Abnormal cause and effect reasoning and root cause positioning method and system based on heterogeneous power grid data

Abstract

The application discloses an abnormality causal reasoning and root cause positioning method and system based on heterogeneous power grid data, and relates to the technical field of electric digital data processing. The method is based on the heterogeneous data of the power grid accessed by the power grid dispatching center in real time, dynamically updates the graph structure corresponding to the initial graph model pre-stored in the power grid dispatching center, so that the graph structure is kept synchronous with the real-time power grid topology; based on the dynamically updated node attributes and edge attributes, a heterogeneous power grid graph atlas with integrated structure and attribute update is formed, and the node attributes are analyzed in space-time correlation with the global voltage time series data to identify voltage abnormal events; based on the nonlinear coupling relationship of the voltage abnormal events in the dynamic power grid topology, the interaction effect between the load fluctuation and the voltage regulation action is quantitatively analyzed, so that the root cause type corresponding to the voltage abnormal event is positioned, the reliability of the abnormality causal reasoning and the accuracy of the root cause positioning are effectively improved, and the operation risk of the power grid is reduced.

Description

Technical Field

[0001] This invention relates to the field of electrical digital data processing technology, and in particular to a method and system for anomaly causal reasoning and root cause localization based on heterogeneous power grid data. Background Technology

[0002] In the process of industrial intelligent transformation, the operational complexity and uncertainty of various industrial systems have significantly increased. Multi-device collaboration, multi-process linkage, and dynamic changes in the external environment have resulted in industrial data exhibiting typical characteristics of being multi-source, heterogeneous, high-dimensional, and dynamic. This data encompasses multi-dimensional information such as equipment operation monitoring, production process parameters, environmental perception, and topological correlations, forming the core foundation for accurate perception and rapid handling of abnormal events. However, the system characteristics and data requirements of different industrial sectors vary significantly. Power grid systems, as critical infrastructure, have much higher requirements for topological correlations and strong temporal correlations in their data compared to general industrial systems. Directly applying existing industrial methods is insufficient to meet the accuracy and timeliness requirements for causal reasoning and root cause localization of power grid anomalies.

[0003] Based on this, existing technologies are mainly divided into two categories: one is based on traditional machine learning methods, which construct classification or regression models to identify anomalies by encoding and fusing features of heterogeneous power grid data, and then combine association rule mining to infer root causes; the other is based on graph model methods, which use the power grid topology to construct graph models, map heterogeneous data to the attributes of nodes and edges in the graph, and use graph neural networks and other technologies to analyze anomaly propagation paths and locate root causes.

[0004] While the aforementioned existing technical solutions can achieve preliminary identification and root cause prediction of power grid anomalies, in complex and dynamic power grid operation scenarios, detailed technical defects lead to insufficient accuracy in root cause localization, making it difficult to adapt to actual operation and maintenance needs. Taking the anomaly diagnosis scenario of a regional power grid dispatch center as an example, although existing graph model-based methods can construct a static graph model based on the power grid physical topology and complete the attribute mapping of equipment operation data and line status data, in scenarios of dynamic power grid topology adjustments, there are detailed problems such as lag in graph structure updates and rigid adaptation of heterogeneous data, which directly affect the accuracy of root cause localization. Specifically, this manifests as follows: Within the same dispatch center's jurisdiction, power grids dynamically adjust their topology during peak hours by switching standby transformers and during off-peak hours by disconnecting some branch lines. This topology is accompanied by dynamic updates of heterogeneous data, including fluctuations in photovoltaic power plant output and interference from meteorological factors. When extreme rainfall in a region causes a reduction in insulation in local lines, it leads to abnormal line impedance in that area, which then spreads to surrounding power supply areas, causing voltage fluctuations. In such cases, it is necessary to precisely coordinate voltage regulation equipment based on the differences in operating conditions in each region. However, the impact of regional meteorological data has spatial diffusion, and the response time of voltage regulation equipment varies across different regions. Furthermore, line monitoring data only corresponds to specific spatial nodes, with a fixed-dimensional attribute mapping method. This results in the inaccurate capture of the cross-regional impact of meteorological factors and the spatial linkage response of voltage regulation equipment. Graph models cannot analyze the root cause of voltage fluctuations based on cross-temporal and spatial correlations, mistakenly attributing the root cause of voltage fluctuations to voltage regulation equipment parameter drift rather than load-side voltage imbalance caused by insulation reduction due to extreme rainfall.

[0005] The reasoning bias caused by the lag in graph structure updates and the rigid adaptation of heterogeneous data across time and space will further weaken the accuracy of voltage fluctuation propagation path and impact range analysis. This will make voltage regulation commands lack specificity and unable to accurately control voltage deviations based on the actual causes in different regions. Even if manual intervention is subsequently implemented, it will delay the timing of voltage regulation, exacerbate the accumulation of voltage deviations, increase the risk of grid equipment overload and abnormal power consumption by users, and result in low closed-loop adaptive capability of voltage management. This will in turn affect the credibility of abnormal causal reasoning and the accuracy of root cause location. Summary of the Invention

[0006] To address the technical problems in the prior art, embodiments of the present invention provide a method and system for anomaly causal reasoning and root cause localization based on heterogeneous power grid data. The technical solution is as follows: On the one hand, a method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data is provided. This method includes: S1, dynamically updating the graph structure corresponding to the initial graph model pre-stored in the power grid dispatch center based on the real-time access of heterogeneous power grid data, so that the graph structure keeps synchronized with the real-time power grid topology; the topology operation data, equipment monitoring data and photovoltaic power output data contained in the heterogeneous power grid data form a one-to-one corresponding node attribute or many-to-one edge attribute with the physical nodes and associated edges in the graph structure; S2, based on the dynamically updated node attributes and edge attributes in S1, forming a heterogeneous power grid map with integrated structure and attribute updates, and performing spatiotemporal correlation analysis on the node attributes in the heterogeneous power grid map and the global voltage time series data to identify voltage anomaly events; S3, based on the nonlinear coupling relationship of voltage anomaly events in the dynamic power grid topology, quantitatively analyzing the interaction effect between load fluctuations and voltage regulation actions, thereby locating the root cause type corresponding to the voltage anomaly event; the root cause type includes voltage regulation response time difference caused by cross-regional collaborative anomalies and / or electrical parameter drift caused by voltage regulation equipment anomalies.

[0007] On the other hand, an anomaly causal reasoning and root cause localization system based on heterogeneous power grid data is provided. This system includes: a graph structure dynamic update module, which dynamically updates the graph structure corresponding to the initial graph model pre-stored in the power grid dispatch center based on the heterogeneous power grid data accessed in real time by the power grid dispatch center, so that the graph structure keeps synchronized with the real-time power grid topology; a voltage anomaly event identification module, which forms a heterogeneous power grid map with integrated structure and attribute updates based on the dynamically updated node attributes and edge attributes, and performs spatiotemporal correlation analysis on the node attributes in the heterogeneous power grid map and the global voltage time series data to identify voltage anomaly events; and an anomaly root cause localization module, which quantitatively analyzes the interaction effect between load fluctuations and voltage regulation actions based on the nonlinear coupling relationship of voltage anomaly events in the dynamic power grid topology, thereby locating the root cause type corresponding to the voltage anomaly event.

[0008] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: This invention addresses the shortcomings of existing graph model methods in complex dynamic power grid scenarios, which suffer from inaccurate root cause localization due to lagging graph structure updates and rigid adaptation of heterogeneous data. It combines regional power grid dispatching requirements and achieves accurate diagnosis through a three-step process. First, based on real-time heterogeneous data from the dispatch center, the graph structure of the initial graph model is dynamically updated to synchronize the graph structure with the real-time topology of the power grid. Simultaneously, topology, equipment monitoring, and photovoltaic output data are mapped to node and edge attributes, preventing data and structure from becoming disconnected during dynamic topology adjustments. Second, an integrated heterogeneous power grid graph is constructed based on the dynamically updated attributes and structure. Anomalies are identified through spatiotemporal correlation analysis of node attributes and global voltage time-series data. This accurately captures cross-temporal and spatiotemporal influencing factors such as meteorological and photovoltaic output factors, solving the problem of inaccurate capture of cross-regional impacts and equipment linkage responses, thus improving the accuracy of voltage anomaly event identification. Finally, based on the nonlinear coupling relationship of anomaly events, the interaction effect of load fluctuations and voltage regulation actions is quantified to locate the root causes of cross-regional collaborative anomalies and equipment parameter drift. This effectively avoids the problem of misjudging root causes using traditional methods, clarifies the anomaly propagation path and impact range, makes voltage regulation commands more targeted, avoids delays caused by manual intervention, reduces the risk of equipment overload and abnormal power consumption, and significantly improves the reliability and accuracy of root cause location in complex dynamic scenarios, adapting to actual operation and maintenance needs.

[0009] Specifically, to address the differences in grid topology and data characteristics between peak and off-peak electricity consumption periods, separate graph structure attribute update processes are developed to achieve accurate data mapping that dynamically adapts to changes in grid load. During peak periods, the process first collects time-series data on bus voltage and photovoltaic output fluctuations at physical nodes, as well as peak line power, power difference between high and low voltage sides of transformers, and time-series change rates for associated edges. By calculating voltage stability coefficients, photovoltaic output fluctuations, peak energy transmission ratios, and peak energy losses, these parameters are mapped to graph structure nodes and edges according to preset rules, completing the attribute update. During off-peak periods, similar core data is collected, and the cumulative average voltage, total off-peak power generation, total energy transmission, and losses are obtained through time-period integration, before completing the corresponding attribute mapping update. This design adapts to topology changes caused by off-peak line disconnections and load declines, weakening instantaneous data fluctuation interference through integral statistics and enhancing the characteristics of time-period operating conditions. The overall process adopts differentiated processing methods for data characteristics at different time periods to ensure accurate adaptation of heterogeneous data to graph structures. It accurately depicts dynamic parameters such as peak photovoltaic fluctuations and equipment power differences, and clearly presents steady-state characteristics such as off-peak energy transmission and accumulated voltage. This provides accurate and real-time graph structure data support for subsequent voltage anomaly identification and root cause localization, and improves the reliability of anomaly diagnosis in complex power grid scenarios.

[0010] First, based on the updated physical nodes and associated edges of the graph structure attributes, an initial topology framework is built according to the preset power grid topology connection relationship. Then, according to the attribute-entity mapping rule, the normalized peak-valley time period heterogeneous power grid data is matched to the corresponding nodes and edges, completing the construction of the heterogeneous power grid map, realizing the accurate fusion of structure and attributes, and improving the basic data quality for anomaly identification. Subsequently, spatiotemporal window units are divided, and the spatiotemporal voltage fluctuation value is calculated by comparing the current and previous spatiotemporal window average voltage values ​​within the same range and compared with the corresponding preset value. This design takes into account the temporal continuity and spatial correlation of voltage data, and can quickly capture potential voltage anomalies, improving the timeliness of anomaly monitoring. For cases where the fluctuation value exceeds the threshold, a graded judgment is made according to the number of node anomalies, deviation characteristics, and verification results: multiple nodes exceeding the same direction are judged as an abnormal event; a single node exceeding the threshold is temporarily recorded as a suspected disturbance and verified; two consecutive anomalies are judged as an event; otherwise, only a local warning is issued. The graded judgment mechanism effectively distinguishes between real anomalies and instantaneous disturbances, reducing false positives and false negatives. The overall process enhances the ability to perform cross-temporal and spatial correlation analysis, provides a reliable basis for precise voltage regulation, reduces the operational risks of power grid equipment, and adapts to the dynamic operation and maintenance needs of complex power grids.

[0011] First, the nonlinear coupling process of anomalies in the dynamic topology is monitored, and the anomaly-related time periods, including load fluctuations and voltage regulation actions, are defined to clearly capture the step-by-step transmission impact of the two on the bus voltage. By calculating the load fluctuation slope, voltage regulation action amplitude, and voltage change, two major indicators—the nonlinear interaction factor and the electrical parameter deviation rate—are constructed. This achieves dual quantification of the interaction effect between load and voltage regulation actions and the equipment parameter status, and collaborative calculations are performed to obtain an interaction effect judgment value reflecting the coupling strength between load fluctuations and voltage regulation actions. This multi-dimensional parameter collaborative calculation method avoids the one-sidedness of single-indicator judgment. Based on the numerical range and allowable deviation rate of the interaction effect judgment value, three types of root cause scenarios are classified. For coexisting scenarios, a cross-regional collaborative-equipment linkage composite voltage regulation process is executed. By setting a baseline voltage regulation amount and adjacent node compensation amount, targeted voltage regulation is achieved. This design takes into account anomalies caused by both single and composite factors, compensating for the lack of specificity in traditional voltage regulation commands and avoiding the accumulation of voltage deviations. After voltage regulation, closed-loop optimization is performed by reviewing the interaction effect judgment value to ensure that the voltage regulation effect meets the standards. The overall process accurately identifies three types of root causes: differences in voltage regulation response time, equipment parameter drift, and the coexistence of both. This solves the problem of ambiguous root cause location in complex scenarios, improves the accuracy and timeliness of voltage regulation commands, reduces the risk of equipment overload and abnormal power consumption, strengthens the closed-loop adaptive capability of power grid voltage management, and adapts to the dynamic operation and maintenance needs of regional power grids. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart of an anomaly causal reasoning and root cause localization method based on heterogeneous power grid data provided in an embodiment of the present invention; Figure 2 A flowchart of the composite voltage regulation process provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an anomaly causal reasoning and root cause localization system based on heterogeneous power grid data provided in an embodiment of the present invention; Figure 4 The logic diagram of the heterogeneous power grid provided in the embodiments of the present invention. Detailed Implementation

[0014] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0015] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0016] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0017] Embodiment 1 of this invention provides an anomaly causal reasoning and root cause localization method based on heterogeneous power grid data, such as... Figure 1 The flowchart shown is for an anomaly causal reasoning and root cause localization method based on heterogeneous power grid data. The processing flow of this method may include the following steps: S1. Based on the real-time access of heterogeneous power grid data from the power grid dispatch center, the graph structure corresponding to the initial graph model pre-stored in the power grid dispatch center is dynamically updated to keep the graph structure synchronized with the real-time power grid topology. The heterogeneous power grid data and the physical nodes and associated edges in the graph structure form a one-to-one correspondence of node attributes or many-to-one edge attributes. S2. Based on the dynamically updated node attributes and edge attributes in S1, a heterogeneous power grid map with integrated structure and attribute updates is formed. The node attributes in the heterogeneous power grid map are then subjected to spatiotemporal correlation analysis with the global voltage time series data to identify voltage anomaly events. S3. Based on the nonlinear coupling relationship of voltage anomaly events in the dynamic power grid topology, the interaction effect between load fluctuations and voltage regulation actions is quantitatively analyzed to locate the root cause type of the voltage anomaly event. The root cause types include voltage regulation response time differences caused by cross-regional collaborative anomalies and / or electrical parameter drift caused by voltage regulation equipment anomalies.

[0018] The power grid dispatch center is a management and control agency responsible for power grid operation monitoring, dispatching and command, load allocation, fault handling, and optimal allocation of power resources. It undertakes the function of ensuring the safe, stable, and efficient operation of the power grid. Heterogeneous power grid data refers to multi-dimensional power grid operation data with different sources and formats that are accessed by the dispatch center in real time. This data covers electrical quantity data (such as voltage, current, and power), status quantity data (such as switch opening and closing status and protection action signals), environmental quantity data (such as temperature and humidity), and business management data (such as equipment ledgers and maintenance records). Its core purpose is to provide comprehensive and accurate data support for power grid operation status perception, anomaly identification, causal reasoning, and root cause localization. The initial graph model is a basic graph structure model pre-built and stored in the dispatch center based on the power grid topology, equipment relationships, and historical operating experience. It represents the connection edges and associated attributes between power grid element nodes and elements. The construction process of this model is as follows: First, core equipment in the power grid, such as generators, transformers, lines, and switches, are designated as nodes in the graph, and each node is assigned attributes such as equipment type and rated parameters. Second, based on the actual wiring method and electrical connection relationship of the power grid, directly related nodes are connected as edges, and each edge is assigned attributes such as connection type, transmission capacity, and impedance value. Then, combining historical fault cases and operation and maintenance experience, logical connection edges between nodes (such as the connection edges between protection devices and protected equipment) are added, ultimately forming the initial graph model. The input of the initial graph model is static topology information of the power grid, inherent attribute parameters of the equipment, and historical operation association rules. The output is a basic graph structure that can intuitively reflect the topological connection, attribute characteristics, and logical association of power grid components. This graph structure is the concrete representation of the initial graph model, consisting of three parts: a set of nodes (representing power grid equipment), a set of edges (representing the topological or logical association between equipment), and a set of attributes (representing the characteristic parameters of nodes and edges). It serves as the basic carrier for subsequent dynamic updates of the graph structure, anomaly propagation path tracing, and root cause localization based on real-time heterogeneous data.

[0019] Specifically, S1 includes: The system acquires instantaneous time-series data of bus voltage and photovoltaic power output fluctuations for each physical node in the initial graph model. Simultaneously, it acquires peak line power time-series data, instantaneous power difference between high and low voltage sides of the transformer, and time-series change rate for each associated edge. The instantaneous bus voltage time-series data refers to the instantaneous voltage amplitude of each bus node over a continuous period. This data is generated through high-frequency sampling by voltage transformers and monitoring terminals at the bus, transmitted to the power grid dispatch center via the dispatch data network, and pre-processed. The photovoltaic power output fluctuation time-series data refers to the output power of the grid-connected photovoltaic power station, obtained from the grid connection point of the photovoltaic power station. Power monitoring devices collect data; peak power time-series data refers to the sequence of peak power of the transmission line within a preset time window, based on the instantaneous power collected by power metering devices at both ends of the line, and the extreme values ​​are calculated by the power grid dispatch center; instantaneous power difference between the high and low voltage sides of the transformer refers to the difference in power between the high and low voltage sides of the transformer at the same moment, calculated by the power grid dispatch center after data is collected synchronously by power monitoring devices on both sides; time-series change rate refers to the change in the instantaneous power difference between the high and low voltage sides of the transformer per unit time, calculated based on the difference data at continuous moments.

[0020] Within a preset power consumption period: First, the peak and valley values ​​of the instantaneous time-series data of the bus voltage are obtained (if there is more than one corresponding peak and valley value, the average is taken). The difference between the peak and valley values ​​is processed as the ratio of the integral mean of the instantaneous time-series data of the bus voltage within this period, and the result is recorded as the voltage stability coefficient, which is used to quantify the fluctuation stability of the bus voltage within this period. The sliding window integration method is used to perform sliding window integration processing on the photovoltaic power output fluctuation time-series data to obtain the photovoltaic power output fluctuation per unit time. Here, the sliding window integration processing refers to setting a fixed time window (such as 5 minutes, 10 minutes), sliding along the instantaneous time-series data of the bus voltage segment by segment with a preset step size (such as 1 minute, 2 minutes), performing integration calculation on the photovoltaic power output data within each window, and then dividing by the window duration to obtain the photovoltaic power output fluctuation per unit time. This can accurately capture the dynamic fluctuation characteristics of new energy power output in different periods and avoid single-point data errors.

[0021] Secondly, the maximum value of the corresponding line power peak time series data within the time period is taken as the line power peak value. Combined with the total integral of the line power time series data within the time period, the ratio of the line power peak value to the total integral value is used to obtain the energy transmission peak ratio. The instantaneous difference of power on the high and low voltage sides of the transformer is integrated to obtain the basic energy loss. Then, it is multiplied with the time series change rate to calculate the energy loss of the time period. Here, the time series change rate can correct the influence of instantaneous fluctuations on loss calculation, taking into account both the cumulative effect of loss and the dynamic change trend. The core of the product calculation is to dynamically adjust the basic energy loss with the time series change rate, so that the result is closer to the actual loss law of the transformer and avoids the loss calculation deviation caused by ignoring the influence of fluctuations by relying solely on integration.

[0022] Finally, the voltage stability coefficient and photovoltaic power output fluctuation are mapped one by one to the corresponding physical nodes in the graph structure according to the preset topology mapping rules, completing the attribute update of the physical nodes in the graph structure for the application electricity period. The preset topology mapping rules refer to the pre-defined matching criteria between parameters and node attributes, including the correspondence between parameter types and node attribute fields, and the verification logic of topology identifiers, ensuring that different types of parameters are accurately written to the corresponding nodes, while filtering invalid identifiers to avoid mapping errors. The peak percentage of line energy transmission and the energy loss value of the time period are mapped to the corresponding associated edges in the graph structure according to the preset topology affiliation relationship, completing the attribute update of the associated edge nodes for the application electricity period. The preset peak topology affiliation relationship refers to the binding rules between associated edges and parameters, clarifying the data types corresponding to the associated edges of lines and transformers, and the matching mechanism between topology identifiers and edge unique IDs. The entire process, through rules and relationship constraints, achieves accurate mapping of parameters to physical nodes and associated edges, completing the attribute update.

[0023] S2 includes: First, based on the pre-defined power grid topology connections, and using the updated physical nodes and associated edges as a foundation, a topological spatial relationship framework is constructed as the initial topological structure of the graph. The pre-defined power grid topology connections refer to fixed association rules established based on the actual wiring methods of the power grid, clearly defining the actual electrical connection logic, spatial correspondence, and topological hierarchy of each physical node (bus, photovoltaic power station, etc.) and associated edge (line, transformer, etc.), ensuring that the initial structure closely matches the actual operating architecture of the power grid. Following the pre-defined attribute-entity mapping rules, after normalization to eliminate dimensional differences in heterogeneous power grid data, each data point is matched to the physical nodes and associated edges of the initial topology. The pre-defined attribute-entity mapping rules refer to predefined binding criteria between heterogeneous data and topological entities, clearly defining the node / edge attribute fields corresponding to different types of data such as voltage and power, as well as data validity verification standards. Through rule constraints, precise adaptation between data and topological entities is achieved, supplementing the attribute information of the initial structure and ultimately completing the construction of the heterogeneous power grid graph.

[0024] Subsequently, the voltage monitoring values ​​of each physical node in the heterogeneous power grid map are obtained. Combined with the corresponding topology attributes and the power supply range of the current power grid, the corresponding spatiotemporal window units are divided by the preset personnel. The spatiotemporal window unit represents the power grid operation status monitoring unit built based on the time dimension and the spatial dimension, which is used to realize the time-segmented and regional analysis of voltage data. The time dimension covers the fixed monitoring period and data sampling interval, and the spatial dimension corresponds to the power supply zone boundary and the set of topology-related nodes.

[0025] Finally, the average voltage value of the voltage monitoring values ​​corresponding to each physical node within the current spatiotemporal window unit is obtained, and compared with the average voltage value of the same range in the previous spatiotemporal window unit. The absolute value of the difference between the two is recorded as the spatiotemporal voltage fluctuation value. If the spatiotemporal voltage fluctuation value is less than the preset spatiotemporal voltage fluctuation value, it is determined that there is no spatiotemporal correlation anomaly within the current spatiotemporal window, and the monitoring continues in the next spatiotemporal window unit. The preset spatiotemporal voltage fluctuation value is represented by the sum and average of historical spatiotemporal voltage fluctuation values ​​during the historical voltage anomaly event identification process. If the previous spatiotemporal window unit does not exist (e.g., at the initial moment of the system or due to missing data), the average value of the current window is directly used as the baseline, and its fluctuation value is temporarily recorded as 0. It is determined that there is no spatiotemporal correlation anomaly within the current spatiotemporal window, and the monitoring continues in the next spatiotemporal window unit. If the spatiotemporal voltage fluctuation value is not less than the preset spatiotemporal voltage fluctuation value, it is determined that there is a potential voltage anomaly, and the following steps are executed: If, within the current spatiotemporal window unit, the voltage monitoring values ​​of multiple physical nodes deviate from the corresponding reference voltage in the same direction, and the deviations all exceed the preset allowable range, this is considered a voltage anomaly event. Since the excessive deviations in the same direction by multiple nodes are not random fluctuations at a single point, they are likely caused by systemic problems such as regional power grid voltage imbalance, sudden load changes, or cross-regional coordination anomalies, rather than a single node failure. The preset allowable range is usually set based on the power grid voltage level (e.g., ±7% for 10kV bus and ±5% for 220kV bus). In actual power grid dispatching and maintenance, it can be fine-tuned according to regional load characteristics, the proportion of new energy grid connection, and equipment tolerance to adapt to different operating conditions.

[0026] If the deviation of the voltage monitoring value of a single physical node exceeds the preset allowable range, it may be caused by non-systematic and occasional factors such as sensor error at that node or transient electromagnetic interference, rather than an overall power grid anomaly. Therefore, it is temporarily recorded as a suspected voltage disturbance, prompting maintenance personnel to review it in the next time window unit to avoid misjudgment.

[0027] If two consecutive verification results are abnormal, it indicates that the deviation of the node is not accidental and has formed a stable abnormal state. It is likely caused by hidden dangers in the node-related equipment or local line faults, so it is judged as a voltage abnormality event. If the deviation is not abnormal after two consecutive verifications, it indicates that the deviation is random and there is no stable abnormal trend. Only a local voltage fluctuation warning is needed to remind the operation and maintenance personnel to pay attention to subsequent changes. There is no need to initiate a full abnormality handling process.

[0028] S3 includes: In power grid operation, the occurrence and development of voltage anomalies are essentially nonlinear dynamic processes involving the interaction and cascading transmission of loads, power sources, and voltage regulating equipment within the grid topology. When a voltage exceedance is detected at a node, the complete time interval covered by the voltage anomaly from its occurrence to its recovery is defined as the anomaly-related period. This period includes two key sub-stages: first, the load fluctuation period that triggers or exacerbates voltage deviations, manifested as significant power changes at relevant nodes; and second, the voltage regulation action period where the system intervenes to stabilize the voltage, encompassing control behaviors such as capacitor switching and transformer tap adjustment. This nonlinear coupling process is specifically manifested as random load fluctuations and discrete actions of voltage regulating equipment, propagating cascading through the electrical connection links of the power grid (lines, transformers, etc.), and their interaction ultimately manifests in the stability changes of bus voltages at each level. Spatiotemporal correlation analysis of the graph nodes and edge attributes within this period is crucial for locating the root cause of the anomaly and evaluating the control effect.

[0029] During periods of load fluctuation, differential calculations are performed on the collected load power time-series data to measure the degree and direction of load changes, yielding the load fluctuation slope per unit time. This slope is the ratio of the difference between the load power time-series data at the end of the load fluctuation period and the load power time-series data at the beginning of the period, to the corresponding duration of the load fluctuation period. The load power time-series data refers to the sequence of total active power values ​​recorded at fixed sampling intervals (e.g., 1-5 minutes) from SCADA monitoring, reflecting the net electricity demand of the load point or load area. This calculation quantifies the continuous load fluctuation process into a concise slope index (kW / min or MW / h). The sign of this index intuitively indicates whether the load is increasing (positive) or decreasing (negative), while its absolute value characterizes the severity of the load change, providing crucial input for subsequent analysis of its impact on voltage.

[0030] During the voltage regulation operation period, the corresponding voltage regulation amplitude is obtained by converting the voltage change of the tap corresponding to the tap of the voltage regulating equipment into the tap voltage. That is, the product of the tap change and the rated voltage regulation capacity of the corresponding tap. Each tap of the transformer tap or on-load tap changer corresponds to a certain change step size. Its rated voltage regulation capacity (usually expressed in kV / tap or per unit value) is calibrated by the equipment manufacturer at the factory. The rated voltage regulation capacity is set by the preset personnel based on the equipment nameplate parameters, technical manual and on-site measured calibration data. It is a constant that reflects the inherent voltage regulation performance of the equipment.

[0031] By performing differential calculations on the bus voltage time series data within the abnormal correlation period, the voltage change during the abnormal correlation period is obtained. That is, the difference between the bus voltage time series data at the end of the abnormal correlation period and the bus voltage time series data at the beginning of the abnormal correlation period. The bus voltage time series data refers to the sequence of measured values ​​collected and transmitted in real time from the bus voltage transformer of the substation. It is usually the effective value (kV) of the line voltage or phase voltage, and is the most direct state variable reflecting the voltage level of the node.

[0032] The product of the load fluctuation slope and the voltage regulation amplitude is used as a reference value. The absolute value of the difference between the voltage change and the reference value is then processed as the ratio of its absolute value to the maximum voltage change during the abnormal correlation period, and this ratio is denoted as the nonlinear interaction factor. The product of the load fluctuation slope and the voltage regulation amplitude constitutes a linear reference value. Its physical meaning is that, assuming a perfectly linear system response, the load fluctuation combined with the voltage regulation action should produce a voltage change. The absolute value of the difference between the voltage change and the reference value represents the deviation between the actual voltage change and the expected linear change. The ratio of its absolute value to the maximum voltage change during the abnormal correlation period quantifies the proportion of the actual linear deviation in the total fluctuation amplitude of this voltage anomaly event, achieving a normalized representation of the degree of nonlinear interaction. This ratio eliminates numerical differences caused by different voltage levels and anomaly severity, giving the nonlinear interaction factor a unified and comparable measurement standard. It can accurately reflect the nonlinear strength of the coupling effect between load fluctuation and voltage regulation action. The closer the nonlinear interaction factor is to 0, the smaller the deviation between the actual voltage change and the linear expectation, and the closer the system response is to linearity.

[0033] The ratio of the absolute value of the difference between the impedance or admittance of the corresponding power supply branch of the voltage regulating equipment and its corresponding rated value to the corresponding rated value is recorded as the electrical parameter deviation rate. By co-calculating the nonlinear interaction factor and the electrical parameter deviation rate, an interaction effect judgment value reflecting the coupling strength between load fluctuations and voltage regulation actions is obtained. This value is the product of the nonlinear interaction factor and the electrical parameter deviation rate plus 1, reflecting the amplification mechanism of static parameter mismatch on dynamic nonlinear effects. This judgment value not only quantifies the complexity of the dynamic process but also reveals that when the electrical parameters of the power grid (such as impedance) deviate from the design value, it will systematically change the voltage sensitivity of the network, thereby significantly exacerbating the unpredictability and fluctuation intensity of the actual interaction between the load and the voltage regulating equipment.

[0034] To determine the root cause of voltage anomalies based on interaction effect judgment values, it is necessary to combine the judgment interval classification with the electrical parameter deviation rate for comprehensive judgment. The first judgment interval is the low to moderate interaction effect range (e.g., 0.1-0.5), corresponding to the weak to moderate degree of collaborative interaction between cross-regional power grids. The second judgment interval is the medium to high interaction effect range (e.g., 0.4-0.9), corresponding to the medium to strong degree of interaction between equipment within a single region. The overlapping interval (0.4-0.5) corresponds to the intermediate state of the superposition of cross-regional and single-regional interaction effects. In actual power grid dispatching and operation and anomaly investigation, the specific judgment interval values ​​mentioned above can be fine-tuned according to the complexity of the power grid topology, the proportion of grid-connected capacity of new energy sources (such as photovoltaic and wind power), the aging degree of voltage regulating equipment, and the regional power supply load characteristics. If the interaction effect judgment value is only in the first judgment interval that does not contain overlapping intervals, and the electrical parameter deviation rate is lower than the lower limit of the corresponding allowable deviation rate critical fluctuation interval, since this interval corresponds to the cross-regional power grid collaborative interaction characteristics, the fact that the electrical parameter deviation rate does not exceed the standard indicates that there is no essential equipment fault. The anomaly originates from the time difference caused by the asynchronous transmission of cross-regional dispatching instructions and voltage regulation response. Therefore, it is judged to be a voltage regulation response time problem caused by cross-regional collaborative anomaly.

[0035] If the interaction effect judgment value is only in the second judgment interval that does not contain overlapping intervals, and the electrical parameter deviation rate is not lower than the upper limit of the corresponding allowable deviation rate critical fluctuation interval, since this interval corresponds to the single-area equipment operation interaction characteristics, the significant parameter deviation indicates that the equipment itself is abnormal. It is highly likely that the voltage regulating equipment (such as transformers, reactive power compensation devices) is aging or malfunctioning, causing the electrical parameters to drift. Therefore, it is judged as this type of root cause.

[0036] If the interaction effect judgment value is in the overlapping part of the two intervals, and the electrical parameter deviation rate is within the corresponding allowable deviation rate critical fluctuation range, it indicates that the anomaly is simultaneously affected by cross-regional coordination asynchrony and equipment performance fluctuations. The superposition of these two factors leads to critical parameter fluctuations, thus it is judged as a coexistence of dual anomalies. The allowable deviation rate critical fluctuation range is the upper and lower limits of the allowable deviation rate critical fluctuation range, which needs to be comprehensively set in conjunction with power grid design standards, equipment rated parameters, and historical operation and maintenance data. It can also be fine-tuned according to the actual fault tolerance capability and operating condition differences of the power grid to adapt to the anomaly judgment requirements in different scenarios, ensuring the accuracy and practicality of the judgment results. At this time, a single voltage regulation method cannot completely solve the problem. Cross-regional coordination can correct time differences, and equipment linkage can calibrate parameter drift. The combination of the two can take into account both types of anomaly causes. Therefore, a composite voltage regulation process of cross-regional coordination and equipment linkage needs to be executed, such as... Figure 2 The flowchart shown below illustrates the combined voltage regulation process: The voltage regulating equipment identified as malfunctioning is designated as the target voltage regulating node. A pre-set reference voltage regulation amount is applied to the target voltage regulating node. This reference voltage regulation amount is set based on the equipment's rated voltage regulation range, the grid voltage standard value, and historical voltage regulation experience (e.g., the reference voltage regulation amount for transformers is set to ±0.5kV, and for reactive power compensation devices, it is set to ±2Mvar). The product of the difference between the electrical parameter deviation rate and the corresponding allowable deviation rate and the reference voltage regulation amount, and the ratio of this product to the allowable deviation rate, is recorded as the voltage compensation amount for adjacent voltage regulating nodes. Adjacent voltage regulating nodes are those that have a direct electrical connection with the target voltage regulating node and whose voltage coupling is higher than that of the target voltage regulating node. For a power grid voltage regulation unit with a preset coupling degree, the preset coupling degree is represented by the sum and average of historical voltage coupling degrees during the historical voltage regulation equipment anomaly identification process. Voltage coupling degree refers to the ratio calculated by multiplying the active power transmission between the target voltage regulation node and the adjacent voltage regulation node by the product of the impedance value and the square of the rated voltage. The result is recorded as voltage coupling degree. Based on the addition of the reference voltage regulation amount and the voltage compensation amount, the total voltage regulation amount of the corresponding target voltage regulation node and the adjacent voltage regulation nodes is obtained, and used as the input of the voltage regulation equipment (transformer tap changer, reactive power compensation device) of the corresponding voltage regulation node to perform bus side voltage adjustment.

[0037] After adjusting the voltage on the bus side, the interaction effect judgment value is reacquired. If the reacquired interaction effect judgment value is only in the interval of the first judgment interval that does not contain overlapping intervals, or only in the interval of the second judgment interval that does not contain overlapping intervals, it indicates that the single abnormal cause has been eliminated, the cross-regional coordinated or equipment linkage voltage regulation has achieved the expected effect, and the power grid has returned to the stable operation state under the influence of a single factor. Therefore, the cross-regional coordinated-equipment linkage composite voltage regulation process is terminated, and the voltage regulation parameters (baseline voltage regulation amount, compensation amount) and execution sequence are recorded. Otherwise, a voltage regulation ratio redistribution prompt is sent, such as suggesting that the baseline voltage regulation amount of the target node be increased by 20% and the compensation amount of the adjacent node be reduced by 10%.

[0038] It should be noted that the preset values ​​involved in the embodiments of the present invention, including but not limited to preset coupling degree, preset reference voltage regulation amount, and allowable deviation rate, all need to be further manually verified based on the power grid operation procedures to exclude abnormal extreme values ​​under extreme operating conditions such as voltage surges and drops and equipment failures, so as to ensure that the set values ​​conform to the normal operating scenario and improve the reliability of the voltage regulation strategy.

[0039] It provides an anomaly causal reasoning and root cause localization system based on heterogeneous power grid data, such as Figure 3The diagram shows the structure of an anomaly causal reasoning and root cause localization system based on heterogeneous power grid data. This system may include: a graph structure dynamic update module, used to dynamically update the graph structure corresponding to the initial graph model pre-stored in the power grid dispatch center based on real-time access to heterogeneous power grid data, ensuring the graph structure remains synchronized with the real-time power grid topology; a voltage anomaly event identification module, used to form a heterogeneous power grid graph with integrated structure and attribute updates based on dynamically updated node and edge attributes, and to perform spatiotemporal correlation analysis between node attributes in the heterogeneous power grid graph and global voltage time-series data to identify voltage anomaly events; and an anomaly root cause localization module, used to quantitatively analyze the interaction effect between load fluctuations and voltage regulation actions based on the nonlinear coupling relationship of voltage anomaly events in the dynamic power grid topology, thereby locating the root cause type corresponding to the voltage anomaly event.

[0040] In this first embodiment, the initial graph model is dynamically updated through real-time heterogeneous data. Combined with multi-dimensional parameter calculation and rule-based mapping, the graph structure iterates synchronously with the power grid topology and operating status, providing accurate data support for anomaly analysis. The spatiotemporal window unit division and multi-level anomaly judgment mechanism effectively distinguish between systemic anomalies and single-point occasional disturbances, avoiding false positives and false negatives, and improving the reliability of anomaly identification. By quantifying the interaction effect between load and voltage regulation actions, and combining dual judgment intervals and parameter deviation rates to comprehensively locate the root cause, accurate differentiation of cross-regional collaborative anomalies, equipment anomalies, and the coexistence of dual anomalies is achieved. The targeted composite voltage regulation process and preset value optimization mechanism take into account both cross-regional collaboration and equipment linkage, adapting to different power grid operating conditions, quickly eliminating anomalies, and ensuring the safe and stable operation of the power grid. The overall method and system realize closed-loop management of anomaly identification, root cause location, and voltage regulation handling, improving the intelligent level of power grid operation and maintenance, and adapting to the complex power grid operation needs under a high proportion of new energy grid connection.

[0041] Step S1 in Example 1 focuses on capturing the power grid operation characteristics under conditions of severe load fluctuations and unstable renewable energy output. The calculated indicators focus on volatility and peak characteristics. Therefore, this step is a graph structure attribute update process corresponding to peak electricity consumption periods, where the preset electricity consumption period refers to peak electricity consumption periods. However, during off-peak electricity consumption periods, the load level is low and gradual, and the equipment loss patterns and renewable energy output characteristics differ significantly from those during peak periods. Updating the graph structure solely based on peak periods cannot fully cover the power grid's operation status throughout all periods, affecting the completeness and accuracy of subsequent anomaly identification and root cause localization. Therefore, based on Example 1, a supplementary example corresponding to off-peak electricity consumption periods needs to be added, namely Example 2 of this invention.

[0042] Embodiment 2 of the present invention: A graph structure attribute update process during off-peak electricity consumption periods includes: acquiring the bus voltage time-series data and photovoltaic output time-series data of each physical node in the initial graph model, and simultaneously acquiring the line power time-series data and the power difference between the high and low voltage sides of the transformer for each associated edge in the initial graph model; within a preset off-peak electricity consumption period, performing time-series integration processing on the bus voltage time-series data, photovoltaic output time-series data, line power time-series data, and the power difference between the high and low voltage sides of the transformer to obtain the cumulative voltage value, total power generation during the off-peak period, total energy transmission, and energy loss during the off-peak period; mapping the cumulative voltage value and total power generation during the off-peak period to the corresponding physical nodes of the graph structure one by one according to the preset off-peak topology mapping rules, thus completing the graph structure physical node attribute update during the off-peak electricity consumption period; mapping the total energy transmission and energy loss during the off-peak period to the corresponding associated edges of the graph structure according to the preset off-peak topology affiliation, thus completing the graph structure associated edge attribute update during the off-peak electricity consumption period.

[0043] The preset off-peak topology attribution and mapping rules share the same core logic, parameter and topology entity matching criteria, topology identifier verification mechanism, and binding logic as the preset topology attribution and mapping rules in the corresponding process during peak electricity consumption periods. Both provide rule constraints to ensure accurate mapping of parameters to graph structure nodes and associated edges. The only difference between the two electricity consumption periods lies in the parameter calculation methods, which are adapted to the operating characteristics of each period: peak periods emphasize the calculation of fluctuation and peak values, while off-peak periods emphasize the calculation of cumulative values. This differentiated indicator design adapts to all time-related operating conditions, enabling the graph structure to represent the power grid's operating status in all dimensions.

[0044] Peak electricity consumption periods refer to the periods during which the real-time load monitoring values ​​collected by the power grid dispatch center are consistently higher than the first-level rated load capacity of the power grid for a preset monitoring period (e.g., 1 hour) or longer, and the peak value of the load monitoring value at any given moment is not lower than the first-level rated load capacity of the power grid. The first level is usually set at 80%-85% (e.g., when the rated load is 1000MW, the first level corresponds to 800-850MW), which aligns with the normal peak load of the power grid and takes into account both equipment redundancy and dispatch early warning requirements. Low electricity consumption periods refer to the periods during which the real-time load monitoring values ​​collected by the power grid dispatch center are consistently lower than the second-level rated load capacity of the power grid for the same preset monitoring period or longer, and the valley value of the load monitoring value at any given moment is not higher than the second-level rated load capacity of the power grid. The second level is set at 30%-40% (e.g., corresponding to 300-400MW), which is suitable for off-peak conditions such as nighttime. The first and second amplitudes mentioned above are both set according to the rated load ratio of the power grid. Because the load ratio is more adaptable to power grids of different voltage levels, the first amplitude ensures that the equipment is not overloaded during peak hours, while the second amplitude distinguishes between off-peak and flat load periods, providing a precise time period definition basis for updating the time-sharing diagram structure and anomaly analysis.

[0045] The off-peak electricity consumption period graph structure attribute update process in Example 2 complements the peak period process in Example 1, significantly improving the completeness and accuracy of the power grid's full-time graph structure representation. The process adapts to the conditions of smooth load and stable losses during off-peak periods. By calculating cumulative indicators through time-period integration, it avoids the inapplicability of peak-fluctuation indicators to off-peak conditions, accurately capturing voltage stability, total renewable energy output, and equipment energy consumption characteristics during off-peak periods. Relying on core mapping rules consistent with those during peak periods, it ensures the standardization and uniformity of attribute updates across all time periods, laying the foundation for subsequent cross-time period anomaly comparison analysis. Simultaneously, the accurately updated off-peak period nodes and edge attributes fill the gap left by traditional updates that only focus on peak periods, helping to comprehensively understand the power grid's full-load cycle operation status. This provides data support for identifying potential anomalies during off-peak periods and optimizing equipment low-load losses, further enhancing the comprehensiveness and intelligence of power grid operation and maintenance.

[0046] It should be added that, Figure 1 The first condition refers to: the voltage monitoring values ​​of non-single physical nodes deviate from the corresponding reference voltage in the same direction, and the magnitude of the deviation from the corresponding reference voltage exceeds the preset allowable range; the second condition refers to: the voltage monitoring value of only a single physical node deviates from the corresponding reference voltage by a magnitude exceeding the preset allowable range; the first judgment condition refers to: only within the first judgment interval that does not contain overlapping intervals, and the electrical parameter deviation rate is lower than the lower limit of the corresponding allowable deviation rate critical fluctuation range; the second judgment condition refers to: only within the second judgment interval that does not contain overlapping intervals, and the electrical parameter deviation rate is not lower than the upper limit of the corresponding allowable deviation rate critical fluctuation range; the third judgment condition refers to: within the overlapping interval of the first and second judgment intervals, and the electrical parameter deviation rate is within the corresponding allowable deviation rate critical fluctuation range.

[0047] like Figure 4The logical diagram of the heterogeneous power grid shown uses 10kV feeders F15 / F16 (distribution network event clusters) as the core information hub. Through color and hierarchical division, it constructs a four-layer heterogeneous structure: logical clusters, physical devices, topology hubs, and attribute scenarios. In this diagram, green nodes represent core physical devices such as the ZW32-12 circuit breaker, 100kVA distribution transformer T01, residential load L05, and 5MW photovoltaic power station PV01, clearly presenting core distribution network units such as transformers, circuit breakers, photovoltaic power stations, and load nodes. Yellow nodes are distribution network event clusters, aggregating all feeder information. Light blue nodes represent topology hubs such as the Guangming 10kV substation and the Chengxi 220kV power supply zone, as well as type / scenario nodes such as 10kV overhead lines and distributed photovoltaic access points. Gray nodes are labeled with attribute information such as commissioning and maintenance times. The heterogeneous power grid map accurately maps the physical transmission links between equipment and across regions through electrical connection edges such as busbar connections, feeder connections, and 10kV tie lines. Simultaneously, it uses heterogeneous association edges containing equipment, equipment type, and commissioning time to synchronously mount topology and attribute information. It not only meets the needs of refined operation and maintenance of distribution networks but also supports macro-level business decisions such as fault diagnosis and load analysis. With dynamic update capabilities, it provides integrated information support for the digital operation of distribution networks.

[0048] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0049] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0050] In various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0051] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0052] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data, characterized in that, The method includes: S1, Based on the heterogeneous power grid data accessed in real time by the power grid dispatch center, the graph structure corresponding to the initial graph model pre-stored in the power grid dispatch center is dynamically updated so that the graph structure is synchronized with the real-time power grid topology; The heterogeneous data of the power grid and the physical nodes and associated edges in the graph structure form a one-to-one correspondence of node attributes or many-to-one edge attributes. S2, based on the dynamically updated node attributes and edge attributes in S1, forms a heterogeneous power grid map with integrated structure and attribute updates, and performs spatiotemporal correlation analysis on the node attributes in the heterogeneous power grid map and the global voltage time series data to identify voltage anomaly events; S3. Based on the nonlinear coupling relationship of the voltage anomaly event in the dynamic power grid topology, quantitatively analyze the interaction effect between load fluctuation and voltage regulation action to locate the root cause type of the voltage anomaly event. The root cause types include differences in voltage regulation response time caused by cross-regional coordination anomalies and / or electrical parameter drift caused by voltage regulation equipment anomalies.

2. The method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data as described in claim 1, characterized in that, S1 includes: The process of updating graph structure attributes during peak electricity demand periods by the power grid dispatch center, or the process of updating graph structure attributes during off-peak electricity demand periods by the power grid dispatch center. The peak electricity consumption period refers to the period during which the real-time load monitoring value collected by the power grid dispatch center is continuously higher than the first level of the rated load capacity of the power grid during a preset monitoring period or longer, and the peak value of the load monitoring value at a single moment is not lower than the first level of the rated load capacity of the power grid. The term "off-peak electricity consumption period" refers to the period during which the real-time load monitoring value collected by the power grid dispatch center is continuously lower than the second-level rated load capacity of the power grid during the aforementioned preset monitoring period and above, and the valley value of the load monitoring value at any given moment is not higher than the second-level rated load capacity of the power grid.

3. The method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data as described in claim 2, characterized in that, The process for updating graph structure attributes during peak electricity usage periods by the power grid dispatch center includes: Acquire the instantaneous time-series data of bus voltage and photovoltaic power output fluctuation of each physical node in the initial graph model, and simultaneously acquire the peak power time-series data of each associated edge, the instantaneous power difference between the high and low voltage sides of the transformer, and the time-series change rate. During the preset peak electricity consumption period, the peak and valley values ​​of the instantaneous time-series data of the bus voltage are obtained, and the ratio of the difference between them to the integral mean of the instantaneous time-series data of the bus voltage during that period is recorded as the voltage stability coefficient. The photovoltaic power output fluctuation time series data is integrated by a sliding window to obtain the photovoltaic power output fluctuation per unit time. The maximum value of the line power peak time series data is taken as the line power peak value. Combined with the total integral of the line power time series data within this period, the energy transmission peak percentage is obtained. The instantaneous power difference between the high and low voltage sides of the transformer is integrated and corrected by synchronously superimposing the time-series change rate to obtain the energy loss during peak hours. The voltage stability coefficient, the photovoltaic power output fluctuation, and the topology identifier of the corresponding physical node are mapped one by one to the corresponding physical node of the graph structure according to the preset peak topology mapping rules, thereby completing the update of the graph structure physical node attributes during peak electricity consumption periods. The peak percentage of line energy transmission, the energy loss value during peak hours, and the topology identifier of the corresponding associated edges are mapped to the corresponding associated edges of the graph structure according to the preset peak topology affiliation relationship, thereby completing the attribute update of the associated edge nodes during peak electricity consumption periods.

4. The anomaly causal reasoning and root cause localization method based on heterogeneous power grid data as described in claim 2, characterized in that, The process for updating graph structure attributes during off-peak electricity hours by the power grid dispatch center includes: Obtain the bus voltage time series data and photovoltaic power output time series data of each physical node in the initial graph model, and simultaneously obtain the line power time series data and the power difference between the high and low voltage sides of the transformer for each associated edge in the initial graph model; During the preset off-peak electricity consumption period, the bus voltage time series data, photovoltaic output time series data, line power time series data and transformer high and low voltage side power difference are integrated by time period to obtain the time period voltage accumulation value, the total power generation during the off-peak period, the total energy transmission amount and the energy loss during the off-peak period. The cumulative voltage value during the time period and the total power generation during the off-peak period are mapped one by one to the corresponding physical nodes of the graph structure according to the preset off-peak topology mapping rules, thereby completing the update of the graph structure physical node attributes during the off-peak period. The total energy transmission and the energy loss during off-peak hours are mapped to the corresponding associated edges of the graph structure according to the preset off-peak topology, thus completing the update of the graph structure associated edge attributes during off-peak hours.

5. The method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data as described in claim 1, characterized in that, S2 includes: The updated physical nodes and associated edges are connected according to the preset power grid topology. The framework corresponding to the topological spatial relationship between the physical nodes and the associated edges is used as the initial topology structure of the heterogeneous power grid graph. According to the preset attribute-entity mapping rules, after updating the graph structure attributes, the heterogeneous power grid data during peak electricity consumption periods or off-peak electricity consumption periods are normalized and then matched one by one to each physical node and associated edge in the original topology to complete the construction of the heterogeneous power grid graph.

6. The method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data as described in claim 1, characterized in that, S2 further includes: The voltage monitoring values ​​of each physical node in the heterogeneous power grid map are obtained. Combined with the corresponding topological attributes and the power supply range of the current power grid, the corresponding spatiotemporal window units are divided. The spatiotemporal window units represent the power grid operation status monitoring units constructed based on the time and space dimensions. Obtain the average voltage value of the voltage monitoring values ​​of each physical node in the current spatiotemporal window unit, and compare it with the average voltage value of the same range in the previous spatiotemporal window unit. Record the absolute value of the difference between the two as the spatiotemporal voltage fluctuation value. If the spatiotemporal voltage fluctuation value is less than the preset spatiotemporal voltage fluctuation value, it is determined that there is no spatiotemporal correlation anomaly within the current spatiotemporal window, and the monitoring continues in the next spatiotemporal window unit. If the spatiotemporal voltage fluctuation value is not less than the preset spatiotemporal voltage fluctuation value, then a potential voltage anomaly is determined to exist, and the following steps are performed: If, within the current spatiotemporal window unit, there are multiple physical nodes whose voltage monitoring values ​​deviate from the corresponding reference voltage in the same direction, and the magnitude of the deviation from the corresponding reference voltage exceeds the preset allowable range, then it is determined to be a voltage anomaly event. If, within the current spatiotemporal window unit, only a single physical node's voltage monitoring value deviates from the corresponding reference voltage by more than the preset allowable range, it is temporarily recorded as a suspected voltage disturbance, and the preset personnel are prompted to verify it in the next spatiotemporal window unit. If the results of two consecutive reviews are both abnormal, it is determined to be a voltage anomaly event. If the results of two reviews are not both abnormal, a local voltage fluctuation warning is issued.

7. The method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data as described in claim 1, characterized in that, S3 includes: The nonlinear coupling process of voltage anomaly events in the corresponding dynamic grid topology in the heterogeneous grid map is monitored. The time interval corresponding to the voltage anomaly triggering period is recorded as the anomaly association period. The anomaly association period includes the load fluctuation period associated with the voltage anomaly and the voltage regulation action period when the voltage regulating equipment is activated. The nonlinear coupling process represents the dynamic association process in which load power fluctuation and voltage regulating equipment action are transmitted step by step in the grid topology link and affect the stability of the bus voltage. During periods of load fluctuation, the load fluctuation slope per unit time is obtained by performing differential calculations on the collected load power time-series data. During the voltage regulation operation period, the corresponding voltage regulation operation amplitude is obtained by converting the voltage change of the tap corresponding to the voltage regulation device into the voltage level. By performing differential calculations on the bus voltage time series data during the abnormal correlation period, the voltage change during the abnormal correlation period is obtained. The product of the load fluctuation slope and the voltage regulation action amplitude is used as the reference value. The ratio of the absolute value of the difference between the voltage change and the reference value to the maximum value of the voltage change during the abnormal correlation period is recorded as the nonlinear interaction factor. The ratio of the absolute value of the difference between the impedance or admittance of the corresponding power supply branch of the voltage regulating equipment and the corresponding rated value to the corresponding rated value is recorded as the electrical parameter deviation rate. By co-calculating the nonlinear interaction factor and the electrical parameter deviation rate, an interaction effect judgment value reflecting the coupling strength between load fluctuation and voltage regulation action is obtained, and the root cause type corresponding to the voltage anomaly event is located.

8. The method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data as described in claim 7, characterized in that, The root cause types corresponding to the location voltage anomaly events specifically include: If the interaction effect judgment value is only in the interval of the first judgment interval that does not contain overlapping intervals, and the electrical parameter deviation rate is lower than the lower limit of the corresponding allowable deviation rate critical fluctuation interval, then it is judged as a voltage regulation response time difference caused by cross-regional coordination anomaly. If the interaction effect judgment value is only in the second judgment interval that does not contain overlapping intervals, and the electrical parameter deviation rate is not lower than the upper limit of the corresponding allowable deviation rate critical fluctuation interval, then it is determined to be electrical parameter drift caused by voltage regulating equipment abnormality. If the interaction effect judgment value is in the overlapping range of the first judgment interval and the second judgment interval, and the electrical parameter deviation rate is in the corresponding allowable deviation rate critical fluctuation range, then it is determined that the cross-regional coordination anomaly and the voltage regulation equipment anomaly coexist, and the composite voltage regulation process of cross-regional coordination-equipment linkage is executed.

9. The method for anomaly causal reasoning and root cause localization based on heterogeneous power grid data as described in claim 8, characterized in that, The aforementioned composite voltage regulation process involving cross-regional collaboration and equipment linkage is specifically as follows: The faulty component identified as an abnormal voltage regulating device is designated as the target voltage regulating node. A pre-set reference voltage regulation amount is applied to the target voltage regulating node. Simultaneously, the voltage compensation amount of adjacent voltage regulating nodes is determined based on the ratio of the difference between the electrical parameter deviation rate and the corresponding allowable deviation rate. The adjacent voltage regulating node refers to a power grid voltage regulating unit that has a direct electrical connection with the target voltage regulating node and has a voltage coupling degree higher than the corresponding preset coupling degree. The voltage coupling degree is calculated based on the impedance value and active power transmission amount between the target voltage regulating node and the adjacent voltage regulating node. Based on the linear superposition calculation result of the reference voltage regulation amount and the voltage compensation amount, the total voltage regulation amount of the corresponding target voltage regulation node and adjacent voltage regulation nodes is obtained, so as to adjust the bus side voltage. After the bus voltage is adjusted, if the newly acquired interaction effect judgment value is only in the first judgment interval that does not contain overlapping intervals, or only in the second judgment interval that does not contain overlapping intervals, the cross-regional collaborative-equipment linkage composite voltage regulation process is terminated, and the voltage regulation parameters and execution sequence are recorded. Otherwise, a voltage regulation ratio redistribution prompt is sent.

10. An anomaly causal reasoning and root cause localization system based on heterogeneous power grid data, employing the anomaly causal reasoning and root cause localization method based on heterogeneous power grid data as described in any one of claims 1-9, characterized in that, include: The graph structure dynamic update module is used to dynamically update the graph structure corresponding to the initial graph model pre-stored in the power grid dispatch center based on the heterogeneous power grid data accessed in real time by the power grid dispatch center, so that the graph structure keeps synchronized with the real-time power grid topology. The voltage anomaly event identification module is used to form a heterogeneous power grid map with integrated structure and attribute updates based on dynamically updated node attributes and edge attributes, and to perform spatiotemporal correlation analysis between the node attributes in the heterogeneous power grid map and the global voltage time series data to identify voltage anomaly events. The abnormal root cause localization module is used to quantitatively analyze the interaction effect between load fluctuations and voltage regulation actions based on the nonlinear coupling relationship of the voltage anomaly event in the dynamic power grid topology, thereby locating the root cause type corresponding to the voltage anomaly event.

Images