Flow data differentiated analysis method applied to a substation

Abstract

The application provides a flow data differential analysis method applied to a transformer substation, and relates to the technical fields of electric power and artificial intelligence, and comprises the following steps: a digital twin model of a transformer substation operation process is constructed in advance; wherein the digital twin model comprises various flow monitoring points; flow data collected by the flow monitoring points covering various electric power equipment arranged in the transformer substation are acquired and imported into the digital twin model; electric power equipment abnormal information representing that the electric power equipment monitored by the flow monitoring points is abnormal is generated according to the flow data; a target flow monitoring point for generating the electric power equipment abnormal information is acquired from the digital twin model, and target flow data corresponding to the target flow monitoring point is acquired; and a differential evaluation model based on deep learning and pre-trained is used to perform differential analysis on the target flow data to obtain a differential analysis result. The scheme can improve the accuracy of differential analysis of transformer substation flow data, and further improve the accuracy of transformer substation abnormal state monitoring.

Description

Technical Field

[0001] This invention relates to the fields of power and artificial intelligence technologies, and in particular to a method for differential analysis of flow data applied to substations. Background Technology

[0002] Substations are locations where voltage is changed. They contain various electrical equipment such as transformers, reactors, switchgear, voltage transformers, and current transformers. When electrical equipment like transformers experiences a fault, it doesn't immediately stop working; it only ceases operation as the fault gradually worsens. When a fault in electrical equipment becomes severe, it can paralyze multiple operating systems, leading to power outages and disrupting people's lives. Detecting a fault in its early stages can prevent it from escalating and causing damage to electrical equipment. Therefore, monitoring substations is crucial for timely detection of anomalies and for taking appropriate countermeasures, thus preventing potential safety risks and improving the stability of the power grid.

[0003] Monitoring substation anomalies based on substation flow data is currently the primary monitoring method. For example, threshold monitoring technology is used to analyze flow data, and then anomalies are determined based on the differential analysis of the flow data. However, the accuracy of existing substation flow data differential analysis is poor, thus reducing the accuracy of anomaly analysis. Summary of the Invention

[0004] In view of this, and to address the above shortcomings, it is necessary to propose a method for differential analysis of flow data in substations, so as to improve the accuracy of differential analysis of flow data in substations, and thus improve the accuracy of monitoring abnormal conditions in substations.

[0005] This invention provides a method for differential analysis of flow data applied to substations, comprising:

[0006] A digital twin model of the substation operation process is pre-constructed; wherein the digital twin model includes various flow monitoring points; wherein each flow monitoring point is used to monitor at least one type of power data from the power equipment;

[0007] The flow data collected from the flow monitoring points covering each power equipment in the substation is obtained and imported into the digital twin model.

[0008] Based on the flow data, generate power equipment anomaly information indicating that the power equipment monitored at the flow monitoring point is abnormal;

[0009] Obtain the target flow monitoring points from the digital twin model to generate the abnormal information of the power equipment;

[0010] Obtain the target flow data corresponding to the target flow monitoring point from the flow data;

[0011] The target traffic data is subjected to differential analysis using a pre-trained differential evaluation model based on deep learning, and the differential analysis results are obtained.

[0012] Preferably, the pre-construction of the digital twin model of the substation operation process includes:

[0013] Construct a three-dimensional model corresponding to the physical entities of each power equipment during the current operation of the substation;

[0014] Obtain the communication link relationships between various flow monitoring points during substation operation;

[0015] The three-dimensional models corresponding to each power device are used as nodes in a directed topology graph, and the communication link relationships between each flow monitoring point are used as the connection relationships between nodes to construct a directed topology graph.

[0016] The traffic data collected from each traffic monitoring point is matched with the three-dimensional model corresponding to each node of the directed topology graph to obtain the digital twin model.

[0017] Preferably, the acquisition of flow data collected from flow monitoring points covering various power equipment within the substation includes:

[0018] Obtain the process unit characteristics of each power equipment in the current substation; wherein, the process unit characteristics include functional characteristics, operating parameter characteristics, structural characteristics, and safety and reliability characteristics;

[0019] Process information is extracted based on the structural features of the process unit; wherein, the process information includes the assembly sequence and assembly relationship of the power equipment;

[0020] Based on the process information, determine the flow monitoring points for each power device;

[0021] Based on the functional characteristics of the process unit, determine the power data monitoring type for each flow monitoring point;

[0022] Real-time acquisition of flow data for each power data monitoring type corresponding to each flow monitoring point.

[0023] Preferably, the step of generating power equipment anomaly information indicating that the power equipment monitored at the flow monitoring point is abnormal, based on the flow data, includes:

[0024] Preset the flow threshold, flow deviation threshold, and abnormal time threshold corresponding to each power data monitoring type at each flow monitoring point;

[0025] For each flow monitoring point, the difference between the flow data of that flow monitoring type and the corresponding flow threshold is calculated to obtain the deviation flow.

[0026] The deviation flow rate is compared with the flow deviation threshold, and when the deviation flow rate is greater than the flow deviation threshold, the time when the deviation flow rate corresponding to the current power data monitoring type is greater than the flow deviation threshold is accumulated.

[0027] When the cumulative time exceeds the abnormal time threshold, the corresponding power data monitoring type is marked as an abnormal type, and power equipment abnormal information of the power equipment at the flow monitoring point where the abnormal type is located is generated.

[0028] Preferably, the step of performing differential analysis on the target traffic data using a pre-trained differential evaluation model based on deep learning includes:

[0029] Based on deep learning, a differential assessment model is pre-trained for differential analysis of flow data in substations.

[0030] The target traffic data is subjected to temporal and spatial feature extraction to obtain a traffic spatiotemporal feature sequence;

[0031] The spatiotemporal characteristic sequence of the flow rate is input into the differential evaluation model, and the differential analysis results of the substation flow rate data are output.

[0032] Preferably, the step of extracting temporal and spatial features from the target traffic data includes:

[0033] A pre-trained LTSM neural network for predicting traffic change trends from traffic data;

[0034] The target flow data is input into the LTSM neural network to obtain the target flow change trend characteristics of the target flow monitoring point within the current monitoring period;

[0035] The flow change trend features are imported into the nodes corresponding to the pre-trained directed topology graph of the substation; wherein, each flow monitoring point in the directed topology graph of the substation constitutes a node of the topology graph.

[0036] Using the flow monitoring point corresponding to the target flow data as the target node, determine the coupling weight between the target node and each adjacent node in the directed topology graph of the substation;

[0037] Based on the coupling weights, the feature aggregation representation of each adjacent node acting on the target node is determined, and the spatiotemporal feature sequence of the traffic is obtained.

[0038] Preferably, determining the coupling weights between the target node and each adjacent node in the directed topology graph of the substation includes:

[0039] The system acquires the first historical flow time series sequence collected by the target node and its adjacent nodes in the directed topology graph of the substation when no abnormal information of power equipment is generated within the historical monitoring period; and acquires the second historical flow time series sequence collected by the adjacent nodes when the target node in the directed topology graph of the substation generates abnormal information of power equipment within the historical monitoring period.

[0040] Calculate the first average flow information of the power data monitoring type corresponding to the target node and its neighboring nodes based on the first historical flow time series; and calculate the second average flow information of the power data monitoring type corresponding to the neighboring nodes based on the second historical flow time series.

[0041] Based on the first historical flow time series and the first average flow information, a first flow amplitude time series of the power data type corresponding to the target node and its adjacent nodes is determined; and based on the second historical flow time series and the second average flow information, a second flow amplitude time series of the power data type corresponding to the adjacent nodes is determined.

[0042] Based on the first flow amplitude time series and the second flow amplitude time series, calculate the difference coefficient of the power data monitoring type corresponding to adjacent nodes;

[0043] The coupling weight is determined based on the difference coefficient.

[0044] Preferred,

[0045] When determining the first flow amplitude time series of the power data type corresponding to the target node and its adjacent nodes based on the first historical flow time series and the first average flow information, or when determining the corresponding second flow amplitude time series based on the second historical flow time series and the second average flow information, the following calculation formula is used:

[0046]

[0047] Among them, H i X represents the time series sequence of the flow amplitude at node i. iL Let Q represent the flow amplitude of node i at time L, N represent the total number of time points, and Q represent the flow amplitude of node i at time L. iL Let Q represent the flow of node i at time L. iL A represents the average flow rate of node i, and δ is the conversion coefficient;

[0048] And / or,

[0049] When calculating the differentiation coefficient of the power data monitoring type corresponding to adjacent nodes based on the first flow amplitude time series and the second flow amplitude time series, the following formula is used:

[0050]

[0051] Where y(ij) represents the difference coefficient between node i and node j, X iL X represents the flow amplitude of node i at time L, obtained from the first flow amplitude time series. jL Y represents the flow amplitude of node j at time L, obtained from the first flow amplitude time series. iL Y represents the flow amplitude of node i at time L, obtained from the second flow amplitude time series. jL This represents the flow amplitude of node j at time L, obtained from the second flow amplitude time series.

[0052] And / or,

[0053] When determining the coupling weight based on the difference coefficient, the following calculation formula is used:

[0054] e ij =θy(ij)

[0055] Among them, e ij The coupling weight between node i and node j is represented by y(ij), which represents the differentiation coefficient between node i and node j, and θ represents the transformation coefficient.

[0056] Preferably, determining the feature aggregation representation of each adjacent node acting on the target node based on the coupling weight includes obtaining it using the following calculation formula:

[0057] F r * =ReLu(pF) r +∑ z∈ur y(rz)pF z )

[0058] Among them, F r * Let represent the feature aggregation representation of node r, ReLU represent the activation function, p represent the parameter matrix of the feature transformation, y(rz) represent the coupling weights of node r to node z, and u r Let F represent the set of neighboring nodes of node r. r and F z These represent the target traffic change trend characteristics for nodes r and z, respectively.

[0059] Preferably, after obtaining the differential analysis results of the substation flow data, the process further includes:

[0060] Obtain the estimated flow time series of power data monitoring types for other flow monitoring points adjacent to the target flow monitoring point;

[0061] The estimated flow time series of each adjacent flow monitoring point is compared with the corresponding flow threshold, flow deviation threshold and abnormal time threshold;

[0062] Based on the comparison results, determine whether each adjacent flow monitoring point has generated potential anomalies, and the estimated timestamp of potential anomaly generation.

[0063] Based on the adjacent traffic monitoring points that generate the potential anomaly type and the estimated timestamp, corresponding anomaly prevention measures are generated.

[0064] As can be seen from the above technical solution, the fault monitoring method for substation power equipment provided by the present invention firstly constructs a digital twin model of the substation operation process in advance, and then imports the flow data collected from the flow monitoring points of each power equipment in the substation into the digital twin model. Further, based on the flow data, abnormal information indicating abnormalities in the power equipment at the flow monitoring points is generated and displayed through the digital twin model. This allows for the determination of the target flow monitoring point from the digital twin model, and the acquisition of the target flow data corresponding to the target flow monitoring point from the flow data. Furthermore, by using a pre-trained differential evaluation model to perform differential analysis on the target flow data, differential analysis results of the substation flow data can be obtained, enabling the analysis of abnormal states in the substation. Therefore, this solution constructs a digital twin model of the substation operation process, which can simulate the substation operation process in real time, thereby achieving dynamic online monitoring of each piece of equipment in the substation. Furthermore, based on the digital twin model, a differentiated assessment model was constructed using deep learning. This model enables real-time dynamic differential analysis of substation flow data collected from various flow monitoring points. This real-time, dynamic online analysis improves the accuracy of substation flow data differential analysis, allowing for timely identification of any abnormal conditions within the substation. Moreover, by taking effective measures to address substation anomalies before they fully develop, more serious accidents and losses can be avoided once the early stages of a fault have matured. Attached Figure Description

[0065] Figure 1 A flowchart illustrating a method for differential analysis of flow data applied to substations, provided as an embodiment of the present invention. Detailed Implementation

[0066] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] like Figure 1 As shown, this invention provides a method for differential analysis of flow data applied to substations, which may include the following steps:

[0068] Step 101: Pre-construct a digital twin model of the substation operation process; wherein the digital twin model includes various flow monitoring points; wherein each flow monitoring point is used to monitor at least one type of power data from the power equipment;

[0069] Step 102: Obtain the flow data collected from the flow monitoring points covering each power equipment in the substation, and import them into the digital twin model;

[0070] Step 103: Based on the flow data, generate power equipment anomaly information indicating that the power equipment monitored at the flow monitoring point is abnormal;

[0071] Step 104: Obtain the target flow monitoring points for generating power equipment anomaly information from the digital twin model;

[0072] Step 105: Obtain the target flow data corresponding to the target flow monitoring point from the flow data;

[0073] Step 106: Use a pre-trained deep learning-based differential evaluation model to perform differential analysis on the target traffic data and obtain the differential analysis results.

[0074] In this embodiment, by constructing a digital twin model of the substation operation process, the substation's operation can be simulated in real time, thereby achieving dynamic online monitoring of various devices within the substation. Furthermore, based on the digital twin model, a differentiated evaluation model is built using deep learning. This model can perform dynamic differentiated analysis of the substation's flow data in real time based on the flow data collected from each flow monitoring point. This real-time dynamic online analysis improves the accuracy of the differentiated analysis of the substation's flow data, enabling timely analysis to determine whether any abnormal conditions exist within the substation. Moreover, by taking effective measures to address abnormal conditions in the substation before a fault matures, more serious accidents and losses can be avoided once the early fault has matured.

[0075] For step 101, a digital twin model of the substation operation process is pre-constructed;

[0076] In this embodiment, a digital twin model of the substation's operation process is constructed to achieve online monitoring of the substation's power equipment. Simultaneously, by combining the flow data collected from various flow monitoring points, differential analysis of the substation's flow data is performed to monitor abnormal states of the substation. The constructed digital twin model includes various flow monitoring points; each flow monitoring point is used to monitor at least one type of power data from the power equipment. Specifically, the digital twin model can be constructed in the following way:

[0077] Construct a three-dimensional model corresponding to the physical entities of each power equipment during the current operation of the substation;

[0078] Obtain the communication link relationships between various flow monitoring points during substation operation;

[0079] The three-dimensional models corresponding to each power device are used as nodes in a directed topology graph, and the communication link relationships between each flow monitoring point are used as the connection relationships between nodes to construct a directed topology graph.

[0080] The flow data collected from each flow monitoring point is matched with the 3D model corresponding to each node of the directed topology graph to obtain a digital twin model.

[0081] In this embodiment, the physical entity parameters of each power device during the current operation of the substation are obtained. Then, a simulation model of each power device's physical entity is constructed to obtain a three-dimensional model of each power device entity. Further, the communication link connections between each flow monitoring point during the current operation of the substation are obtained. The three-dimensional models corresponding to each power device are used as nodes in a directed topology graph, and the communication link connections between the flow monitoring points are used as the connection relationships between nodes to construct a directed topology graph. Finally, the flow data collected by each flow monitoring point is obtained, twin data is generated based on the flow, and the twin data is matched with the three-dimensional models corresponding to each node in the directed topology graph to generate a digital twin model. Thus, the constructed digital twin model can fully reflect the actual operation process of the substation. Based on this digital twin model, online monitoring of the substation's operation process can be achieved. Furthermore, based on this digital twin model, differentiated analysis of the substation's flow data can be achieved, thereby enabling real-time monitoring of abnormal states in the substation and preventing more serious losses caused by the maturation of faults in the substation.

[0082] For step 102, obtain the flow data collected from the flow monitoring points covering each power equipment in the substation and import them into the digital twin model.

[0083] In this step, we consider importing the traffic data collected from each traffic monitoring point into the digital twin model for display, storage, and analysis. Specifically, step 102 can be implemented as follows:

[0084] Obtain the process unit characteristics of each power equipment in the current substation; among which, the process unit characteristics include functional characteristics, operating parameter characteristics, structural characteristics, and safety and reliability characteristics;

[0085] Process information is extracted based on the structural features of the process unit; the process information includes the assembly sequence and assembly relationship of the power equipment.

[0086] Based on the process information, determine the flow monitoring points for each power device;

[0087] Based on the functional characteristics of the process unit, determine the power data monitoring type for each flow monitoring point;

[0088] Real-time acquisition of flow data for each power data monitoring type corresponding to each flow monitoring point.

[0089] In this embodiment, the power equipment included in the substation may include different types of power equipment such as transformers, circuit breakers, current transformers, voltage transformers, and grounding devices. The technological unit characteristics of each power equipment in the substation may include functional characteristics, operating parameter characteristics, structural characteristics, and safety and reliability characteristics. These characteristics determine the application and performance of the power equipment in the power system.

[0090] Specifically, in terms of functional characteristics, each power device has its own specific function. For example, transformers are used for voltage transformation, circuit breakers are used to control the on / off state of circuits, and current transformers and voltage transformers are used to monitor current and voltage data. These functions determine the role and status of power devices in the power system.

[0091] Different types of power equipment have different operating parameters, such as the rated capacity, rated voltage, and short-circuit impedance of transformers, and the rated current and rated short-circuit current carrying capacity of circuit breakers. These operating parameters determine the scope of use and performance of power equipment in the power system.

[0092] Regarding structural features, each type of power equipment has its specific structural features, such as the oil tank, windings, and cooling system of a transformer, and the contacts, spring mechanism, and arc-extinguishing chamber of a circuit breaker. These structural features determine the manufacturing process and maintenance methods of the equipment.

[0093] Regarding safety and reliability characteristics, the safety and reliability of power equipment is a crucial aspect of its technological unit characteristics. This includes characteristics such as the equipment's insulation performance, thermal stability, and mechanical strength, to ensure the equipment can operate safely and stably under various operating conditions.

[0094] Process information primarily indicates the assembly sequence and relationships of electrical equipment. Within a substation, the assembly sequence and relationships of various electrical devices are typically determined according to the structural and functional requirements of the power system to ensure its safe and stable operation. For example, circuit breakers are usually installed between the power input terminal and various circuits to control the on / off state of the circuits. Circuit breakers are often connected to transformers, cables, and other equipment to achieve precise control and protection of the circuits.

[0095] For example, current transformers and voltage transformers are usually connected to circuit breakers or transformers to monitor current and voltage data and transmit the data to the monitoring system. They are usually installed in key locations in the power system to ensure accurate monitoring of the power system's operating status.

[0096] For example, grounding devices are used to ground the system's grounding point to ensure the safe operation of the system. Grounding devices are usually installed near transformers or circuit breakers and connected to the metal casing of electrical equipment.

[0097] For example, control and monitoring equipment includes monitoring systems, protection devices, remote signaling and control devices, etc., which are used to monitor and control the operating status of the power system. These devices are usually installed in the control room or operation and maintenance center and connected to various power equipment to achieve centralized monitoring and control of the entire power system.

[0098] The types of power data monitoring can include, but are not limited to, electrical parameters such as rated voltage, rated current, and rated capacity; operating parameters such as temperature, pressure, and humidity; and power quality parameters such as voltage fluctuation, frequency fluctuation, and harmonic content.

[0099] For example, for transformers, the corresponding power data monitoring types include, but are not limited to, oil temperature monitoring, oil level monitoring, transformer winding temperature monitoring, partial discharge monitoring, and humidity monitoring. Monitoring transformer oil temperature ensures that the internal temperature of the transformer is within a safe range, preventing damage due to excessively high oil temperatures; monitoring the oil level in the transformer tank ensures the required amount of oil for normal operation, preventing insufficient oil levels from affecting the insulation performance of the equipment; monitoring the temperature of the transformer windings ensures that the windings will not be damaged due to overheating; monitoring potential partial discharges within the transformer, such as voltage fluctuations, frequency fluctuations, and harmonic content, can prevent internal insulation faults; and monitoring the humidity of the environment surrounding the transformer ensures that the transformer's insulation performance is not affected by a humid environment.

[0100] For step 103, based on the flow data, generate power equipment anomaly information indicating that the power equipment monitored at the flow monitoring point is abnormal.

[0101] In this step, it is determined whether there is a continuous abnormal flow data in the flow data corresponding to each power data monitoring type, and based on this, power equipment abnormality information is generated to indicate that the corresponding power equipment is abnormal.

[0102] Specifically, step 103 can be implemented in the following way:

[0103] Preset the flow threshold, flow deviation threshold, and abnormal time threshold corresponding to each power data monitoring type at each flow monitoring point;

[0104] For each flow monitoring point, the difference between the flow data of that flow monitoring type and the corresponding flow threshold is calculated to obtain the deviation flow.

[0105] Compare the deviation flow rate with the flow deviation threshold, and when the deviation flow rate is greater than the flow deviation threshold, accumulate the time when the deviation flow rate corresponding to the current power data monitoring type is greater than the flow deviation threshold.

[0106] When the cumulative time exceeds the abnormal time threshold, the corresponding power data monitoring type is marked as an abnormal type, and power equipment abnormal information for the power equipment at the flow monitoring point where the abnormal type is located is generated.

[0107] In this embodiment, firstly, flow thresholds, flow deviation thresholds, and abnormal time thresholds are preset for each flow monitoring point's power data monitoring type. Then, the flow data corresponding to the power data type collected from each flow monitoring point is compared with the corresponding flow threshold. By calculating the difference, the deviation flow between the flow data corresponding to the power data monitoring type and the corresponding flow threshold can be obtained. Further, by determining the relationship between the deviation flow corresponding to each power data monitoring type and the flow deviation threshold, if the deviation flow corresponding to a certain power data monitoring type is greater than the deviation flow threshold, and the accumulated time is greater than the preset abnormal time threshold, it indicates that the power equipment corresponding to the current power data monitoring type is abnormal, and thus, power equipment abnormality information corresponding to that power equipment is generated.

[0108] For steps 104 and 105, the target flow monitoring points for generating abnormal power equipment information are obtained from the digital twin model, and the target flow data corresponding to the target flow monitoring points are obtained from the flow data.

[0109] After generating abnormal information about the power equipment, this step can use a digital twin model to query the target flow monitoring point corresponding to the generated abnormal information. Then, the target flow data collected by the target flow monitoring point can be obtained from the flow data collected by the target flow monitoring point, that is, the flow data of the power equipment in abnormal condition.

[0110] For step 106, a differential evaluation model based on deep learning pre-trained is used to perform differential analysis on the target traffic data to obtain the differential analysis results.

[0111] This step considers building a differentiated assessment model based on deep learning on top of the digital twin model, and using this differentiated assessment model to perform differentiated analysis on substation flow data, thereby enabling real-time monitoring of abnormal substation conditions. Specifically, step 106 can be achieved through the following steps:

[0112] S61: A pre-trained differential assessment model based on deep learning for differential analysis of flow data in substations;

[0113] S62: Extract time and space features from the target traffic data to obtain a traffic spatiotemporal feature sequence;

[0114] S63: Input the spatiotemporal characteristic sequence of flow into the differential assessment model and output the differential analysis results of substation flow data.

[0115] For step S61, when pre-training a differentiated assessment model for differential analysis of substation flow data based on deep learning, sample flow data from several historical monitoring periods are first acquired for each flow monitoring point. This sample flow data from several historical monitoring periods is used as the training and testing sets for the differentiated assessment model. The training set is then input into the initial differentiated assessment model for training until the loss function converges or the model's accuracy requirements are met, and the model parameters are saved. Finally, the differentiated assessment model is tested using the test set until it meets the model's reliability requirements, and the differentiated assessment model is output.

[0116] For step S62, when performing temporal and spatial feature extraction on the target traffic data, it can be achieved in the following way:

[0117] A pre-trained LTSM neural network for predicting traffic change trends from traffic data;

[0118] Input the target flow data into the LTSM neural network to obtain the target flow change trend characteristics of the corresponding target flow monitoring points within the current monitoring period;

[0119] The flow change trend features are imported into the nodes corresponding to the pre-trained directed topology graph of the substation; where each flow monitoring point in the directed topology graph of the substation constitutes a node of the topology graph.

[0120] Using the flow monitoring point corresponding to the target flow data as the target node, determine the coupling weight between the target node and each adjacent node in the directed topology graph of the substation.

[0121] Based on the coupling weight, the feature aggregation representation of each adjacent node acting on the target node is determined, and the spatiotemporal feature sequence of traffic is obtained.

[0122] In this embodiment, historical flow time-series sequences corresponding to the power data types collected by each flow monitoring point within the historical monitoring period are obtained, and an LTSM neural network is constructed based on this to predict the changing trends of flow data. Then, after power equipment anomaly information is generated at the flow monitoring points, the target flow data corresponding to the anomaly type is input into the trained LTSM neural network, outputting the target flow change trend features of the corresponding target flow monitoring point. Furthermore, a graph attention network can be used to learn the directed topology graph of the substation. The target flow change trend features obtained from each flow monitoring point in the current monitoring period are imported into the corresponding nodes of the directed topology graph of the substation. The node corresponding to the flow monitoring point that generated the power equipment anomaly information in the directed topology graph of the substation is taken as the target node, and the coupling weight of the target node to its adjacent nodes in the directed topology graph of the substation is determined. Further, the coupling weights are assigned to each adjacent node, and the neighbor aggregation mechanism of the graph attention network is used to obtain the feature aggregation representation of each adjacent node acting on the target node, thereby obtaining the spatiotemporal feature sequence of flow. By fully considering the impact of adjacent nodes on the target node, that is, the correlation between the flow data collected by adjacent flow monitoring points and the flow data collected by the current flow monitoring point, the obtained data not only includes time characteristics but also spatial characteristics. It fully considers the correlation between various flow monitoring points or various power equipment, and thus, based on the obtained flow spatiotemporal characteristic sequence, differential analysis can be performed to obtain more accurate differential analysis results, thereby making the monitoring results of substation abnormal conditions more accurate.

[0123] Furthermore, when determining the coupling weights between the target node and its neighboring nodes in the directed topology graph of the substation, this can be achieved in the following way:

[0124] The first historical flow time series sequence collected by the target node and its adjacent nodes in the directed topology graph of the substation when no abnormal information of power equipment is generated within the historical monitoring period is obtained; and the second historical flow time series sequence collected by the adjacent nodes when the target node in the directed topology graph of the substation generates abnormal information of power equipment within the historical monitoring period is obtained.

[0125] Calculate the first average flow information of the target node and its neighboring nodes according to the first historical flow time series; and calculate the second average flow information of the neighboring nodes according to the second historical flow time series.

[0126] Based on the first historical flow time series and the first average flow information, determine the first flow amplitude time series of the power data type corresponding to the target node and its adjacent nodes; and based on the second historical flow time series and the second average flow information, determine the second flow amplitude time series of the power data type corresponding to the adjacent nodes.

[0127] Based on the first and second flow amplitude time series, calculate the difference coefficient of the power data monitoring type corresponding to adjacent nodes;

[0128] The coupling weights are determined based on the differentiation coefficients.

[0129] Specifically, when determining the first flow amplitude time series of the power data type corresponding to the target node and its adjacent nodes based on the first historical flow time series and the first average flow information, and when determining the corresponding second flow amplitude time series based on the second historical flow time series and the second average flow information, the following calculation formula can be used:

[0130]

[0131] Among them, H i X represents the time series sequence of the flow amplitude at node i. iL Let Q represent the flow amplitude of node i at time L, N represent the total number of time points, and Q represent the flow amplitude of node i at time L. iL Let Q represent the flow of node i at time L. iL A represents the average flow rate of node i, and δ is the conversion coefficient.

[0132] When calculating the differentiation coefficient of the power data monitoring type corresponding to adjacent nodes based on the first flow amplitude time series and the second flow amplitude time series, the following formula can be used:

[0133]

[0134] Where y(ij) represents the difference coefficient between node i and node j, X iL X represents the flow amplitude of node i at time L, obtained from the first flow amplitude time series. jL Y represents the flow amplitude of node j at time L, obtained from the first flow amplitude time series. iL Y represents the flow amplitude of node i at time L, obtained from the second flow amplitude time series. jL This represents the flow amplitude of node j at time L, obtained from the second flow amplitude time series.

[0135] When determining the coupling weight based on the differentiation coefficient, the following formula can be used:

[0136] eij =θy(ij)

[0137] Among them, e ij The coupling weight between node i and node j is represented by y(ij), which represents the differentiation coefficient between node i and node j, and θ represents the transformation coefficient.

[0138] Furthermore, the feature aggregation representation after determining the interaction between each adjacent node and the target node based on the coupling weights can be obtained using the following formula:

[0139] F r * =ReLu(pF) r +∑ z∈ur y(rz)pF z )

[0140] Among them, F r * Let represent the feature aggregation representation of node r, ReLU represent the activation function, p represent the parameter matrix of the feature transformation, y(rz) represent the coupling weights of node r to node z, and u r Let F represent the set of neighboring nodes of node r. r and F z These represent the target traffic change trend characteristics for nodes r and z, respectively.

[0141] Furthermore, after step 106, corresponding anomaly prevention measures can be generated based on the results of the differential analysis. Specifically, after obtaining the differential analysis results of the substation flow data, the following can be further included:

[0142] Obtain the estimated flow time series of power data monitoring types from other flow monitoring points adjacent to the target flow monitoring point;

[0143] The estimated flow time series of each adjacent flow monitoring point is compared with the corresponding flow threshold, flow deviation threshold and abnormal time threshold;

[0144] Based on the comparison results, determine whether each adjacent flow monitoring point has generated potential anomalies, and the estimated timestamp of potential anomaly generation.

[0145] Based on the adjacent traffic monitoring points that generate potential anomalies and the estimated timestamps, corresponding anomaly prevention measures are generated.

[0146] In this embodiment, after determining the differential analysis results of the substation flow data, the estimated flow time series of other adjacent nodes can also be obtained. Then, based on the method in step 103, it is compared with the corresponding flow threshold, flow deviation threshold, and abnormal time threshold to determine the potential anomaly type and the estimated timestamp of the anomaly type. Based on the potential anomaly type and the estimated timestamp, corresponding anomaly prevention measures can be generated to further avoid the losses and risks brought about by the maturity of potential faults.

[0147] This specification also provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the methods in any of the embodiments of the specification.

[0148] This specification also provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it implements the method in any of the embodiments of the specification.

[0149] The device embodiments provided by the present invention are based on the same inventive concept as the method embodiments in this specification. For details, please refer to the description in the method embodiments of this specification, which will not be repeated here.

[0150] The modules or units in the device of this invention can be merged, divided, and deleted according to actual needs. The above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the invention. Those skilled in the art will understand that implementing all or part of the processes of the above embodiments and making equivalent changes according to the claims of this invention still fall within the scope of the invention.

Claims

1. A method for differential analysis of flow data applied to substations, characterized in that, include: A digital twin model of the substation operation process is pre-constructed; wherein the digital twin model includes various flow monitoring points; wherein each flow monitoring point is used to monitor at least one type of power data from the power equipment; The flow data collected from the flow monitoring points covering each power equipment in the substation is obtained and imported into the digital twin model. Based on the flow data, generate power equipment anomaly information indicating that the power equipment monitored at the flow monitoring point is abnormal; Obtain the target flow monitoring points from the digital twin model to generate the abnormal information of the power equipment; Obtain the target flow data corresponding to the target flow monitoring point from the flow data; The target traffic data is subjected to differential analysis using a pre-trained differential evaluation model based on deep learning, and the differential analysis results are obtained. The differential analysis of the target traffic data using a pre-trained deep learning-based differential evaluation model includes: Based on deep learning, a differential assessment model is pre-trained for differential analysis of flow data in substations. The target traffic data is subjected to temporal and spatial feature extraction to obtain a traffic spatiotemporal feature sequence; The spatiotemporal characteristic sequence of the flow rate is input into the differential evaluation model, and the differential analysis results of the substation flow rate data are output. The extraction of temporal and spatial features from the target traffic data includes: A pre-trained LSTM neural network is used to predict the trend of traffic changes in traffic data. The target flow data is input into the LSTM neural network to obtain the target flow change trend characteristics of the target flow monitoring point within the current monitoring period; The flow change trend features are imported into the nodes corresponding to the pre-trained directed topology graph of the substation; wherein, each flow monitoring point in the directed topology graph of the substation constitutes a node of the topology graph. Using the flow monitoring point corresponding to the target flow data as the target node, determine the coupling weight between the target node and each adjacent node in the directed topology graph of the substation; Based on the coupling weights, the feature aggregation representation of each adjacent node acting on the target node is determined, and the spatiotemporal feature sequence of the traffic is obtained. Determining the coupling weights between the target node and its neighboring nodes in the directed topology graph of the substation includes: The system acquires the first historical flow time series sequence collected by the target node and its adjacent nodes in the directed topology graph of the substation when no abnormal information of power equipment is generated within the historical monitoring period; and acquires the second historical flow time series sequence collected by the adjacent nodes when the target node in the directed topology graph of the substation generates abnormal information of power equipment within the historical monitoring period. Calculate the first average flow information of the power data monitoring type corresponding to the target node and its neighboring nodes based on the first historical flow time series; and calculate the second average flow information of the power data monitoring type corresponding to the neighboring nodes based on the second historical flow time series. Based on the first historical flow time series and the first average flow information, a first flow amplitude time series of the power data type corresponding to the target node and its adjacent nodes is determined; and based on the second historical flow time series and the second average flow information, a second flow amplitude time series of the power data type corresponding to the adjacent nodes is determined. Based on the first flow amplitude time series and the second flow amplitude time series, calculate the difference coefficient of the power data monitoring type corresponding to adjacent nodes; The coupling weight is determined based on the difference coefficient.

2. The method for differential analysis of flow data applied to substations according to claim 1, characterized in that, The pre-built digital twin model of the substation operation process includes: Construct a three-dimensional model corresponding to the physical entities of each power equipment during the current operation of the substation; Obtain the communication link relationships between various flow monitoring points during substation operation; The three-dimensional models corresponding to each power device are used as nodes in a directed topology graph, and the communication link relationships between each flow monitoring point are used as the connection relationships between nodes to construct a directed topology graph. The traffic data collected from each traffic monitoring point is matched with the three-dimensional model corresponding to each node of the directed topology graph to obtain the digital twin model.

3. The method for differential analysis of flow data applied to substations according to claim 1, characterized in that, The acquisition of flow data collected from flow monitoring points covering various power equipment within the substation includes: Obtain the process unit characteristics of each power equipment in the current substation; wherein, the process unit characteristics include functional characteristics, operating parameter characteristics, structural characteristics, and safety and reliability characteristics; Process information is extracted based on the structural features of the process unit; wherein, the process information includes the assembly sequence and assembly relationship of the power equipment; Based on the process information, determine the flow monitoring points for each power device; Based on the functional characteristics of the process unit, determine the power data monitoring type for each flow monitoring point; Real-time acquisition of flow data for each power data monitoring type corresponding to each flow monitoring point.

4. The method for differential analysis of flow data applied to substations according to claim 1, characterized in that, The step of generating power equipment anomaly information based on the flow data, indicating that the power equipment monitored at the flow monitoring point is abnormal, includes: Preset the flow threshold, flow deviation threshold, and abnormal time threshold corresponding to each power data monitoring type at each flow monitoring point; For each flow monitoring point, the difference between the flow data of that flow monitoring type and the corresponding flow threshold is calculated to obtain the deviation flow. The deviation flow rate is compared with the flow deviation threshold, and when the deviation flow rate is greater than the flow deviation threshold, the time when the deviation flow rate corresponding to the current power data monitoring type is greater than the flow deviation threshold is accumulated. When the cumulative time exceeds the abnormal time threshold, the corresponding power data monitoring type is marked as an abnormal type, and power equipment abnormal information of the power equipment at the flow monitoring point where the abnormal type is located is generated.

5. The method for differential analysis of flow data applied to substations according to claim 1, characterized in that, When determining the first flow amplitude time series of the power data type corresponding to the target node and its adjacent nodes based on the first historical flow time series and the first average flow information, or when determining the corresponding second flow amplitude time series based on the second historical flow time series and the second average flow information, the following calculation formula is used: ， in, Represents a node The time series of flow amplitude, Represents a node In the The flow rate amplitude at any given time, where N represents the total number of times. Represents a node In the Flow rate at any moment Represents a node Average flow, The conversion factor; And / or, When calculating the differentiation coefficient of the power data monitoring type corresponding to adjacent nodes based on the first flow amplitude time series and the second flow amplitude time series, the following formula is used: ， in, Represents a node With nodes The coefficient of difference between them This represents the node obtained from the first flow amplitude time series. In the The flow rate at any given moment This represents the node obtained from the first flow amplitude time series. In the The flow rate at any given moment This represents the node obtained from the second flow amplitude time series. In the The flow rate at any given moment This represents the node obtained from the second flow amplitude time series. In the The flow rate at any given moment; And / or, When determining the coupling weight based on the difference coefficient, the following calculation formula is used: ， in, Representation Nodes With nodes The coupling weights between them Represents a node With nodes The coefficient of difference between them This represents the conversion factor.

6. The method for differential analysis of flow data applied to substations according to claim 1, characterized in that, The step of determining the feature aggregation representation of each adjacent node acting on the target node based on the coupling weight includes obtaining it using the following calculation formula: ， in, Represents a node The feature aggregation representation, where ReLU represents the activation function. The parameter matrix representing the eigentransformation. This represents the coupling weight between node r and node z. Let r represent the set of neighboring nodes. and These represent the target traffic change trend characteristics for nodes r and z, respectively.

7. The method for differential analysis of flow data applied to substations according to any one of claims 1-6, characterized in that, After obtaining the differential analysis results of the substation flow data, further steps include: Obtain the estimated flow time series of power data monitoring types for other flow monitoring points adjacent to the target flow monitoring point; The estimated flow time series of each adjacent flow monitoring point is compared with the corresponding flow threshold, flow deviation threshold and abnormal time threshold; Based on the comparison results, determine whether each adjacent flow monitoring point has generated potential anomalies, and the estimated timestamp of potential anomaly generation. Based on the adjacent traffic monitoring points that generate the potential anomaly type and the estimated timestamp, corresponding anomaly prevention measures are generated.

Images