range index, For the The number of components affected by each failure, is the fault type weight coefficient; through the fault feature vector Reflects the reliability status of components; Calculate the spatial distance matrix between component nodes based on the geographic location information , the spatial distance matrix Component association matrix Perform weighted fusion to obtain the spatiotemporal correlation tensor , the fusion formula is expressed as: , , in, Representation Node and nodes The strength of the spatiotemporal correlation between and is the fusion weight, is the distance characteristic scale; Representation Node and nodes The correlation coefficient between For components and The Euclidean distance between and Components and The three-dimensional coordinates of .

[0009] As a preferred solution of the reliability analysis method of the power system of the present invention, the time-space correlation tensor is subjected to characteristic decomposition to extract the previous The principal eigenvectors are expressed as: , , in, represents the extracted main eigenvector matrix, Indicates feature vectors, is the number of retained eigenvectors, represents the total number of eigenvalues, Indicates eigenvalues, Represents the preset cumulative contribution rate threshold; the main eigenvector and the fault feature vector Combine to form a graph feature matrix , expressed as: , in, Represents the horizontal concatenation operation of the matrix, the graph feature matrix Contains complete information about spatiotemporal correlation features and fault features; uses a sliding time window mechanism to map the feature matrix Update by sliding the time window step Control the frequency of graph update, the updated graph feature matrix It is expressed as: , in, Indicates the current time The global state representation vector is is the smoothing factor, the value range is [0,1], is the time interval, The feature matrix computed for the new time window.

[0010] As a preferred solution of the reliability analysis method of the power system described in the present invention, wherein: a dual-channel graph attention layer is constructed based on the spatiotemporal correlation features and fault features for feature extraction; the first channel is the spatiotemporal correlation channel to process the spatiotemporal correlation tensor The second channel is the fault feature channel to process the fault feature vector The attention coefficient of the spatiotemporal correlation channel is expressed as: , in, represents the attention coefficient of the spatiotemporal correlation channel, Indicates the current center node The spatiotemporal correlation characteristics of Neighbor node The spatiotemporal correlation characteristics of is the feature transformation matrix; Representation Node The set of neighboring nodes; represents the rectified linear activation function with leakage, Representation Node All neighbor nodes of The attention coefficient of the fault feature channel is expressed as: , in, represents the attention coefficient of the fault feature channel, Current central node The fault feature vector is Neighbor node The fault feature vector of is the feature transformation matrix, Representation Node All neighbor nodes of The fault feature vector of is obtained; the importance of the two types of information is dynamically adjusted through the learnable fusion coefficient, which is expressed as: , in, and is the fusion coefficient of adaptive learning, represents the Sigmoid activation function, is the final combined attention weight, representing the node For Node Based on the combined attention weight, the node features are weighted and aggregated to generate enhanced features that consider spatial correlation, which is expressed as: , in, is the transformation matrix of spatiotemporal correlation features, is the transformation matrix of fault characteristics, is the spatial domain enhanced feature matrix, is the spatiotemporal correlation feature vector, representing the node The spatiotemporal correlation characteristics of is the fault feature vector, representing the node The fault related information.

[0011] As a preferred solution of the reliability analysis method of the power system described in the present invention, based on the enhanced features, a causal convolution layer based on a gating mechanism is constructed within the time window determined by the graph feature matrix to obtain the time series features, which is expressed as: , The gate control unit calculation is expressed as: , in, is the causal convolution kernel size, is the gating unit parameter, For the The convolution kernel parameters at each position, is the weight matrix of the gating unit, is the bias term of the gating unit, Indicates the current time The state feature vector of is the time step The state feature vector of is the time step Based on the temporal convolutional features, the time domain attention score is calculated to identify the key temporal patterns, which is expressed as: , in, is the temporal attention parameter matrix, represents the transposed matrix of the temporal attention parameter matrix, is the time window length The length of the time series within Indicates that at time step Convolutional features when ; is the time step The time domain attention weight is weighted and combined with the temporal convolution feature to generate the spatiotemporal fusion feature, which is expressed as: , in, The spatiotemporal fusion feature representation of the system state contains complete spatiotemporal information. Represents the time step The attention weight, Represents the time step Temporal convolution features.

[0012] As a preferred solution of the reliability analysis method of the power system of the present invention, the time-space fusion feature is double-mapped to form a double state representation, which is a risk state representation vector for the local component state and a global state representation vector describing the overall operation status of the system, expressed as: , , in, represents the risk state representation vector, describing the local component state, Represents the global state representation vector, describing the overall state of the system, represents the mapping weight matrix, represents the bias vector; Based on the dual state representation, the risk index of the local component and the reliability index of the system as a whole are calculated respectively, and the state information is converted into a quantifiable risk metric through nonlinear transformation, which is expressed as: , , in, Represents the local component risk index vector, reflecting the risk level of each component, Represents the global system reliability index, characterizing the overall reliability status of the system. is the activation function, To evaluate the weight matrix, Represents the bias vector.

[0013] As a preferred solution of the reliability analysis method of the power system described in the present invention, a two-layer strategy network architecture is constructed based on the risk index and the reliability index; the two-layer strategy network architecture includes a local intervention strategy network and a global coordination strategy network, and the intervention decisions of corresponding levels are generated based on the risk state representation and the global state representation, respectively, and are expressed as: , , in, represents the local policy network, generating component-level control strategies, represents the global policy network, generating the system-level coordination policy, represents the network parameters, Represents a local decision network based on local node features Generate local control strategies; Represents the global decision network, based on the global state representation Generate a global control strategy; through adaptive weight calculation, dynamically integrate the local intervention strategy and the global coordination strategy to form the final collaborative decision-making strategy, which is expressed as: , , in, is the collaborative weight vector, reflecting the importance of strategies at different levels, represents the fusion weight matrix, For the final decision strategy, represents the weight coefficient of the local control strategy, Represents the weight coefficient of the global control strategy.

[0014] In the second aspect, an embodiment of the present invention provides a reliability analysis system for an electric power system, which includes: an acquisition module, which is used to collect data to construct a dynamic spatiotemporal graph network of the electric power system, including time dimension information of component operating status, maintenance records, and fault history, as well as spatial dimension information of physical connection, electrical association, and geographic location; a feature extraction module, which extracts features from the spatiotemporal graph network based on a spatiotemporal feature learning framework constructed based on a dual-channel graph attention network and a gated causal convolutional network, realizes spatial domain feature extraction through adaptive fusion of physical association and fault features, captures time series evolution patterns using gated causal convolution and time attention, and generates spatiotemporal fusion features containing complete spatiotemporal information; an analysis module, which converts the spatiotemporal fusion features into local risk states and global system states through a state feature mapping network, calculates local component risk indicators and overall system reliability indicators, generates collaborative decision-making strategies through a risk weight adaptive fusion mechanism, and realizes reliability analysis of the electric power system.

[0015] The beneficial effect of the present invention is that the present invention realizes a full-process intelligent solution from data perception, feature extraction to risk assessment and decision optimization by constructing a system architecture that combines multi-scale deep learning and reinforcement learning. It can not only detect and respond to single component failures in a timely manner, but also effectively prevent and handle multi-component coupling failures and cascading failures, greatly improving the reliability and stability of the system. It significantly improves the intelligent operation and maintenance level of the power system, reduces system downtime and maintenance costs, and achieves continuous improvement of system performance through preventive maintenance and active optimization strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. 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 work, among which: Figure 1 Scenario diagram for reliability analysis method of power system. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0020] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and which provides a reliability analysis method for a power system, comprising: S1: Collect data to build a dynamic spatiotemporal graph network of the power system, which contains the time dimension information of component operating status, maintenance records, and fault history, as well as the spatial dimension information of physical connection, electrical connection, and geographic location, and continuously optimizes the graph structure through a real-time data update mechanism.

[0021] Collect power system operation data, including component operation parameters, historical fault data and geographic location information, and store the collected data in a spatiotemporal database; wherein the component operation parameters include voltage values , Current value , Active Power , reactive power And power factor ; The historical fault data includes the time of fault occurrence, fault type, fault duration and fault recovery time; The geographic location information includes the longitude, latitude and altitude coordinates of the component.

[0022] Establish component nodes according to component operation parameters, and generate component association matrix based on component nodes , the incidence matrix The rows and columns correspond to different component nodes, and the matrix elements Characterization Components With components The physical connection strength between them is expressed as: , in, , , is the weight coefficient, and Components and The active power, and Components and The reactive power, and are the maximum values ​​of active power and reactive power in the system, respectively. and Components and The power factor of adjacent components is evaluated by comparing the power parameter differences between them. Divide historical fault data into multiple time windows according to time series , the length of each time window is , calculate the fault feature vector of the component node in each time window , fault feature vector It is expressed as: , , , , , in, is the fault frequency, i.e. the number of faults occurring per unit time, is the total number of failures, is the length of the time window, is the mean time to repair, For the The time to repair a fault, is the fault duration, i.e. the cumulative value of the duration of all faults. For the The duration of the fault, is the fault impact range index, For the The number of components affected by each failure, is the fault type weight coefficient; through the fault feature vector Reflects the reliability status of the component.

[0023] Calculate the spatial distance matrix between component nodes based on geographic location information , the spatial distance matrix Component association matrix Perform weighted fusion to obtain the spatiotemporal correlation tensor , the fusion formula is expressed as: , , in, Representation Node and nodes The strength of the spatiotemporal correlation between and is the fusion weight, and + = 1, The distance characteristic scale is used to normalize the effect of spatial distance; Representation Node and nodes The correlation coefficient between For components and The Euclidean distance between them is calculated by longitude and latitude coordinates; and Components and The three-dimensional coordinates of . Space-time correlation tensor Comprehensively consider the physical connection relationship and geographic spatial relationship between components.

[0024] Space-time correlation tensor Perform feature decomposition before extraction principal eigenvectors, where The cumulative contribution rate of the characteristic value is greater than the preset threshold Determine, expressed as: , , in, represents the extracted main eigenvector matrix, Indicates feature vectors, is the number of retained eigenvectors, Indicates eigenvalues, sorted from largest to smallest; represents the total number of eigenvalues, The cumulative contribution rate of the characteristic value is greater than the preset threshold Determine the main eigenvector and the fault feature vector Combine to form a graph feature matrix , expressed as: , in, Represents the horizontal concatenation operation of the matrix, the graph feature matrix Contains complete information on temporal and spatial correlation characteristics and fault characteristics.

[0025] The sliding time window mechanism is used to map the feature matrix Update by sliding the time window step Control the frequency of graph update, the updated graph feature matrix It is expressed as: , in, Indicates the current time The global state representation vector is is the smoothing factor, the value range is [0,1], is the time interval, The feature matrix computed for the new time window.

[0026] By smoothing factor Adjust the weight ratio of historical information and new information, the larger The smaller the value, the more emphasis is placed on historical stability. The value emphasizes real-time performance.

[0027] S2: A spatiotemporal feature learning framework based on a dual-channel graph attention network and a gated causal convolutional network is used to extract features from the spatiotemporal graph network. The spatial domain feature extraction is achieved through the adaptive fusion of physical association and fault features. The gated causal convolution and temporal attention are used to capture the temporal evolution pattern and generate spatiotemporal fusion features containing complete spatiotemporal information.

[0028] Based on the generated graph feature matrix , construct a dual-channel graph attention layer. Considering the heterogeneity of physical correlation and fault characteristics in power systems, two independent attention channels are designed respectively: the first channel is dedicated to processing the spatiotemporal correlation tensor The second channel processes the fault feature vector Fault information in.

[0029] The attention coefficient of the spatiotemporal correlation channel is expressed as: , in, represents the attention coefficient of the spatiotemporal correlation channel, Indicates the current center node The spatiotemporal correlation characteristics of Neighbor node The spatiotemporal correlation characteristics of Representation Node All neighbor nodes of The spatiotemporal correlation characteristics of the , including physical correlation information such as topological structure, electrical distance, etc.; is the feature transformation matrix, which is used to map the input features to the attention space; Representation Node The set of neighboring nodes; Represents a leaky rectified linear activation function to avoid the vanishing gradient problem.

[0030] The attention coefficient of the fault feature channel is expressed as: , in, represents the attention coefficient of the fault feature channel, Current central node The fault feature vector is Neighbor node The fault feature vector is Representation Node All neighbor nodes of The fault feature vector contains fault information such as fault type and severity; is the feature transformation matrix, which is used to map the input features to the attention space.

[0031] In order to make full use of the complementary information of the two channels, the importance of the two types of information is dynamically adjusted through the learnable fusion coefficient, which is expressed as: , in, and is the fusion coefficient of adaptive learning, initialized to equal weights, and dynamically adjusted during the training process. Represents the Sigmoid activation function, which compresses the fusion result to the (0,1) interval to ensure the validity of the weight. is the final combined attention weight, representing the node For Node the overall level of attention.

[0032] Using the fused combined attention weights, we perform weighted aggregation operations on the node features to generate enhanced features that take spatial correlation into account, expressed as: , in, is the transformation matrix of spatiotemporal correlation features, which is used to adjust the dimension and importance of physical features. is the transformation matrix of the fault feature, which is used to adjust the dimension and importance of the fault feature. It is a spatial domain enhanced feature matrix, which contains the association information between nodes. is the spatiotemporal correlation feature vector, representing the node The spatiotemporal correlation characteristics of is the fault feature vector, representing the node The fault related information.

[0033] The features of neighborhood nodes are weighted and combined to highlight the contribution of important nodes. The expressions of spatiotemporal correlation features and fault features are adjusted separately through learnable transformation matrices to achieve optimal mapping of feature space. The weighted features and transformed features are combined to generate a node representation containing complete spatial correlation information, effectively extracting and fusing spatial domain information, and providing high-quality feature representation for subsequent analysis tasks.

[0034] Furthermore, in order to capture the temporal pattern of fault evolution, based on the enhanced feature matrix In the graph feature matrix Within a certain time window, a causal convolution layer based on a gating mechanism is constructed to obtain temporal features, which can be expressed as: , The gate control unit calculation is expressed as: , in, is the size of the causal convolution kernel, which determines the time span for considering historical information. is the gating unit parameter, which controls the importance of information at different time steps. For the The convolution kernel parameters at each position are used for feature conversion. is the weight matrix of the gating unit, is the bias term of the gating unit, Indicates the current time The state feature vector of is the time step The state feature vector of is the time step Temporal convolution features.

[0035] Causal convolution is used to ensure the unidirectionality of time series information, the transmission of historical information is controlled through the gating mechanism, the importance of features at different time steps is dynamically adjusted, and time series convolution features are obtained.

[0036] Based on the temporal convolutional features, the temporal domain attention score is calculated to identify the key temporal patterns, which is expressed as: , in, is the temporal attention parameter matrix, used to learn the importance of the temporal dimension, represents the transposed matrix of the temporal attention parameter matrix, is the time window length The time series length within is consistent with the sliding window setting of the graph feature matrix; Indicates that at time step Convolutional features when ; is the time step The attention weight indicates the importance of the moment.

[0037] By calculating the feature importance of each time step, softmax normalization is used to obtain the attention distribution in the time dimension, thereby highlighting the contribution of key timing patterns.

[0038] The time domain attention weights are weighted and combined with the temporal convolution features to generate the final spatiotemporal fusion features, which are expressed as: , in, The spatiotemporal fusion feature representation of the system state contains complete spatiotemporal information. Represents the time step The attention weight, Represents the time step Temporal convolution features.

[0039] Through multi-level feature extraction and fusion mechanism, deep mining of spatiotemporal characteristics of power system is achieved, providing high-quality feature representation for subsequent reliability analysis.

[0040] S3: The spatiotemporal fusion features are converted into local risk states and global system states through the state feature mapping network, the local component risk indicators and the overall system reliability indicators are calculated, and the collaborative decision-making strategy is generated through the risk weight adaptive fusion mechanism to realize the reliability analysis of the power system.

[0041] The acquired spatiotemporal fusion features are double mapped, and two state representations at different semantic levels are constructed through the feature mapping network to generate risk state representation vectors for local component states respectively. and the global state representation vector that describes the overall operating status of the system , the transformation of the mapping process is expressed as: , , in, represents the risk state representation vector, describing the local component state, Represents the global state representation vector, describing the overall state of the system, represents the mapping weight matrix, Represents the bias vector.

[0042] Through the constructed dual state representation, the risk index of the local component and the reliability index of the system as a whole are calculated respectively, and the state information is converted into a quantifiable risk measure through nonlinear transformation, which is expressed as: , , in, Represents the local component risk index vector, reflecting the risk level of each component, Represents the global system reliability index, characterizing the overall reliability status of the system. is the activation function, To evaluate the weight matrix, Represents the bias vector.

[0043] According to the risk index and reliability index, a two-layer strategy network architecture is constructed. This architecture includes a local intervention strategy network and a global coordination strategy network, which generate intervention decisions at corresponding levels based on the risk state representation and the global state representation, respectively, and are expressed as: , , in, represents the local policy network, generating component-level control strategies, represents the global policy network, generating the system-level coordination policy, represents the network parameters, Represents a local decision network based on local node features Generate local control strategies; Represents the global decision network, based on the global state representation Generate a global control strategy.

[0044] Through adaptive weight calculation, the local intervention strategy and the global coordination strategy are dynamically integrated to form the final collaborative decision-making strategy, which is expressed as: , , in, is the collaborative weight vector, reflecting the importance of strategies at different levels, represents the fusion weight matrix, For the final decision strategy, represents the weight coefficient of the local control strategy, Represents the weight coefficient of the global control strategy.

[0045] Through a risk-oriented hierarchical decision-making mechanism, the extracted spatiotemporal features are effectively mapped to the risk assessment and intervention strategy space, achieving collaborative decision-making of local risk control and global reliability optimization.

[0046] Furthermore, based on the obtained collaborative decision-making strategy, different fault scenarios are handled separately, including: If the local risk index A component in If the risk value exceeds the warning threshold, the local intervention strategy of the component is activated, the component operating parameters are adjusted to a safe range, the intervention effect is recorded, and the risk assessment model is updated; If the local risk indicators of multiple related components increase at the same time and there is a significant correlation, the global coordination strategy will be triggered to calculate the load redistribution plan between components, coordinate the operating parameters of each component, establish a complementary mechanism between components, and monitor the overall stability of the system.

[0047] If a potential fault propagation path is detected, the global reliability index If it is below the safety threshold, the system-level prevention strategy is activated to isolate high-risk areas, activate backup resources, and reconstruct the system operation mode.

[0048] If the system reliability index does not reach the set target, identify the weak links, strengthen the protection of key nodes, optimize the backup mechanism, and improve the monitoring system.

[0049] In summary, the present invention realizes a full-process intelligent solution from data perception, feature extraction to risk assessment and decision optimization by constructing a system architecture that combines multi-scale deep learning with reinforcement learning. It can not only detect and respond to single component failures in a timely manner, but also effectively prevent and handle multi-component coupling failures and cascading failures, greatly improving the reliability and stability of the system. It significantly improves the intelligent operation and maintenance level of the power system, reduces system downtime and maintenance costs, and achieves continuous improvement of system performance through preventive maintenance and active optimization strategies. It not only improves the reliability and safety of the system, but also significantly reduces the operation and maintenance costs through intelligent decision-making, which has important engineering application value and promotion significance.

[0050] This embodiment further provides a reliability analysis system for a power system, including: The acquisition module is used to collect data to build a dynamic spatiotemporal graph network of the power system, including time dimension information of component operation status, maintenance records, and fault history, as well as spatial dimension information of physical connection, electrical association, and geographic location; The feature extraction module extracts features from the spatiotemporal graph network based on a spatiotemporal feature learning framework constructed based on a dual-channel graph attention network and a gated causal convolutional network, realizes spatial domain feature extraction through adaptive fusion of physical association and fault features, and uses gated causal convolution and temporal attention to capture the temporal evolution pattern, generating spatiotemporal fusion features containing complete spatiotemporal information; The analysis module converts the spatiotemporal fusion features into local risk states and global system states through a state feature mapping network, calculates local component risk indicators and system overall reliability indicators, generates a collaborative decision-making strategy through a risk weight adaptive fusion mechanism, and realizes reliability analysis of the power system.

[0051] This embodiment also provides a computer device, which is suitable for the reliability analysis method of the power system, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to implement the reliability analysis method of the power system proposed in the above embodiment.

[0052] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0053] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing reliability analysis of a power system as proposed in the above embodiment is implemented.

[0054] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0055] Embodiment 2: This embodiment provides a reliability analysis method for a power system. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0056] Real fault case data are extracted from the historical operation database of a provincial power dispatching and control center from January 2022 to December 2023, including equipment status quantities, electrical quantities and fault recording data.

[0057] The dynamic model of the IEEE 118-node system is constructed. Based on the characteristics of actual fault cases, various typical scenarios including single faults and cascading faults are designed. The Monte Carlo method is used to randomly inject faults of different types and locations to generate 2000 groups of fault samples.

[0058] The comparative experiments include: traditional rule-based method, CNN method, LSTM method, and the proposed method.

[0059] The reliability of the system is comprehensively evaluated from the perspectives of fault identification accuracy, fault location time, risk assessment accuracy, and decision response time. The sampling frequency is set to 100 Hz, and each set of samples contains 10 seconds of time series data before, during, and after the fault.

[0060] The experimental results are shown in Table 1: Table 1 Comparative experimental results Evaluation Metrics Rule-based CNN LSTM Proposed Fault identification accuracy (%) 82.5 88.7 90.3 95.8 Fault location time (ms) 156 89 75 42 Risk assessment accuracy (%) 78.3 85.4 87.6 93.2 Decision response time (ms) 235 145 128 85 Memory usage (MB) 45 186 213 167 Computational complexity (GFLOPs) 0.8 3.2 3.8 2.9 It can be seen from Table 1 that the method of the present invention shows significant performance advantages in the entire process of fault handling. In terms of fault identification accuracy, the method of the present invention achieves a high accuracy of 95.8%, which is 13.3 percentage points higher than the traditional rule-based method and 5.5 percentage points higher than the optimal LSTM method. This is due to the dual-channel graph attention network's ability to simultaneously capture the spatiotemporal coupling patterns of physical associations and fault characteristics.

[0061] In terms of fault location time, the method of the present invention only takes 42ms to complete fault location, which is 73% shorter than the 156ms of the traditional method. In particular, in terms of risk assessment accuracy, the method of the present invention achieves an accuracy rate of 93.2%, which is significantly better than other comparison methods. At the same time, the decision response time of the present invention only takes 85ms, which is 63.8% faster than the 235ms of the traditional method, while the computing resource consumption remains at a reasonable level, showing good engineering practicality.

[0062] These performance improvements are mainly due to the spatiotemporal graph analysis framework and risk-aware collaborative decision-making mechanism proposed in this invention, which together achieve rapid identification, accurate assessment and timely intervention of power system faults, and provide reliable technical support for the intelligent operation and maintenance of power systems. Through large-scale fault scenario testing on the IEEE 118 node system, the advancement and practical value of the method of this invention in the operation and maintenance of complex power systems have been fully verified.

[0063] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A reliability analysis method for a power system, characterized in that: include: Collect data to build a dynamic spatiotemporal graph network of the power system, which contains time dimension information of component operation status, maintenance records, and fault history, as well as spatial dimension information of physical connection, electrical association, and geographic location; A spatiotemporal feature learning framework based on a dual-channel graph attention network and a gated causal convolutional network is used to extract features from the spatiotemporal graph network, and spatial domain feature extraction is achieved through adaptive fusion of physical association and fault features. Gated causal convolution and temporal attention are used to capture the temporal evolution pattern, generating spatiotemporal fusion features containing complete spatiotemporal information. The spatiotemporal fusion features are converted into local risk states and global system states through a state feature mapping network, local component risk indicators and overall system reliability indicators are calculated, and a collaborative decision-making strategy is generated through a risk weight adaptive fusion mechanism to achieve reliability analysis of the power system.

2. The reliability analysis method of the power system according to claim 1, characterized in that: Collect power system operation data, including component operation parameters, historical fault data and geographic location information, and store the collected data in a spatiotemporal database; The component operating parameters include voltage values , Current value , Active Power , reactive power And power factor ; The historical fault data includes the time of fault occurrence, fault type, fault duration and fault recovery time; The geographic location information includes the longitude, latitude and altitude coordinates of the component.

3. The reliability analysis method of the power system according to claim 2, characterized in that: Establish component nodes based on the component operation parameters, and generate component association matrices based on the component nodes , the incidence matrix The rows and columns correspond to different component nodes, and the matrix elements Characterization Components With components The physical connection strength between them is expressed as: ， in, , , is the weight coefficient, and Components and The active power, and Components and The reactive power, and are the maximum values ​​of active power and reactive power in the system, respectively. and Components and The power factor of adjacent components is evaluated by comparing the power parameter differences between them. Based on the historical fault data, it is divided into multiple time windows according to the time series , the length of each time window is , calculate the fault feature vector of the component node in each time window , expressed as: ， ， ， ， ， in, is the fault frequency, i.e. the number of faults occurring per unit time, is the total number of failures, is the length of the time window, is the mean time to repair, For the The repair time of each failure, is the fault duration, For the The duration of the fault, is the fault impact range index, For the The number of components affected by each failure, is the fault type weight coefficient; through the fault feature vector Reflects the reliability status of components; Calculate the spatial distance matrix between component nodes based on the geographic location information , the spatial distance matrix Component association matrix Perform weighted fusion to obtain the spatiotemporal correlation tensor , the fusion formula is expressed as: ， ， in, Representation Node and nodes The strength of the spatiotemporal correlation between and is the fusion weight, is the distance characteristic scale; Representation Node and nodes The correlation coefficient between For components and The Euclidean distance between and Components and The three-dimensional coordinates of .

4. The reliability analysis method of the power system according to claim 3, characterized in that: The spatiotemporal correlation tensor is subjected to feature decomposition to extract The principal eigenvectors are expressed as: ， ， in, represents the extracted main eigenvector matrix, Indicates feature vectors, is the number of retained eigenvectors, represents the total number of eigenvalues, Indicates eigenvalues, Indicates the preset cumulative contribution rate threshold; The main eigenvector and the fault feature vector Combine to form a graph feature matrix , expressed as: ， in, Represents the horizontal concatenation operation of the matrix, the graph feature matrix Contains complete information on temporal and spatial correlation features and fault characteristics; The sliding time window mechanism is used to map the feature matrix Update by sliding the time window step Control the update frequency of the graph, the updated graph feature matrix It is expressed as: ， in, Indicates the current time The global state representation vector is is the smoothing factor, the value range is [0,1], is the time interval, The feature matrix computed for the new time window.

5. The reliability analysis method of the power system according to claim 4, characterized in that: Constructing a dual-channel graph attention layer based on the spatiotemporal correlation features and fault features to perform feature extraction; The first channel is the spatiotemporal correlation channel to process the spatiotemporal correlation tensor The second channel is the fault feature channel to process the fault feature vector Fault information in The attention coefficient of the spatiotemporal correlation channel is expressed as: ， in, represents the attention coefficient of the spatiotemporal correlation channel, Indicates the current center node The spatiotemporal correlation characteristics of Neighbor node The spatiotemporal correlation characteristics of is the feature transformation matrix; Representation Node The set of neighboring nodes; represents the rectified linear activation function with leakage, Representation Node All neighbor nodes of The spatiotemporal correlation characteristics of The attention coefficient of the fault feature channel is expressed as: ， in, represents the attention coefficient of the fault feature channel, Current central node The fault feature vector is Neighbor node The fault feature vector of is the feature transformation matrix, Representation Node All neighbor nodes of The fault feature vector of The importance of the two types of information is dynamically adjusted through the learnable fusion coefficient, which is expressed as: ， in, and is the fusion coefficient of adaptive learning, represents the Sigmoid activation function, is the final combined attention weight, representing the node For Node The overall level of attention; Based on the combined attention weights, a weighted aggregation operation is performed on the node features to generate enhanced features that consider spatial correlation, which is expressed as: ， in, is the transformation matrix of spatiotemporal correlation features, is the transformation matrix of fault characteristics, is the spatial domain enhanced feature matrix, is the spatiotemporal correlation feature vector, representing the node The spatiotemporal correlation characteristics of is the fault feature vector, representing the node The fault related information.

6. The reliability analysis method of the power system according to claim 5, characterized in that: Based on the enhanced features, a causal convolution layer based on a gating mechanism is constructed within the time window determined by the graph feature matrix to obtain the temporal features, which are expressed as: ， The gate control unit calculation is expressed as: ， in, is the causal convolution kernel size, is the gating unit parameter, For the The convolution kernel parameters at each position, is the weight matrix of the gating unit, is the bias term of the gating unit, Indicates the current time The state feature vector of is the time step The state feature vector of is the time step Temporal convolution features of Based on the temporal convolutional features, the temporal domain attention score is calculated to identify the key temporal patterns, which is expressed as: ， in, is the temporal attention parameter matrix, represents the transposed matrix of the temporal attention parameter matrix, is the time window length The length of the time series within Indicates that at time step Convolutional features when ; is the time step The attention weight of The time domain attention weights are weighted and combined with the temporal convolution features to generate spatiotemporal fusion features, which can be expressed as: ， in, The spatiotemporal fusion feature representation of the system state contains complete spatiotemporal information. Represents the time step The attention weight, Represents the time step Temporal convolution features.

7. The reliability analysis method of the power system according to claim 6, characterized in that: The spatiotemporal fusion features are double-mapped to form a double state representation, which is a risk state representation vector for the local component state and a global state representation vector that describes the overall operation status of the system, expressed as: ， ， in, represents the risk state representation vector, describing the local component state, Represents the global state representation vector, describing the overall state of the system, represents the mapping weight matrix, represents the bias vector; Based on the dual state representation, the risk index of the local component and the reliability index of the system as a whole are calculated respectively, and the state information is converted into a quantifiable risk metric through nonlinear transformation, which is expressed as: ， ， in, Represents the local component risk index vector, reflecting the risk level of each component, Represents the global system reliability index, characterizing the overall reliability status of the system. is the activation function, To evaluate the weight matrix, Represents the bias vector.

8. The reliability analysis method of the power system according to claim 7, characterized in that: Constructing a two-layer strategy network architecture based on the risk index and reliability index; The two-layer strategy network architecture includes a local intervention strategy network and a global coordination strategy network, which generate intervention decisions at corresponding levels based on risk state representation and global state representation, respectively, and are expressed as: ， ， in, represents the local policy network, generating component-level control strategies, represents the global policy network, generating the system-level coordination policy, represents the network parameters, Represents a local decision network based on local node features Generate local control strategies; Represents the global decision network, based on the global state representation Generate global control strategy; Through adaptive weight calculation, the local intervention strategy and the global coordination strategy are dynamically integrated to form the final collaborative decision-making strategy, which is expressed as: ， ， in, is the collaborative weight vector, reflecting the importance of strategies at different levels, represents the fusion weight matrix, For the final decision strategy, represents the weight coefficient of the local control strategy, Represents the weight coefficient of the global control strategy.

9. A reliability analysis system for a power system, based on the reliability analysis method for a power system according to any one of claims 1 to 8, characterized in that: include: The acquisition module is used to collect data to build a dynamic spatiotemporal graph network of the power system, including time dimension information of component operation status, maintenance records, and fault history, as well as spatial dimension information of physical connection, electrical association, and geographic location; The feature extraction module extracts features from the spatiotemporal graph network based on a spatiotemporal feature learning framework constructed based on a dual-channel graph attention network and a gated causal convolutional network, and realizes spatial domain feature extraction through adaptive fusion of physical association and fault features, and uses gated causal convolution and temporal attention to capture the temporal evolution pattern, generating spatiotemporal fusion features containing complete spatiotemporal information; The analysis module converts the spatiotemporal fusion features into local risk states and global system states through a state feature mapping network, calculates local component risk indicators and system overall reliability indicators, generates a collaborative decision-making strategy through a risk weight adaptive fusion mechanism, and realizes reliability analysis of the power system.

Images