Emergency strategy generation method and system for distribution network disasters

Abstract

This invention provides a method and system for generating emergency strategies for distribution network disasters. It acquires multimodal sensing data of the distribution network operating environment, including equipment status sensing sequences, environmental disturbance sensing sequences, and load response sensing sequences. The method then models disaster evolution characteristics of the multimodal sensing data to generate a disaster evolution feature stream. Based on the disaster evolution feature stream, it constructs a strategy generation knowledge graph. Through the strategy generation knowledge graph, it performs strategy reasoning to generate an emergency strategy candidate stream. Finally, it performs collaborative verification of the emergency strategy candidate stream to generate a distribution network disaster emergency strategy. This invention can improve the overall effectiveness and feasibility of emergency strategies, thus providing more scientific and reliable strategy support for distribution network disaster response.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method and system for generating emergency strategies for disasters in power distribution networks. Background Technology

[0002] Currently, emergency response strategies for power distribution network disasters are generated by acquiring equipment status data or environmental data, extracting equipment anomaly thresholds or environmental indicators, and then generating emergency strategies based on a preset rule base or simple mapping table. This method is prone to one-sided information and fails to reflect the dynamic correlation of disaster development. Simple rule matching cannot adapt to changes in different disaster scenarios, and the generated emergency strategies are insufficient in terms of scientificity, adaptability, and feasibility of implementation, making it difficult to meet the actual needs of responding to complex power distribution network disasters. Summary of the Invention

[0003] This invention provides a method and system for generating emergency strategies for disasters in power distribution networks.

[0004] In a first aspect, embodiments of the present invention provide a method for generating emergency strategies for power distribution network disasters. The method includes: acquiring multimodal sensing data of the power distribution network operating environment, the multimodal sensing data including equipment status sensing sequences, environmental disturbance sensing sequences, and load response sensing sequences; performing disaster evolution feature modeling on the multimodal sensing data to generate a disaster evolution feature stream; constructing a strategy generation knowledge graph based on the disaster evolution feature stream; performing strategy reasoning through the strategy generation knowledge graph to generate an emergency strategy candidate stream; and performing collaborative verification on the emergency strategy candidate stream to generate a power distribution network disaster emergency strategy.

[0005] Secondly, embodiments of the present invention provide a computer system, comprising: a memory storing a computer program; and a processor for loading the computer program to implement the emergency strategy generation method for power distribution network disasters as described above.

[0006] The emergency strategy generation method for power distribution network disasters provided by this invention integrates three types of sensing sequences—equipment status, environmental disturbance, and load response—by acquiring multimodal sensing data. This provides a more comprehensive input foundation for disaster analysis and avoids the information bias caused by traditional methods relying on only a single data type. By modeling disaster evolution features, it generates a disaster evolution feature stream containing equipment status degradation correlation features, environmental disturbance coupling features, and load response sensitivity features. This transforms traditional static features into dynamic correlation features reflecting temporal synchronicity, spatial overlap, and causal dependence, better reflecting the dynamic correlation essence of disaster development and enhancing the depth of understanding of disaster causes. Furthermore, by constructing a strategy generation knowledge graph, it presents the information in a structured manner. This approach identifies disaster-specific entities, strategy entities, and their relationships, overcoming the limitations of traditional rule bases or simple mapping tables that can only handle single relationships. It provides a more flexible and scalable knowledge base for strategy generation. By using knowledge graphs for strategy reasoning to generate emergency strategy candidate flows, it can dynamically match strategies based on real-time disaster characteristics, adapting to changes in different disaster scenarios and avoiding the rigidity of traditional fixed rule matching. Through collaborative verification of candidate strategy flows, it comprehensively considers the synergy between strategies in terms of resource consumption, execution time, and effect objectives, avoiding strategy conflicts that may result from verifying a single strategy. This improves the overall effectiveness and feasibility of emergency strategies, thus providing more scientific and reliable strategy support for distribution network disaster response. Attached Figure Description

[0007] Figure 1 This is a flowchart of an emergency strategy generation method for power distribution network disasters provided by an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of the composition of a computer system provided in an embodiment of the present invention. Detailed Implementation

[0009] Please see Figure 1 , Figure 1 A flowchart illustrating an emergency strategy generation method for power distribution network disasters, provided as an embodiment of the present invention, is included. This method can be executed by a computer system and includes the following steps:

[0010] Step S100: Obtain multimodal sensing data of the power distribution network operating environment. The multimodal sensing data includes equipment status sensing sequence, environmental disturbance sensing sequence and load response sensing sequence.

[0011] Multimodal sensing data is data acquired from multiple dimensions and types of sensors, comprehensively reflecting different aspects of the distribution network's operating environment. Equipment status sensing sequences are a series of data about the operating status of equipment in the distribution network, reflecting the various parameters of the equipment at different times, such as the changes in voltage, current, and temperature over time. Environmental disturbance sensing sequences record data sequences of various disturbance factors in the environment in which the distribution network is located, which may affect it. These disturbance factors may include the changes in environmental factors such as wind speed, precipitation, and thunderstorms over time. Load response sensing sequences describe the changes in the load response of the distribution network under different conditions, covering relevant data of different states such as load growth, stability, and reduction, such as information on load power changes and load fluctuation areas. For equipment status sensing sequences, voltage sensors can be used to collect voltage data from equipment in real time, current sensors to collect current data, and temperature sensors to collect temperature data. These sensors collect data at certain time intervals to form the equipment status sensing sequence. For environmental disturbance sensing sequences, anemometers can measure wind speed, rain gauges can measure precipitation, and lightning monitoring equipment can monitor thunderstorm conditions. Data is also recorded at fixed time intervals to obtain the environmental disturbance sensing sequence. For load response sensing sequences, power sensors can be used to monitor changes in load power. By monitoring loads in different areas, load fluctuation areas can be determined, thereby forming a load response sensing sequence.

[0012] Step S200: Model the disaster evolution characteristics of the multimodal sensing data and generate a disaster evolution characteristic stream.

[0013] Disaster evolution feature modeling involves analyzing and processing multimodal sensing data to extract features related to the evolution of disasters in the distribution network. The disaster evolution feature stream integrates these extracted features according to certain temporal relationships, forming a data stream that reflects the characteristic changes of a distribution network disaster from its occurrence to its development. In-depth analysis of multimodal sensing data can reveal the interrelationships and patterns of change among equipment status, environmental disturbances, and load response.

[0014] As one implementation method, step S200 can be specifically implemented as the following steps S210~S240:

[0015] Step S210: Perform time-series correlation analysis on the equipment status sensing sequence and the environmental disturbance sensing sequence to generate equipment status degradation correlation features. The equipment status degradation correlation features include the time-series synchronization information of equipment parameter changes and environmental disturbances, as well as the dependence information of change trends.

[0016] Temporal correlation analysis examines the temporal relationship between equipment status perception sequences and environmental disturbance perception sequences, identifying the synchronicity and trend dependencies between them. Equipment status degradation correlation characteristics, obtained through this analysis, reflect the deterioration of equipment status due to environmental disturbances. Temporal synchronicity information reflects the degree of overlap between equipment parameter change events and environmental disturbance events on the time axis; for example, whether an abnormal voltage increase occurs simultaneously with a sudden increase in wind speed. Trend dependency information reflects the intrinsic connection between equipment parameter change trends and environmental disturbance trends; for example, whether an upward trend in equipment temperature is correlated with a decreasing trend in precipitation.

[0017] As one implementation method, step S210 involves performing a time-series correlation analysis on the equipment state sensing sequence and the environmental disturbance sensing sequence to generate equipment state degradation correlation features. Specifically, this can be implemented as follows: steps S211~S215:

[0018] Step S211: Extract the device parameter change event points from the device status sensing sequence. The device parameter change event points include abnormal voltage increase event points, abnormal current decrease event points, and abnormal temperature duration event points.

[0019] Equipment parameter change events are critical time points in the equipment status sensing sequence where abnormal changes occur in equipment parameters. An abnormal voltage increase event occurs when the equipment voltage exceeds the upper limit of its normal operating range, which may be caused by line faults, sudden load changes, etc. An abnormal current decrease event occurs when the equipment current falls below the lower limit of its normal operating range, which may be caused by equipment faults, line breaks, etc. A sustained abnormal temperature event occurs when the equipment temperature remains above the normal temperature threshold for a period of time; prolonged temperature anomalies can severely impact equipment performance and lifespan.

[0020] To extract these event points, corresponding thresholds can be set. For abnormal voltage rise events, a reasonable upper voltage threshold can be determined based on the device's rated voltage and historical operating data. When the device voltage exceeds this threshold, the time point is recorded as an abnormal voltage rise event. For abnormal current drop events, a lower current threshold is also set. When the current falls below this threshold, the corresponding time point is recorded. For abnormal temperature duration events, a temperature threshold and a duration threshold are set. When the device temperature exceeds the temperature threshold and the duration reaches the duration threshold, the time point is recorded.

[0021] Step S212: Extract environmental disturbance event points from the environmental disturbance perception sequence. These environmental disturbance event points include sudden increase in wind speed, sudden increase in precipitation, and thunderstorm occurrence.

[0022] Environmental disturbance event points are the critical time points in the environmental disturbance perception sequence where abnormal changes in environmental factors occur. A sudden increase in wind speed is the point in time when wind speed rapidly increases to exceed a certain threshold within a short period; strong winds may damage overhead lines, poles, and other equipment in the power distribution network. A sudden increase in precipitation is the point in time when precipitation increases significantly within a short period, exceeding the normal precipitation range; excessive precipitation may lead to equipment dampness, short circuits, and other problems. A thunderstorm occurrence event point is the point in time when a thunderstorm is detected; lightning generated by a thunderstorm may cause direct lightning strike damage to power distribution network equipment.

[0023] To extract these event points, appropriate thresholds need to be set based on environmental monitoring data. For events involving a sudden increase in wind speed, a threshold for wind speed increase can be determined based on local climate conditions and historical wind speed data. When the wind speed increases beyond this threshold within a short period, the time point is recorded. For events involving a sudden increase in precipitation, a threshold for precipitation increase is set. When the increase in precipitation exceeds this threshold within a short period, the time point is recorded. For events involving thunderstorms, lightning signals can be monitored in real time using lightning monitoring equipment. When a lightning signal is detected, the time point is recorded.

[0024] Step S213: Timestamp align the device parameter change event points with the environmental disturbance event points to obtain timing synchronization information. The timing synchronization information includes the overlap ratio of device parameter change events and environmental disturbance events on the time axis.

[0025] Timestamp alignment arranges device parameter change events and environmental disturbance events in chronological order, allowing them to be compared on the same timeline. Timing synchronization information measures the degree of synchronization between device parameter change events and environmental disturbance events by calculating the overlap ratio on the timeline. A higher overlap ratio indicates a greater likelihood of a correlation between the two events.

[0026] In practice, the timestamps of device parameter change events and environmental disturbance events can be converted into a unified time format. Then, these event points can be iterated through to count the number of times that device parameter change events and environmental disturbance events occur simultaneously within a certain time window. The result is then divided by the total number of events to obtain the overlap ratio.

[0027] Step S214: Perform a dependency analysis on the changing trend of equipment parameter change event points and the disturbance trend of environmental disturbance event points to obtain the dependency information of the changing trend. The dependency information of the changing trend includes the covariance information of the upward trend of equipment parameters and the strengthening trend of environmental disturbance.

[0028] Dependency analysis studies the intrinsic relationship between the changing trends of equipment parameters and the trends of environmental disturbances, measuring the degree of dependence between the two by calculating the covariance. Covariance is a statistic indicating whether the changing trends of two variables are consistent. A positive covariance indicates a positive correlation between the upward trend of equipment parameters and the increasing trend of environmental disturbances, meaning their changing trends are consistent; a negative covariance indicates opposite trends; and a covariance close to zero indicates almost no correlation between the two.

[0029] To calculate covariance, it is first necessary to quantify the trends of change in equipment parameters and environmental disturbances. These trends can be represented by calculating the rate of change of parameters between adjacent event points. For example, for equipment parameter change event points, calculate the rate of voltage increase between adjacent abnormal voltage increase event points; for environmental disturbance event points, calculate the rate of wind speed increase between adjacent sudden wind speed increase event points. Then, these rates of change are used as variables, and the covariance calculation formula is applied.

[0030] Step S215: Combine the timing synchronization information with the dependence information of the change trend into equipment condition degradation correlation features.

[0031] By combining temporal synchronization information and trend dependence information, a comprehensive feature is formed: the equipment condition degradation correlation feature. This feature can more comprehensively reflect the degradation of equipment condition due to environmental disturbances. By combining these two aspects of information, it is possible to more accurately determine whether changes in equipment condition are related to environmental disturbances, and the strength of this correlation.

[0032] Step S220: Perform spatial coupling analysis on the environmental disturbance sensing sequence and the load response sensing sequence to generate environmental disturbance coupling features. The environmental disturbance coupling features include spatial overlap information between the environmental disturbance region and the load fluctuation region, as well as gradient correspondence information between the disturbance intensity and the load fluctuation amplitude.

[0033] Spatial coupling analysis examines the spatial correlation between environmental disturbance sensing sequences and load response sensing sequences, investigating the spatial overlap between disturbance and load fluctuation regions, as well as the gradient correspondence between disturbance intensity and load fluctuation amplitude. Environmental disturbance coupling characteristics, obtained through this analysis, reflect the spatial correlation between environmental disturbances and load responses. Spatial overlap information indicates the degree of spatial overlap between disturbance and load fluctuation regions, such as whether high-wind-speed areas geographically overlap with high-load-growth areas. The gradient correspondence between disturbance intensity and load fluctuation amplitude reflects the relationship between changes in disturbance intensity and load fluctuation amplitude, such as the relationship between the attenuation gradient of wind speed from the disturbance center to the boundary and the decreasing gradient of load growth from the load center to the boundary.

[0034] As one implementation method, step S220 involves performing spatial coupling analysis on the environmental disturbance sensing sequence and the load response sensing sequence to generate environmental disturbance coupling features. Specifically, this can be implemented as follows: steps S221 to S227:

[0035] Step S221: Extract the disturbance area distribution information from the environmental disturbance sensing sequence. The disturbance area distribution information includes the geographical location information of the high wind speed area, the coverage information of the heavy precipitation area, and the boundary coordinate information of the high frequency thunderstorm area.

[0036] Disturbance area distribution information describes the spatial distribution of environmental disturbances, including the specific location and extent of different types of environmental disturbance areas. Geographical location information for high-wind-speed areas clarifies the specific geographical location of areas where wind speeds exceed a certain threshold, and can be represented by geographic coordinates. Coverage information for heavy precipitation areas describes the size and shape of areas where precipitation exceeds a certain standard, and can be represented by area, boundary coordinates, etc. Boundary coordinate information for high-frequency thunderstorm areas determines the boundary location of areas with a high frequency of thunderstorms, defining the area through a series of coordinate points.

[0037] To extract this information, Geographic Information Systems (GIS) and environmental monitoring data can be utilized. For the geographic location information of high-wind-speed areas, data collected by anemometers can be imported into the GIS system, high-wind-speed areas can be delineated based on wind speed thresholds, and their geographic coordinates can be obtained. For the coverage information of heavy precipitation areas, the boundaries of heavy precipitation areas can be drawn in the GIS system using rain gauge distribution data and precipitation data, and their coverage area can be calculated. For the boundary coordinate information of high-frequency thunderstorm areas, lightning occurrence location data recorded by lightning monitoring equipment can be used to statistically analyze thunderstorm occurrence frequencies in the GIS system, delineate high-frequency thunderstorm areas, and obtain their boundary coordinates.

[0038] Step S222: Extract the load fluctuation area distribution information from the load response sensing sequence. The load fluctuation area distribution information includes the geographical location information of the high load growth area, the coverage information of the load drop area, and the boundary coordinate information of the load oscillation area.

[0039] Load fluctuation area distribution information describes the spatial fluctuations of the distribution network load, including the specific location and range of different types of load fluctuation areas. Geographical location information for high load growth areas clarifies the specific geographical location of areas where the load growth rate exceeds a certain threshold, and can be represented by geographic coordinates. Coverage information for load drop areas describes the size and shape of the area where the load suddenly drops, and can be represented by area, boundary coordinates, etc. Boundary coordinate information for load oscillation areas determines the boundary location of areas where the load fluctuates significantly within a certain range, and is defined by a series of coordinate points. To extract this information, power load monitoring data and a GIS system can be combined. Data collected by power sensors at each load node is imported into the GIS system, and high load growth areas, load drop areas, and load oscillation areas are divided according to thresholds such as load growth rate, load drop magnitude, and load fluctuation range. Then, the geographic location information, coverage information, and boundary coordinate information of these areas are obtained.

[0040] Step S223: Perform spatial overlap calculation on the distribution information of the disturbance area and the distribution information of the load fluctuation area to obtain spatial overlap information. The spatial overlap information includes the intersection area information of the high wind speed area and the high load growth area.

[0041] Spatial overlap calculation involves determining the overlapping portion of disturbance and load fluctuation regions in geographic space. Spatial overlap information measures the degree of spatial overlap between environmental disturbance and load fluctuation regions by calculating indicators such as the intersection area of ​​these regions. The intersection area information between high wind speed regions and high load growth regions is an important component of spatial overlap information, reflecting the specific situation of high load growth regions under high wind speed conditions.

[0042] In practice, the distribution information of disturbance areas and load fluctuation areas can be overlaid and analyzed in a GIS system. Geographic data of high wind speed areas and high load growth areas are imported into the GIS system, and their intersection area is calculated using the GIS's spatial analysis functions.

[0043] Step S224: Extract the disturbance intensity gradient information from the environmental disturbance sensing sequence. The disturbance intensity gradient information includes the wind speed attenuation gradient, precipitation decrease gradient, and thunderstorm frequency decrease gradient from the disturbance center to the boundary.

[0044] Disturbance intensity gradient information describes the spatial variation of environmental disturbance intensity, reflecting the trend of intensity change from the center to the boundary. Wind speed attenuation gradient indicates the degree of wind speed reduction from the center to the boundary of a high-wind-speed area, which can be represented by calculating the ratio of the wind speed difference between adjacent locations to their distance. Precipitation decrease gradient indicates the degree of precipitation reduction from the center to the boundary of a heavy precipitation area, also represented by calculating the ratio of the precipitation difference between adjacent locations to their distance. Thunderstorm frequency decrease gradient indicates the degree of thunderstorm frequency decrease from the center to the boundary of a high-frequency thunderstorm area, represented by calculating the ratio of the thunderstorm frequency difference between adjacent locations to their distance.

[0045] To extract this information, multiple monitoring points can be set up within the environmental disturbance area to collect environmental data at different locations. For wind speed attenuation gradients, wind speed monitoring points are set up at regular intervals within high-wind-speed areas to record wind speed data at each monitoring point. Then, the ratio of the wind speed difference between adjacent monitoring points to the distance is calculated to obtain the wind speed attenuation gradient. For precipitation decrease gradients, rain gauges are reasonably distributed within areas of heavy precipitation to record precipitation data at different locations. The ratio of the precipitation difference between adjacent locations to the distance is calculated. For thunderstorm frequency decrease gradients, lightning monitoring equipment is used to monitor areas with high-frequency thunderstorms, and the frequency of thunderstorms at different locations is statistically analyzed. The ratio of the thunderstorm frequency difference between adjacent locations to the distance is calculated.

[0046] Step S225: Extract the load fluctuation amplitude gradient information from the load response sensing sequence. The load fluctuation amplitude gradient information includes the decreasing gradient of load increase and the increasing gradient of load decrease from the load center to the boundary.

[0047] Load fluctuation gradient information describes the spatial variation of load fluctuation amplitude in a distribution network, reflecting the trend of load fluctuation amplitude change from the center to the boundary. The load growth decrease gradient indicates the degree to which the load growth gradually decreases from the center to the boundary of a high load growth area, which can be represented by calculating the ratio of the difference in load growth between adjacent locations to the distance. The load decrease increase gradient indicates the degree to which the load decrease gradually increases from the center to the boundary of a sudden load drop area, similarly represented by calculating the ratio of the difference in load decrease between adjacent locations to the distance.

[0048] To extract this information, multiple load monitoring points need to be set up within the load fluctuation area. In areas of high load growth, a load monitoring point is set up at regular intervals to record load power data at different locations. The ratio of the load growth difference between adjacent monitoring points to the distance is calculated to obtain the load growth decrease gradient. Similarly, in areas of sudden load drop, load monitoring points are set up to record load power data, and the ratio of the load reduction difference between adjacent locations to the distance is calculated to obtain the load reduction increase gradient.

[0049] Step S226: Perform gradient correspondence analysis on the disturbance intensity gradient information and the load fluctuation amplitude gradient information to obtain gradient correspondence information, which includes the linear fitting relationship information between the wind speed attenuation gradient and the load growth decrease gradient.

[0050] Gradient correspondence analysis studies the intrinsic relationship between the intensity gradient of environmental disturbances and the amplitude gradient of load fluctuations. By analyzing their correspondence, it identifies the patterns between them. Linear fitting information is an important form of gradient correspondence information, indicating whether a linear relationship exists between the wind speed decay gradient and the load growth decrease gradient, and the specific parameters of this linear relationship.

[0051] To perform gradient correspondence analysis, the extracted disturbance intensity gradient information and load fluctuation amplitude gradient information can be processed. Using the wind speed attenuation gradient and load increase decrease gradient data as two variables, a linear regression algorithm is used for fitting analysis. First, the wind speed attenuation gradient and load increase decrease gradient data are organized into a dataset. Then, a linear regression model is used to train the dataset to obtain a linear fitting equation.

[0052] Step S227: Combine spatial overlap information with gradient correspondence information to form environmental perturbation coupling features.

[0053] By combining spatial overlap information and gradient correspondence information, a comprehensive feature is formed: the environmental disturbance coupling feature. This feature can more comprehensively reflect the spatial correlation and gradient correspondence between environmental disturbances and load response. By combining these two aspects of information, the degree and manner of the impact of environmental disturbances on load fluctuations can be determined more accurately.

[0054] Step S230: Perform causal dependency analysis on the load response sensing sequence and the equipment status sensing sequence to generate load response sensitive features. The load response sensitive features include causal chain information of load change events and equipment status change events, as well as the transmission path information of the impact of load change on equipment status.

[0055] Causal dependency analysis studies the causal relationship between load response sensing sequences and equipment state sensing sequences, identifying the intrinsic connections between load change events and equipment state change events, and how load changes affect equipment state. Causal chain information describes the causal relationship between load change events and equipment state change events, clarifying which load change events directly or indirectly lead to which equipment state change events. Influence propagation path information describes how and through what paths load changes affect equipment state; for example, how an increase in load is propagated to equipment temperature via current parameters.

[0056] To perform causal dependency analysis, data mining and machine learning algorithms can be employed. First, the load response sensing sequence and equipment state sensing sequence are preprocessed to extract key events and parameters. Then, causal relationship mining algorithms, such as the Granger causality test, are used to analyze the causal relationships between load change events and equipment state change events, obtaining causal chain information. For influence propagation path information, a path analysis model can be constructed, combining the physical principles and electrical characteristics of the equipment, to analyze the impact path of load changes on equipment state.

[0057] As one implementation method, step S230 involves performing causal dependency analysis on the load response sensing sequence and the device status sensing sequence to generate load response sensitive features. Specifically, this can be implemented as follows: steps S231 to S237:

[0058] Step S231: Extract the load change event chain from the load response perception sequence. The load change event chain includes load increase trigger event, load stability maintenance event, and load decrease termination event.

[0059] A load change event chain is a combination of events describing load changes at different stages. A load increase trigger event is the initial event that marks the beginning of a load increase, which may be caused by new load connections, production equipment startup, etc. A load stabilization event is an event where the load remains relatively stable for a period of time, during which the load power change is small. A load decrease termination event is an event where the load begins to decrease and eventually stops decreasing, which may be caused by load equipment shutdown, production completion, etc.

[0060] To extract the load change event chain, event detection can be performed on the load response sensing sequence. Load increase trigger events and load decrease termination events are determined by setting a threshold for load power change. When the load power increases by more than a certain threshold within a short period, it is determined as a load increase trigger event; when the load power decreases by more than a certain threshold within a short period, it is determined as a load decrease termination event. For load stability maintenance events, it can be determined by whether the change in load power over a period of time is less than a small threshold.

[0061] Step S232: Extract the equipment status change event chain from the equipment status perception sequence. The equipment status change event chain includes normal equipment operation events, equipment abnormality warning events, and equipment failure shutdown events.

[0062] A device state change event chain describes a series of events that transition a device between different states. A normal operation event occurs when the device is in a normal working state, and all its parameters are within normal ranges. A device anomaly warning event occurs when some parameters of the device become abnormal, but have not yet reached a fault state; timely measures are needed to prevent further failure. A device failure shutdown event occurs when the device stops operating due to a fault; in this case, the device cannot function normally and requires repair or replacement.

[0063] To extract the event chain of equipment state changes, event detection can be performed on the equipment state perception sequence. Normal operation events, equipment anomaly warning events, and equipment failure shutdown events are determined by setting normal ranges and abnormal thresholds for equipment parameters. When all equipment parameters are within the normal range, it is determined to be a normal operation event; when some parameters exceed the normal range but do not reach the fault threshold, it is determined to be an equipment anomaly warning event; when some parameters reach the fault threshold, it is determined to be an equipment failure shutdown event.

[0064] Step S233: Perform causal relationship mining on the load change event chain and the equipment status change event chain to obtain causal chain information. The causal chain information includes direct causal relationship information from the load increase trigger event to the equipment abnormality warning event and indirect causal relationship information from the load stability maintenance event to the equipment normal operation event.

[0065] Causal relationship mining is the process of identifying the causal relationships between load change event chains and equipment state change event chains. Direct causal relationship information clarifies the specific circumstances and probability of a load increase triggering an event that directly leads to an equipment malfunction warning event. Indirect causal relationship information describes how a load stabilization event indirectly leads to a normal equipment operation event through some intermediate link.

[0066] To uncover causal relationships, causal analysis algorithms such as causal graph models and Bayesian networks can be used. First, the data on load change event chains and equipment state change event chains are organized into a dataset. Then, a causal analysis model is trained on the dataset to identify the causal relationships within it.

[0067] Step S234: Extract load change information from the load response sensing sequence. The load change information includes the absolute value of the load increase, the absolute value of the load decrease, and the relative proportion of the load change.

[0068] Load change information describes the degree of load change, reflecting the load variation from different perspectives. The absolute value of the load increase is the specific power increase during the increase, directly reflecting the scale of the load increase. The absolute value of the load decrease is the specific power decrease during the decrease, reflecting the degree of load reduction. The relative proportion of the load change is the ratio of the load change to the initial load value, allowing for a more intuitive comparison of the degree of change in different loads.

[0069] To extract information about load changes, the load response sensing sequence can be processed. The absolute values ​​of load increases and decreases are obtained by calculating the differences between adjacent load data points.

[0070] Step S235: Extract the equipment status impact information from the equipment status perception sequence. The equipment status impact information includes the degree of impact of load changes on equipment voltage, the degree of impact on equipment current, and the degree of impact on equipment temperature.

[0071] Equipment status impact information describes the specific effects of load changes on equipment status, reflecting the impact of load changes on the equipment from different perspectives. The impact of load changes on equipment voltage indicates the specific numerical or proportional change in equipment voltage caused by the load change, which can be obtained by measuring the difference or percentage change in equipment voltage before and after the load change. Similarly, the impact of load changes on equipment current indicates the specific numerical or proportional change in equipment current caused by the load change, which can also be obtained by measuring the difference or percentage change in equipment current before and after the load change. Finally, the impact of load changes on equipment temperature indicates the specific numerical value or trend of temperature change caused by the load change, which can be determined by monitoring the temperature changes of the equipment before and after the load change using a temperature sensor.

[0072] Step S236: Perform transmission path analysis on the load change information and the equipment status impact information to obtain the impact transmission path information, which includes the path information of the load increase being transmitted to the equipment temperature through the current parameter.

[0073] Transmission path analysis is the process of studying how load changes affect equipment status through various intermediate parameters and links. Transmission path information clarifies the specific paths and methods by which load changes influence equipment status. The path information of the load increase being transmitted to equipment temperature via current parameters is an important aspect of transmission path information, describing how the load increase affects the current and thus the equipment temperature.

[0074] To perform transmission path analysis, we can combine circuit principles and the physical characteristics of the equipment. First, we analyze the relationship between the increase in load and the current. According to Ohm's law, an increase in load will lead to an increase in current. Then, we analyze the effect of the increased current on the equipment temperature. Current flowing through the resistance of the equipment generates heat. According to Joule's law, as the current increases, the heat generated by the equipment will increase, thus causing the equipment temperature to rise.

[0075] Step S237: Combine the causal chain information with the influence propagation path information into a load response sensitive feature.

[0076] By combining causal chain information and impact propagation path information, a comprehensive feature is formed: the load response sensitivity feature. This feature can more comprehensively reflect the impact of load changes on equipment status, including the causal relationship between load change events and equipment status change events, as well as the impact path of the load change on the equipment status. By combining these two aspects of information, the degree and manner of the impact of load changes on the equipment can be determined more accurately.

[0077] Step S240: Perform feature time series fusion of equipment condition degradation correlation features, environmental disturbance coupling features and load response sensitivity features to generate disaster evolution feature stream.

[0078] Feature time-series fusion integrates different types of features in chronological order, enabling them to correlate and complement each other over time, forming a continuous feature stream that reflects the evolution of a disaster. The disaster evolution feature stream is a comprehensive feature data stream obtained through this fusion, containing information on changes in equipment status, environmental disturbances, and load response over time, and can more comprehensively reflect the entire process of a distribution network disaster from its occurrence to its development.

[0079] As one implementation method, step S240 involves fusing the equipment condition degradation correlation characteristics, environmental disturbance coupling characteristics, and load response sensitivity characteristics through feature time series fusion to generate a disaster evolution feature stream. Specifically, this can be implemented as follows: steps S241 to S244:

[0080] Step S241: Extract the temporal synchronization information from the equipment condition deterioration correlation features, determine the time weight coefficient of the equipment features, and the time weight coefficient is positively correlated with the overlap ratio of equipment parameter change events and environmental disturbance events.

[0081] Temporal synchronization information is an important component of the correlation characteristics of equipment condition deterioration, reflecting the degree of temporal synchronization between equipment parameter change events and environmental disturbance events. The time weighting coefficient is a coefficient used to measure the importance of equipment characteristics in the disaster evolution process. It is positively correlated with the overlap ratio of equipment parameter change events and environmental disturbance events; that is, the higher the overlap ratio, the larger the time weighting coefficient.

[0082] To determine the time weighting coefficients for device features, they can be calculated based on the overlap ratio in the timing synchronization information. First, the overlap ratio in the extracted timing synchronization information is normalized so that its value ranges from 0 to 1. Then, the normalized overlap ratio is used as the time weighting coefficient.

[0083] Step S242: Extract spatial overlap information from the environmental disturbance coupling features, determine the spatial weight coefficient of the environmental features, and the spatial weight coefficient is positively correlated with the intersection area of ​​the environmental disturbance region and the load fluctuation region.

[0084] Spatial overlap information is a key feature in the coupling characteristics of environmental disturbances, reflecting the degree of spatial overlap between the environmental disturbance region and the load fluctuation region. The spatial weighting coefficient is a coefficient used to measure the importance of environmental features in the disaster evolution process. It is positively correlated with the intersection area of ​​the environmental disturbance region and the load fluctuation region; that is, the larger the intersection area, the larger the spatial weighting coefficient. To determine the spatial weighting coefficient of environmental features, it can be calculated based on the intersection area in the spatial overlap information. First, the intersection area extracted from the spatial overlap information is normalized so that its value ranges between 0 and 1. Then, the normalized intersection area is used as the spatial weighting coefficient.

[0085] Step S243: Extract causal chain information from the load response sensitive features and determine the causal weight coefficient of the load features. The higher the clarity of the direct causal relationship of the causal chain information, the larger the causal weight coefficient.

[0086] Causal chain information is an important component of load response sensitivity characteristics, reflecting the causal relationship between load change events and equipment state change events. The causal weight coefficient is a coefficient used to measure the importance of load characteristics in the disaster evolution process. It is positively correlated with the clarity of the direct causal relationship in the causal chain information; that is, the clearer the direct causal relationship, the larger the causal weight coefficient.

[0087] To determine the causal weight coefficients for load characteristics, causal chain information can be quantitatively analyzed. A direct causal relationship clarity index can be defined to measure the clarity of direct causal relationships in the causal chain information. For example, the direct causal relationship clarity index can be calculated based on factors such as the number and strength of direct causal relationships in the causal chain information. Then, the direct causal relationship clarity index is normalized to a value between 0 and 1, and the normalized index is used as the causal weight coefficient.

[0088] Step S244: Based on the time weight coefficient, spatial weight coefficient and causal weight coefficient, the equipment condition degradation correlation characteristics, environmental disturbance coupling characteristics and load response sensitivity characteristics are weighted and time-series superimposed to generate a disaster evolution feature stream with dynamic weights. The dynamic weights are used to characterize the degree of dominance of different features in the disaster evolution process.

[0089] Weighted time-series superposition combines equipment condition degradation correlation characteristics, environmental disturbance coupling characteristics, and load response sensitivity characteristics in chronological order, and weights them according to time weighting coefficients, spatial weighting coefficients, and causal weighting coefficients. Dynamic weighting, on the other hand, reflects the changing weights of different characteristics during disaster evolution, as these weights change with time and disaster conditions, thus more accurately reflecting the dominant role of different characteristics in the disaster evolution process.

[0090] In practice, the equipment condition degradation correlation features, environmental disturbance coupling features, and load response sensitivity features are first arranged in chronological order. Then, each feature is weighted according to time weighting coefficients, spatial weighting coefficients, and causal weighting coefficients. For example, for the equipment condition degradation correlation features, the feature value at each time point is multiplied by the time weighting coefficient; for the environmental disturbance coupling features, the feature value at each time point is multiplied by the spatial weighting coefficient; and for the load response sensitivity features, the feature value at each time point is multiplied by the causal weighting coefficient. Finally, the weighted features are superimposed to obtain a disaster evolution feature stream with dynamic weights.

[0091] Step S300: Generate a knowledge graph based on a strategy for constructing disaster evolution feature flows.

[0092] A strategy generation knowledge graph is a graph structure used to represent the knowledge and relationships related to disaster emergency response strategies in power distribution networks. It associates and integrates various feature information from the disaster evolution feature flow with corresponding emergency strategies, enabling the rapid and accurate generation of appropriate emergency strategies when a disaster occurs. By constructing a strategy generation knowledge graph, complex disaster features and strategy information can be visualized and structured.

[0093] As one implementation method, step S300, generating a knowledge graph based on the disaster evolution feature flow construction strategy, can be specifically implemented as follows: steps S310~S370:

[0094] Step S310: Extract equipment condition deterioration correlation features from the disaster evolution feature stream, and define the temporal synchronization information and the dependence information of the change trend in the equipment condition deterioration correlation features as the first type of disaster feature entity.

[0095] Equipment condition deterioration correlation features are an important component of the disaster evolution feature stream, reflecting the deterioration of equipment condition due to environmental disturbances. Temporal synchronicity information and trend dependence information are key information in equipment condition deterioration correlation features; these are defined as first-class disaster feature entities for representation and association in the knowledge graph.

[0096] To extract the correlation features of equipment condition deterioration, feature decomposition can be performed on the disaster evolution feature stream. By analyzing the equipment feature portion of the disaster evolution feature stream, temporal synchronization information and trend dependency information can be extracted. For example, in the disaster evolution feature stream, data segments related to equipment condition deterioration correlation features can be identified, and the overlap ratio between equipment parameter change events and environmental disturbance events (temporal synchronization information) and the covariance between the upward trend of equipment parameters and the increasing trend of environmental disturbances (trend dependency information) can be extracted from them. This information can be used as attributes of the first type of disaster feature entity.

[0097] Step S320: Extract equipment intervention strategies from the preset strategy library, define the type information and operation process information of the equipment intervention strategies as the first type of strategy entity; establish a first association relationship between the first type of disaster characteristic entity and the first type of strategy entity.

[0098] The preset strategy library is a database that pre-stores various emergency strategies for power distribution network disasters. This includes equipment intervention strategies, environmental buffering strategies, and load regulation strategies for different disaster situations. Equipment intervention strategies are used to address equipment degradation issues. Their type information and operational process information are defined as first-class strategy entities for representation and association in the knowledge graph.

[0099] To extract equipment intervention strategies, strategies related to equipment condition deterioration can be selected from a pre-defined strategy library. Examples include preventative maintenance strategies and emergency isolation strategies. The type information (e.g., preventative maintenance, emergency isolation) and operational process information (e.g., maintenance scope determination rules, isolation operation priority rules) of these strategies are used as attributes of the first type of strategy entity. Then, a first association is established between the first type of disaster characteristic entity and the first type of strategy entity. This association represents the connection between equipment condition deterioration and the corresponding equipment intervention strategy. For example, when the overlap between equipment parameter change events and environmental disturbance events is high, a preventative maintenance strategy may be necessary; when the covariance between the upward trend of equipment parameters and the increasing trend of environmental disturbance is large, an emergency isolation strategy may be necessary.

[0100] Step S330: Extract environmental disturbance coupling features from the disaster evolution feature flow, and define the spatial overlap information and gradient correspondence information in the environmental disturbance coupling features as the second type of disaster feature entity.

[0101] Environmental disturbance coupling features are another important component of the disaster evolution feature flow, reflecting the spatial correlation and gradient correspondence between environmental disturbances and load response. Spatial overlap information and gradient correspondence information are key information in environmental disturbance coupling features, and they are defined as second-class disaster feature entities for representation and association in the knowledge graph.

[0102] To extract environmental disturbance coupling features, feature decomposition can be performed on the disaster evolution feature stream. By analyzing the environmental feature portion of the disaster evolution feature stream, spatial overlap information and gradient correspondence information can be extracted. For example, in the disaster evolution feature stream, data segments related to environmental disturbance coupling features are found, and the intersection area of ​​the environmental disturbance region and the load fluctuation region (spatial overlap information) and the linear fitting relationship between the wind speed attenuation gradient and the load growth decrease gradient (gradient correspondence information) are extracted from them. This information is used as attributes of the second type of disaster feature entity.

[0103] Step S340: Extract environmental buffering strategies from the preset strategy library, define the type information and implementation requirement information of the environmental buffering strategies as second-class strategy entities; establish a second association relationship between the second-class disaster characteristic entities and the second-class strategy entities.

[0104] Environmental buffering strategies are strategies used to address the impact of environmental disturbances on the distribution network. Their type information and implementation requirements information are defined as second-class strategy entities so that they can be represented and associated in the knowledge graph.

[0105] To extract environmental buffering strategies, strategies related to environmental disturbances can be selected from a pre-defined strategy library. Examples include regional joint protection strategies and gradient-level protection strategies. The type information of these strategies (e.g., regional joint protection, gradient-level protection) and implementation requirements (e.g., rules for coordinated allocation of protection resources, rules for adjusting protection intensity according to gradient changes) are used as attributes of the second type of strategy entities. Then, a second association is established between the second type of disaster characteristic entities and the second type of strategy entities. This association represents the connection between the environmental disturbance situation and the corresponding environmental buffering strategy. For example, when the intersection area of ​​the environmental disturbance area and the load fluctuation area is large, a regional joint protection strategy may be necessary; when the linear fit between the wind speed attenuation gradient and the load increase decrease gradient is good, a gradient-level protection strategy may be necessary.

[0106] Step S350: Extract load response sensitive features from the disaster evolution feature stream, and define the causal chain information and impact transmission path information in the load response sensitive features as the third type of disaster feature entity.

[0107] Load response sensitivity features are another important component of the disaster evolution feature flow, reflecting the impact of load changes on equipment status. Causal chain information and impact transmission path information are key information in load response sensitivity features, and they are defined as the third type of disaster feature entity for representation and association in the knowledge graph.

[0108] To extract load response sensitivity features, feature decomposition can be performed on the disaster evolution feature flow. By analyzing the load feature portion of the disaster evolution feature flow, causal chain information and impact transmission path information can be extracted. For example, in the disaster evolution feature flow, data segments related to load response sensitivity features are identified, and causal chain information between load change events and equipment state change events, as well as path information of load increase being transmitted to equipment temperature through current parameters, are extracted from them. This information is then used as attributes of the third type of disaster feature entity.

[0109] Step S360: Extract load adjustment strategies from the preset strategy library, define the type information and execution rule information of the load adjustment strategies as third-class strategy entities; establish a third association relationship between the third-class disaster characteristic entities and the third-class strategy entities.

[0110] Load regulation strategies are strategies used to address the impact of load changes on the distribution network. Their type information and execution rule information are defined as third-class strategy entities so that they can be represented and associated in the knowledge graph.

[0111] To extract load regulation strategies, strategies related to load changes can be filtered from a pre-defined strategy library. Examples include dynamic load transfer strategies and tiered load control strategies. The type information of these strategies (e.g., dynamic load transfer, tiered load control) and execution rule information (e.g., rules for selecting load transfer types, rules for classifying load control levels) are used as attributes of a third type of strategy entity. Then, a third association is established between the third type of disaster characteristic entity and the third type of strategy entity. This association represents the connection between load changes and the corresponding load regulation strategies. For example, when the direct causal relationship between a load change event and an equipment state change event is clear, a dynamic load transfer strategy may be necessary; when the path from the increase in load to the equipment temperature via current parameters is clear, a tiered load control strategy may be necessary.

[0112] Step S370: Combine the first type of disaster feature entity, the first type of strategy entity and the first association, the second type of disaster feature entity, the second type of strategy entity and the second association, and the third type of disaster feature entity, the third type of strategy entity and the third association into a strategy generation knowledge graph.

[0113] By combining the various disaster characteristic entities, strategy entities, and relationships defined above, a complete strategy generation knowledge graph is formed. In this knowledge graph, the first, second, and third types of disaster characteristic entities serve as nodes representing different disaster characteristics; the first, second, and third types of strategy entities serve as nodes representing different emergency strategies; and the first, second, and third relationships serve as edges representing the connections between disaster characteristics and emergency strategies. In this way, various characteristic information of distribution network disasters and corresponding emergency strategies are organically integrated to form a visualized and structured knowledge graph.

[0114] As one implementation method, the establishment of the first, second, and third association relationships described above can be specifically implemented as follows: steps S301 to S303:

[0115] Step S301: For the first association: when the overlap ratio of the temporal synchronization information of the first type of disaster characteristic entity is greater than the first association threshold, establish the association between the first type of disaster characteristic entity and the equipment preventive maintenance strategy; when the covariance value of the dependency information of the change trend of the first type of disaster characteristic entity is greater than the second association threshold, establish the association between the first type of disaster characteristic entity and the equipment emergency isolation strategy; the equipment preventive maintenance strategy includes maintenance scope determination rules and maintenance time window selection rules, and the equipment emergency isolation strategy includes isolation operation priority rules and system reconstruction principles after isolation.

[0116] The first and second association thresholds are pre-set thresholds used to determine whether to establish an association relationship. The overlap ratio of temporal synchronization information reflects the degree of synchronization between equipment parameter change events and environmental disturbance events in time. When the overlap ratio is greater than the first association threshold, it indicates a strong temporal correlation between equipment condition deterioration and environmental disturbance. In this case, an association is established between the first type of disaster characteristic entity and the equipment preventive maintenance strategy to perform equipment maintenance in advance and prevent failures. The covariance value of the trend dependence information reflects the correlation between the upward trend of equipment parameters and the increasing trend of environmental disturbances. When the covariance value is greater than the second association threshold, it indicates a strong consistency between the trends of equipment condition deterioration and environmental disturbances. In this case, an association is established between the first type of disaster characteristic entity and the equipment emergency isolation strategy to quickly isolate faulty equipment and prevent the fault from escalating.

[0117] To establish these relationships, it is first necessary to determine the first and second association thresholds. Appropriate thresholds can be determined through statistical analysis based on historical data and experience. For example, analysis of past power distribution network disaster data reveals that when the overlap ratio between equipment parameter change events and environmental disturbance events is greater than 0.7, implementing preventative maintenance strategies can effectively reduce equipment failure rates; therefore, the first association threshold is set to 0.7. When the covariance between the upward trend of equipment parameters and the increasing trend of environmental disturbances is greater than 0.8, implementing emergency isolation strategies can quickly block fault propagation; therefore, the second association threshold is set to 0.8. Then, in the knowledge graph, when the overlap ratio of the temporal synchronization information of the first type of disaster characteristic entities is greater than 0.7, an edge is added between the first type of disaster characteristic entities and the preventative maintenance strategy node to represent their association; when the covariance of the dependency information of the change trends of the first type of disaster characteristic entities is greater than 0.8, an edge is added between the first type of disaster characteristic entities and the emergency isolation strategy node to represent their association.

[0118] Step S302: Regarding the second association: when the intersection area of ​​the spatial overlap information of the second type of disaster characteristic entities is greater than the third association threshold, establish the association between the second type of disaster characteristic entities and the regional joint protection strategy; when the linear fit goodness value of the gradient correspondence information of the second type of disaster characteristic entities is greater than the fourth association threshold, establish the association between the second type of disaster characteristic entities and the gradient-level protection strategy; the regional joint protection strategy includes rules for the coordinated allocation of protection resources and rules for the synchronous execution of protection measures, and the gradient-level protection strategy includes rules for adjusting the protection intensity with the gradient and rules for phased verification of protection effects.

[0119] The third and fourth correlation thresholds are pre-set thresholds used to determine whether to establish a correlation. The intersection area of ​​spatial overlap information reflects the degree of spatial overlap between the environmental disturbance area and the load fluctuation area. When the intersection area is greater than the third correlation threshold, it indicates a strong spatial correlation between the environmental disturbance and the load fluctuation. In this case, a correlation is established between the second type of disaster characteristic entity and the regional joint protection strategy to mitigate environmental disturbances through resource coordination and synchronized measures within the region. The linear fit goodness value of gradient correspondence information reflects the goodness of the linear relationship between the wind speed attenuation gradient and the load growth decrease gradient. When the linear fit goodness value is greater than the fourth correlation threshold, it indicates a strong linear relationship between the two. In this case, a correlation is established between the second type of disaster characteristic entity and the gradient-based protection strategy to adjust the protection intensity according to gradient changes.

[0120] To establish these relationships, the third and fourth association thresholds must first be determined. Appropriate thresholds can be determined through statistical analysis based on historical data and experience. For example, analysis of past power distribution network disaster data reveals that when the intersection area of ​​the environmental disturbance area and the load fluctuation area is greater than 12 square kilometers, a regional joint protection strategy can effectively reduce the impact of environmental disturbances on the power distribution network; therefore, the third association threshold is set to 12 square kilometers. When the goodness-of-fit value of the linear fit between the wind speed attenuation gradient and the load growth decrease gradient is greater than 0.9, a gradient-based protection strategy can improve the utilization rate of protection resources; therefore, the fourth association threshold is set to 0.9. Then, in the knowledge graph, when the intersection area of ​​the spatial overlap information of the second type of disaster characteristic entities is greater than 12 square kilometers, an edge is added between the second type of disaster characteristic entities and the regional joint protection strategy nodes to represent their association; when the goodness-of-fit value of the linear fit of the gradient correspondence information of the second type of disaster characteristic entities is greater than 0.9, an edge is added between the second type of disaster characteristic entities and the gradient-based protection strategy nodes to represent their association.

[0121] Step S303: Regarding the third association: When the clarity of the direct causal relationship of the causal chain information of the third type of disaster characteristic entity is greater than the fifth association threshold, establish the association between the third type of disaster characteristic entity and the load dynamic transfer strategy; when the clarity of the path of the impact transmission path information of the third type of disaster characteristic entity is greater than the sixth association threshold, establish the association between the third type of disaster characteristic entity and the load classification and control strategy; the load dynamic transfer strategy includes rules for screening the type of transferred load and rules for selecting the transfer target line, and the load classification and control strategy includes rules for classifying the load control level and rules for ensuring the supply of important loads.

[0122] The fifth and sixth association thresholds are pre-set thresholds used to determine whether to establish an association relationship. The clarity of the direct causal relationship in the causal chain information reflects the clarity of the direct causal relationship between load change events and equipment status change events. When the clarity of the direct causal relationship is greater than the fifth association threshold, it indicates that the load change has a direct impact on the equipment status. In this case, an association is established between the third type of disaster characteristic entity and the dynamic load transfer strategy to alleviate the equipment burden by transferring the load. The clarity of the influence transmission path information reflects the clarity of the path through which the load change affects the equipment status. When the path clarity is greater than the sixth association threshold, it indicates that the path of the load change's impact on the equipment status is clear. In this case, an association is established between the third type of disaster characteristic entity and the load grading control strategy to enable grading control based on the importance of the load and the influence path.

[0123] To establish these relationships, the fifth and sixth correlation thresholds must first be determined. Appropriate thresholds can be determined through statistical analysis based on historical data and experience. For example, analysis of past power distribution network disaster data reveals that when the clarity of the direct causal relationship between load change events and equipment status change events is greater than 0.8, a dynamic load transfer strategy can effectively balance the load; therefore, the fifth correlation threshold is set to 0.8. When the clarity of the path from the increase in load to the equipment temperature via current parameters is greater than 0.9, a load grading control strategy can better ensure the supply of critical loads; therefore, the sixth correlation threshold is set to 0.9. Then, in the knowledge graph, when the clarity of the direct causal relationship of the causal chain information of the third-type disaster characteristic entity is greater than 0.8, an edge is added between the third-type disaster characteristic entity and the load dynamic transfer strategy node to represent their correlation; when the clarity of the path of the influence transmission information of the third-type disaster characteristic entity is greater than 0.9, an edge is added between the third-type disaster characteristic entity and the load grading control strategy node to represent their correlation.

[0124] As another implementation method, the process of establishing the first association, the second association, and the third association can be specifically implemented as follows: S300A~S300F:

[0125] Step S300A: Obtain the effect evaluation data of strategy implementation under historical disaster scenarios. The effect evaluation data includes the fault mitigation rate information of equipment intervention strategy, the disturbance resistance rate information of environmental buffer strategy, and the overload relief rate information of load adjustment strategy.

[0126] The historical disaster scenario strategy implementation effectiveness evaluation data records and evaluates the effectiveness of various emergency strategies implemented in past distribution network disasters. Fault mitigation rate information indicates the effectiveness of equipment intervention strategies in reducing equipment failure rates, which can be obtained by calculating the reduction in the number of equipment failures after implementing the intervention strategy. Disturbance mitigation rate information indicates the effectiveness of environmental buffering strategies in mitigating the impact of environmental disturbances on the distribution network, which can be obtained by calculating the reduction in losses caused by environmental disturbances to the distribution network after implementing environmental buffering strategies. Overload relief rate information indicates the effectiveness of load regulation strategies in relieving distribution network overloads, which can be obtained by calculating the reduction in the number of overloaded loads after implementing load regulation strategies.

[0127] To obtain this effectiveness evaluation data, relevant information can be extracted from historical disaster databases. Past power distribution network disaster events are compiled and analyzed, recording the emergency response strategies adopted for each event and their corresponding effectiveness data. Similarly, the disturbance resilience rate of environmental buffering strategies and the overload relief rate of load regulation strategies are calculated.

[0128] Step S300B: Perform strategy effectiveness analysis on the effect evaluation data to obtain the strategy effectiveness coefficient. The strategy effectiveness coefficient includes the historical fault mitigation rate of the equipment preventive maintenance strategy, the historical fault blocking rate of the equipment emergency isolation strategy, the historical disturbance attenuation rate of the regional joint protection strategy, the historical protection resource utilization rate of the gradient graded protection strategy, the historical load balance of the load dynamic transfer strategy, and the historical important load supply rate of the load graded control strategy.

[0129] Strategy effectiveness analysis involves in-depth analysis of effect evaluation data to determine the effectiveness of various emergency strategies. The strategy effectiveness coefficient is an indicator that measures the effectiveness of each strategy in historical disaster scenarios. This includes the historical fault mitigation rate of equipment preventative maintenance strategies, the historical fault blocking rate of equipment emergency isolation strategies, the historical disturbance attenuation rate of regional joint protection strategies, the historical protection resource utilization rate of gradient-level protection strategies, the historical load balancing degree of load dynamic transfer strategies, and the historical critical load supply guarantee rate of load grading and control strategies.

[0130] To analyze the effectiveness of strategies, statistical analysis can be performed on the effect evaluation data. For preventive maintenance strategies, the average fault mitigation rate across multiple historical disaster events is calculated as the historical fault mitigation rate. For emergency isolation strategies, the proportion of successful fault propagation blocking across multiple historical disaster events is calculated as the historical fault blocking rate. For regional joint protection strategies, the average proportion of environmental disturbance impact reduction across multiple historical disaster events is calculated as the historical disturbance attenuation rate. For tiered protection strategies, the average utilization rate of protection resources across multiple historical disaster events is calculated as the historical protection resource utilization rate. For dynamic load transfer strategies, the degree of load balancing achieved across multiple historical disaster events is calculated as the historical load balancing degree. For tiered load control strategies, the proportion of ensuring supply to critical loads across multiple historical disaster events is calculated as the historical critical load supply guarantee rate.

[0131] Step S300C: When establishing the first correlation, the strategy effectiveness coefficients of the equipment preventive maintenance strategy and the equipment emergency isolation strategy are used as the first correlation strength parameter; when establishing the second correlation, the strategy effectiveness coefficients of the regional joint protection strategy and the gradient graded protection strategy are used as the second correlation strength parameter; when establishing the third correlation, the strategy effectiveness coefficients of the load dynamic transfer strategy and the load graded control strategy are used as the third correlation strength parameter.

[0132] The correlation strength parameter measures the degree of correlation between disaster characteristic entities and strategy entities. Using the strategy effectiveness coefficients of equipment preventative maintenance and emergency equipment isolation strategies as the first correlation strength parameter means that the higher the strategy effectiveness, the stronger the correlation when establishing the first correlation. Similarly, the strategy effectiveness coefficients of regional joint protection and graded protection strategies are used as the second correlation strength parameter, and the strategy effectiveness coefficients of dynamic load transfer and graded load control strategies are used as the third correlation strength parameter.

[0133] In the knowledge graph, when establishing the first association, an edge is added between the first type of disaster characteristic entity and the equipment preventive maintenance strategy and the equipment emergency isolation strategy, and the corresponding strategy effectiveness coefficient is used as the edge weight. Similarly, when establishing the second and third associations, the strategy effectiveness coefficients of the regional joint protection strategy, the gradient hierarchical protection strategy, the load dynamic transfer strategy, and the load hierarchical control strategy are used as the edge weights, respectively.

[0134] Step S300D: The first correlation strength parameter, the second correlation strength parameter, and the third correlation strength parameter are positively correlated with the effectiveness coefficients of the corresponding strategies.

[0135] This indicates a positive correlation between the association strength parameter and the effectiveness coefficient of the corresponding strategy; that is, the higher the effectiveness coefficient of the strategy, the larger the association strength parameter. This positive correlation ensures that in the knowledge graph, disaster-feature entities are more closely associated with more effective strategies, thus allowing for the priority selection of more effective strategies during strategy reasoning.

[0136] Step S300E: Set the confidence threshold for the association. The confidence threshold includes the minimum fault mitigation rate requirement associated with the equipment intervention strategy, the minimum disturbance resistance rate requirement associated with the environmental buffer strategy, and the minimum overload relief rate requirement associated with the load regulation strategy.

[0137] The confidence threshold is used to determine whether a correlation is valid. The minimum fault mitigation rate requirement for equipment intervention strategy correlation indicates that the correlation is considered valid only when the fault mitigation rate of the equipment intervention strategy reaches this threshold. The minimum disturbance resilience rate requirement for environmental buffering strategy correlation indicates that the correlation is considered valid only when the disturbance resilience rate of the environmental buffering strategy reaches this threshold. The minimum overload relief rate requirement for load regulation strategy correlation indicates that the correlation is considered valid only when the overload relief rate of the load regulation strategy reaches this threshold.

[0138] These confidence thresholds can be determined based on actual needs and experience.

[0139] Step S300F: Establish a valid first association only when the first association strength parameter is greater than the confidence threshold of the device intervention strategy association; establish a valid second association only when the second association strength parameter is greater than the confidence threshold of the environmental buffer strategy association; establish a valid third association only when the third association strength parameter is greater than the confidence threshold of the load adjustment strategy association; valid associations are used to preferentially trigger historically validated strategy types.

[0140] In knowledge graphs, an association is considered valid only when the association strength parameter is greater than the corresponding confidence threshold. For example, when the first association strength parameter (historical fault mitigation rate) of a preventive maintenance strategy is greater than the minimum fault mitigation rate requirement (0.6) for an intervention strategy, a valid first association is established between the first type of disaster characteristic entity and the preventive maintenance strategy. Similarly, the second and third associations are judged according to the corresponding confidence thresholds. Valid associations can preferentially trigger historically validated strategy types during strategy reasoning, thereby improving the effectiveness and reliability of emergency strategies.

[0141] Step S400: Perform policy reasoning through policy-generated knowledge graph to generate an emergency policy candidate stream.

[0142] Strategic reasoning utilizes knowledge and relationships within a strategy-generated knowledge graph to deduce appropriate emergency strategies based on current disaster characteristics. The emergency strategy candidate stream is a flow of possible emergency strategies arranged chronologically, containing multiple candidate strategies tailored to the current disaster situation.

[0143] As one implementation method, step S400 involves performing strategy reasoning through a strategy generation knowledge graph to generate an emergency strategy candidate stream, which can be specifically implemented as follows: steps S410~S440:

[0144] Step S410: Extract the equipment status deterioration related feature instance at the current moment from the disaster evolution feature stream, query the disaster feature entity that matches the equipment status deterioration related feature instance in the strategy generation knowledge graph, obtain the equipment intervention strategy entity associated with the disaster feature entity, and generate the equipment intervention strategy unit.

[0145] Equipment condition deterioration-related feature instances are specific feature information in the disaster evolution feature stream that reflects the current moment's equipment condition deterioration due to environmental disturbances. Searching for matching disaster feature entities in the strategy generation knowledge graph involves finding the first-class disaster feature entity most similar to the current equipment condition deterioration-related feature instance. Obtaining the equipment intervention strategy entity associated with this disaster feature entity means finding preventative maintenance strategies or emergency isolation strategies connected to this disaster feature entity through the first association relationship.

[0146] To extract equipment condition deterioration-related feature instances at the current moment, the disaster evolution feature stream can be monitored in real time. At the current moment, data related to equipment condition deterioration-related features is extracted from the disaster evolution feature stream, such as the overlap ratio between equipment parameter change events and environmental disturbance events, and the covariance between the upward trend of equipment parameters and the strengthening trend of environmental disturbances. Then, in the policy generation knowledge graph, these data are compared with the attributes of the first type of disaster feature entities to find the most similar disaster feature entities. For example, by calculating the similarity between equipment condition deterioration-related feature instances and each first type of disaster feature entity in the knowledge graph, the entity with the highest similarity is selected as the matching entity. The equipment intervention strategy entities associated with the matching entities are obtained, and their relevant information is organized into equipment intervention strategy units.

[0147] As one implementation method, in step S410, a disaster feature entity matching the equipment condition deterioration associated feature instance is queried in the strategy generation knowledge graph, the equipment intervention strategy entity associated with the disaster feature entity is obtained, and an equipment intervention strategy unit is generated. Specifically, this can be implemented as follows: steps S411~S415:

[0148] Step S411: Perform similarity calculation between the temporal synchronization information of the equipment condition deterioration associated feature instance and the temporal synchronization information of the disaster feature entity in the strategy-generated knowledge graph to obtain the first similarity value.

[0149] Similarity calculation compares the temporal synchronization information of equipment condition deterioration-related feature instances with the temporal synchronization information of disaster feature entities in a knowledge graph. The first similarity value is an indicator of this similarity and can be calculated using various methods, such as Euclidean distance and cosine similarity.

[0150] To perform similarity calculations, the temporal synchronization information of equipment condition deterioration-related feature instances (such as the overlap ratio between equipment parameter change events and environmental disturbance events) and the temporal synchronization information of disaster feature entities in the knowledge graph are first preprocessed to ensure they have the same format and scope. Then, the Euclidean distance formula is used to calculate the distance between the two, converting the distance into a similarity value.

[0151] Step S412: Perform similarity calculation between the dependency information of the changing trend of the equipment condition deterioration associated feature instance and the dependency information of the changing trend of the disaster feature entity in the knowledge graph generated by the strategy, and obtain a second similarity value.

[0152] Similar to step S411, the dependency information of the changing trend of the equipment condition deterioration associated feature instances (such as the covariance between the upward trend of equipment parameters and the increasing trend of environmental disturbance) is compared with the dependency information of the changing trend of the disaster feature entities in the knowledge graph to obtain a second similarity value. Likewise, the data is first preprocessed to ensure it has the same format and range. Then, a suitable similarity calculation method, such as cosine similarity, is used to calculate the similarity between the two.

[0153] Step S413: Select the disaster feature entity with the largest sum of the first similarity value and the second similarity value as the matching entity.

[0154] The first and second similarity values ​​are added together to obtain the comprehensive similarity score between each disaster feature entity and the equipment condition deterioration associated feature instance. The disaster feature entity with the highest comprehensive similarity score is selected as the matching entity because it is most similar to the equipment condition deterioration associated feature instance.

[0155] Step S414: Obtain the device intervention strategy entity associated with the matching entity. The device intervention strategy entity includes the device preventive maintenance strategy and the device emergency isolation strategy.

[0156] In the strategy generation knowledge graph, identify the device intervention strategy entities that are connected to the matching entities through the first association relationship. These entities may be device preventive maintenance strategies or device emergency isolation strategies, and obtain their relevant information, such as type information and operation procedure information.

[0157] Step S415: Based on the severity information of the equipment status degradation associated feature instances, select the highest priority strategy from the equipment intervention strategy entities as the equipment intervention strategy unit. The severity information includes the urgency and scope of impact of the equipment parameter changes.

[0158] Severity information is the basis for judging the severity of equipment condition deterioration, including the urgency of changes in equipment parameters and the scope of impact. The urgency of changes in equipment parameters can be measured by the speed and magnitude of the changes, while the scope of impact can be determined by the number of affected devices and the affected area.

[0159] Based on the severity information, the equipment intervention strategy entities are filtered. If the equipment condition deterioration is urgent and has a wide impact, the emergency isolation strategy may be selected as the highest priority strategy; if the equipment condition deterioration is relatively minor, the preventive maintenance strategy may be selected. For example, when the equipment parameters change rapidly and a large number of devices are affected, the equipment condition deterioration is considered severe, and the emergency isolation strategy is selected as the equipment intervention strategy unit.

[0160] Step S420: Extract the environmental disturbance coupling feature instance at the current moment from the disaster evolution feature stream, query the disaster feature entity that matches the environmental disturbance coupling feature instance in the strategy generation knowledge graph, obtain the environmental buffer strategy entity associated with the disaster feature entity, and generate the environmental buffer strategy unit.

[0161] Environmental disturbance coupling feature instances are specific feature information in the disaster evolution feature stream that reflects the spatial correlation and gradient correspondence between environmental disturbances and load fluctuations at the current moment. Searching for matching disaster feature entities in the strategy generation knowledge graph involves finding the second type of disaster feature entity most similar to the current environmental disturbance coupling feature instance. Obtaining the environmental buffering strategy entity associated with this disaster feature entity means finding regional joint protection strategies or gradient-level protection strategies, etc., connected to this disaster feature entity through a second association relationship.

[0162] To extract environmental disturbance coupling feature instances at the current moment, the disaster evolution feature flow can be monitored in real time. At the current moment, data related to environmental disturbance coupling features are extracted from the disaster evolution feature flow, such as the intersection area of ​​the environmental disturbance region and the load fluctuation region, and the goodness of fit between the wind speed attenuation gradient and the load increase decrease gradient. Then, in the policy generation knowledge graph, these data are compared with the attributes of the second type of disaster feature entities to find the most similar disaster feature entity. The environmental buffering policy entity associated with this disaster feature entity is obtained, and its relevant information is organized into environmental buffering policy units.

[0163] Step S430: Extract the load response sensitive feature instance at the current moment from the disaster evolution feature stream, query the disaster feature entity that matches the load response sensitive feature instance in the strategy generation knowledge graph, obtain the load adjustment strategy entity associated with the disaster feature entity, and generate a load adjustment strategy unit.

[0164] Load response sensitive feature instances are specific feature information in the disaster evolution feature flow that reflects the impact of load changes on equipment status at the current moment. Searching for matching disaster feature entities in the strategy generation knowledge graph involves finding the third type of disaster feature entity most similar to the current load response sensitive feature instance. Obtaining the load adjustment strategy entity associated with this disaster feature entity means finding the load dynamic transfer strategy or load grading control strategy, etc., connected to this disaster feature entity through a third association relationship.

[0165] To extract the load response sensitive feature instances at the current moment, the disaster evolution feature stream can be monitored in real time. At the current moment, data related to the load response sensitive features are extracted from the disaster evolution feature stream, such as the causal chain information of load change events and device state change events, and the path information of the load increase amount transmitted to the device temperature through current parameters. Then, in the policy generation knowledge graph, these data are compared with the attributes of the third type of disaster feature entities to find the most similar disaster feature entities. The load regulation policy entities associated with the disaster feature entities are obtained, and their relevant information is sorted into load regulation policy units.

[0166] Step S440: Arrange the device intervention policy units, environmental buffer policy units, and load regulation policy units in chronological order of timestamps to generate an emergency policy candidate stream.

[0167] Arrange the generated device intervention policy units, environmental buffer policy units, and load regulation policy units in chronological order of timestamps to form a continuous emergency policy candidate stream. The chronological order of timestamps reflects the execution order of these policies during the disaster evolution process. By arranging them in chronological order of timestamps, it can ensure that the implementation of emergency policies has a reasonable timing sequence, thus more effectively coping with distribution network disasters.

[0168] To achieve the arrangement in chronological order of timestamps, first, each policy unit needs to be marked with its corresponding timestamp. For the device intervention policy units, environmental buffer policy units, and load regulation policy units, according to the time information of the disaster features they target in the disaster evolution feature stream, the corresponding timestamps are assigned to them. For example, when the device state deterioration associated feature instance at the current moment is extracted from the disaster evolution feature stream and a device intervention policy unit is generated, this moment is used as the timestamp of the device intervention policy unit.

[0169] Then, a sorting algorithm is used to sort these policy units. Simple sorting algorithms such as bubble sort and quick sort can be adopted to arrange the device intervention policy units, environmental buffer policy units, and load regulation policy units in ascending order according to the chronological order of timestamps. Suppose the timestamp of the device intervention policy unit is t1, the timestamp of the environmental buffer policy unit is t2, and the timestamp of the load regulation policy unit is t3, and t1 < t2 < t3. Then, the order of the emergency policy candidate stream after sorting is the device intervention policy unit, the environmental buffer policy unit, and the load regulation policy unit.

[0170] In this way, the generated emergency policy candidate stream can clearly show the emergency policies to be taken at different time points, providing clear guidance for subsequent emergency handling.

[0171] Step S500: Conduct a coordination verification on the emergency policy candidate stream to generate a distribution network disaster emergency policy.

[0172] Collaboration verification checks whether the various strategy units in the emergency strategy candidate stream can cooperate and work together to avoid problems such as resource conflicts, time conflicts, or mutual cancellation of effects. Through collaboration verification, it can be ensured that the final generated distribution network disaster emergency strategy is efficient and reliable, and can better cope with distribution network disasters.

[0173] As one implementation method, step S500 involves performing collaborative verification on the emergency strategy candidate streams to generate a distribution network disaster emergency strategy. Specifically, this can be implemented as follows: steps S510~S540:

[0174] Step S510: Perform resource utilization coordination verification on the equipment intervention strategy unit and the environmental buffer strategy unit in the emergency strategy candidate stream to obtain the resource utilization coordination verification result. The resource utilization coordination verification result includes information on the overlap and conflict of human and material resources required for strategy implementation and information on complementary utilization.

[0175] Resource utilization coordination verification analyzes the human and material resource requirements of equipment intervention strategy units and environmental buffering strategy units during implementation, determining whether resource conflicts exist or whether complementary resource utilization can be achieved. Information on overlapping and conflicting human and material resources indicates whether the two strategy units' demand for the same resources at the same time will lead to competition, such as whether the same group of maintenance personnel or the same batch of protective equipment is needed. Information on complementary utilization indicates whether the two strategy units can share certain resources or achieve better results through reasonable resource allocation; for example, transport vehicles required for the equipment intervention strategy can be used for material allocation in the environmental buffering strategy after the equipment transportation is completed.

[0176] To verify resource utilization coordination, the required human and material resources for each of the equipment intervention strategy unit and the environmental buffering strategy unit are first defined. For the equipment intervention strategy unit, human resource requirements may include maintenance personnel and technicians, while material requirements may include maintenance tools and spare equipment. For the environmental buffering strategy unit, human resource requirements may include protection personnel and monitoring personnel, while material requirements may include protective materials and monitoring equipment. Then, the two lists are compared to analyze resource overlap and complementarity. For example, analysis reveals that both the equipment intervention strategy unit and the environmental buffering strategy unit require cranes for material handling, resulting in resource overlap and conflict. However, surplus protective materials from the equipment intervention strategy unit can be used for protection work in the environmental buffering strategy unit, demonstrating complementary resource utilization. Based on the analysis results, the resource utilization coordination verification results are obtained.

[0177] Step S520: Perform time-based collaborative verification on the environmental buffer strategy unit and the load adjustment strategy unit in the emergency strategy candidate stream to obtain the time-based collaborative verification result. The time-based collaborative verification result includes overlapping and conflicting information of strategy start time and duration, as well as timing information.

[0178] Time-based execution coordination verification checks whether the timing of the environment buffering strategy unit and the load conditioning strategy unit is reasonable, whether there are time conflicts, or whether good timing coordination can be achieved. Information on overlapping and conflicting strategy start times and durations indicates whether the start times and durations of the two strategy units will interfere with each other. For example, will the environment buffering strategy unit and the load conditioning strategy unit start at the same time, leading to resource allocation chaos or mutual impact on effects? Timing coordination information indicates whether the two strategy units can be executed sequentially and in an orderly manner. For example, can the load conditioning strategy start after the environment buffering strategy has created certain environmental conditions to achieve better results?

[0179] To perform time-based co-validation, it is necessary to obtain the start-up time and duration information of the environmental buffering strategy unit and the load conditioning strategy unit. This information can be estimated based on the nature of the strategies and the actual situation. For example, the environmental buffering strategy unit may need to start immediately after a disaster, lasting for several hours; the load conditioning strategy unit may need to start after the environmental buffering strategy has achieved some effect, lasting for several days. Then, the timing of the two strategy units is compared to analyze the possibility of time conflicts and timing coordination. For example, if the duration of the environmental buffering strategy unit covers the start-up time of the load conditioning strategy unit, a time conflict exists; if the load conditioning strategy unit can start immediately after the environmental buffering strategy unit ends, and the effects of both can mutually reinforce each other, then good timing coordination is achieved. Based on the analysis results, the time-based co-validation results are obtained.

[0180] Step S530: Perform effect target co-verification on the load adjustment strategy unit and equipment intervention strategy unit in the emergency strategy candidate stream to obtain the effect target co-verification result. The effect target co-verification result includes the consistency information of the strategy's disaster mitigation target and the information on effect superposition or offset.

[0181] The effectiveness-objective synergy verification assesses whether the effectiveness objectives of load regulation and equipment intervention strategies are consistent in responding to distribution network disasters, and whether their implementation effects will be additive or cancel each other out. Information on the consistency of strategies' disaster mitigation objectives indicates whether both strategy units are working towards the same goal of mitigating distribution network disasters, such as reducing equipment failure rates or minimizing load overload. Information on effect additive or canceling out indicates whether the implementation effects of the two strategy units can mutually enhance or weaken each other. For example, if a load regulation strategy reduces equipment burden by shifting load, while an equipment intervention strategy performs preventative maintenance, the effects of both can be additive, leading to better disaster mitigation. Conversely, if the implementation of a load regulation strategy causes changes in equipment operating conditions, and the equipment intervention strategy does not account for these changes, the effects of both may cancel each other out.

[0182] To verify the synergistic effect objectives, it is necessary to clarify the effect objectives of the load regulation strategy unit and the equipment intervention strategy unit. The effect objective of the load regulation strategy unit might be to achieve load balancing or relieve overload, while the effect objective of the equipment intervention strategy unit might be to prevent equipment failure or repair faulty equipment. Then, it is necessary to analyze whether the effect objectives of the two strategy units are consistent and the relationship between their implementation effects. The possibility of effect superposition or cancellation can be assessed through simulation experiments and historical data statistics. For example, analysis of historical data reveals that in some cases, the combined implementation of the load regulation strategy and the equipment intervention strategy can significantly reduce the equipment failure rate, indicating that their effects are superimposed; while in other cases, improper implementation of the load regulation strategy affects the effect of the equipment intervention strategy, indicating that their effects cancel each other out. Based on the analysis results, the synergistic effect objective verification results are obtained.

[0183] Step S540: Based on the resource usage collaborative verification results, time execution collaborative verification results, and effect target collaborative verification results, conflict resolution and collaborative optimization are performed on the emergency strategy candidate flow to generate a conflict-free and collaboratively enhanced distribution network disaster emergency strategy.

[0184] Conflict resolution and collaborative optimization involves adjusting and optimizing the emergency strategy candidate flow based on the collaborative verification results of resource usage, time execution, and effect objectives, thereby eliminating conflicts between strategy units and enhancing the collaborative effect between strategies.

[0185] For resource overlaps and conflicts identified in the resource utilization collaboration verification results, a resource reallocation method is adopted to resolve them. For example, if both the equipment intervention strategy unit and the environmental buffer strategy unit need to use cranes, the crane usage time and task allocation are rationally arranged according to the strategy priority and actual needs to ensure effective resource utilization. For complementary resource utilization information, resource allocation can be further optimized to achieve resource sharing and collaborative use, thereby improving resource utilization efficiency.

[0186] For time conflicts discovered in the time-based execution co-validation results, adjust the start time and duration of the strategies. For example, if there is a time conflict between the environment buffering strategy unit and the load conditioning strategy unit, the start time of one of the strategies can be appropriately delayed, or the duration of a certain strategy can be shortened to avoid the conflict. For timing sequence information, optimize the execution order of the strategies based on the interrelationships between them to ensure that the strategies can be executed sequentially in an orderly manner, achieving good timing sequence.

[0187] For any offsetting effects discovered in the collaborative verification results of the effectiveness objectives, the strategies are adjusted and improved. For example, if the effects of load regulation strategies and equipment intervention strategies offset each other, the reasons are analyzed, and the parameters or implementation methods of the strategies are adjusted so that their effects can mutually enhance each other. Simultaneously, based on the consistency information of the effectiveness objectives, the common goals of the strategies are further clarified, the combination of strategies is optimized, and the overall effectiveness of the strategies is improved.

[0188] Through the above conflict resolution and collaborative optimization process, the candidate flow of emergency strategies is comprehensively adjusted and optimized, ultimately generating a conflict-free and collaboratively enhanced distribution network disaster emergency strategy. This strategy can fully leverage the advantages of each strategy unit to collaboratively respond to distribution network disasters and improve the reliability and stability of the distribution network.

[0189] It is understood that the various algorithms involved in the above descriptions of the embodiments of the present invention, such as the Euclidean distance algorithm, the cosine distance algorithm, the causal analysis algorithm, etc., can all be obtained from relevant content in the prior art. To save space, they will not be elaborated on in the embodiments of the present invention. In addition, those skilled in the art can supplement the details based on common knowledge in the art when implementing the solutions of the present invention. For example, they can use normalization to eliminate dimensional conflicts before feature fusion, use interpolation to eliminate dimensional differences, reasonably set thresholds based on historical data, experience or business scenario requirements, train the model based on a general model training method, set the number of layers in the model structure based on actual needs, select activation functions, etc. The present invention will not provide redundant descriptions of overly detailed implementation processes here.

[0190] Please see Figure 2 , Figure 2This is a schematic diagram of a computer system provided in an embodiment of the present invention. The computer system includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 can be connected via a bus or other means. The processor 101 (or Central Processing Unit, CPU) is the computing and control core of the computer system, capable of parsing various instructions and processing various data within the computer system. The communication interface 102 may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, mobile communication interface, etc.), and can be used to send and receive data under the control of the processor 101; the communication interface 102 can also be used for data transmission and interaction within the computer system. The memory 103 is a storage device in the computer system used to store programs and data. It is understood that the memory 103 here can include the computer system's built-in memory, or it can include extended memory supported by the computer system. The memory 103 provides storage space, which stores the computer system's operating system; this invention does not limit this storage space.

[0191] In one embodiment, the processor 101 executes the emergency strategy generation method for power distribution network disasters provided above in the embodiments of the present invention by running a computer program in the memory 103.

Claims

1. A method for generating emergency strategies for power distribution network disasters, characterized in that, The method includes: Acquire multimodal sensing data of the power distribution network operating environment, including equipment status sensing sequence, environmental disturbance sensing sequence, and load response sensing sequence; Modeling disaster evolution features on the multimodal sensing data to generate a disaster evolution feature stream specifically includes: performing temporal correlation analysis on the equipment state sensing sequence and the environmental disturbance sensing sequence to generate equipment state degradation correlation features, which include temporal synchronization information and trend dependence information of equipment parameter changes and environmental disturbances; performing spatial coupling analysis on the environmental disturbance sensing sequence and the load response sensing sequence to generate environmental disturbance coupling features, which include spatial overlap information of environmental disturbance regions and load fluctuation regions and gradient correspondence information of disturbance intensity and load fluctuation amplitude; performing causal dependency analysis on the load response sensing sequence and the equipment state sensing sequence to generate load response sensitivity features, which include causal chain information of load change events and equipment state change events and information on the transmission path of the impact of load change on equipment state; and performing feature temporal fusion on the equipment state degradation correlation features, the environmental disturbance coupling features, and the load response sensitivity features to generate the disaster evolution feature stream. The strategy for generating a knowledge graph based on the disaster evolution feature flow specifically includes: extracting equipment condition degradation correlation features from the disaster evolution feature flow, defining the temporal synchronization information and the dependence information of the change trend in the equipment condition degradation correlation features as first-type disaster feature entities; extracting equipment intervention strategies from a preset strategy library, defining the type information and operation process information of the equipment intervention strategies as first-type strategy entities; establishing a first association relationship between the first-type disaster feature entities and the first-type strategy entities; extracting environmental disturbance coupling features from the disaster evolution feature flow, defining the spatial overlap information and gradient correspondence information in the environmental disturbance coupling features as second-type disaster feature entities; and extracting environmental buffering strategies from the preset strategy library, defining the type information and implementation requirement information of the environmental buffering strategies as second-type strategies. Entities; establish a second association between the second type of disaster characteristic entity and the second type of strategy entity; extract load response sensitive features from the disaster evolution characteristic flow, and define the causal chain information and impact transmission path information in the load response sensitive features as a third type of disaster characteristic entity; extract load adjustment strategies from a preset strategy library, and define the type information and execution rule information of the load adjustment strategies as a third type of strategy entity; establish a third association between the third type of disaster characteristic entity and the third type of strategy entity; combine the first type of disaster characteristic entity, the first type of strategy entity and the first association, the second type of disaster characteristic entity, the second type of strategy entity and the second association, and the third type of disaster characteristic entity, the third type of strategy entity and the third association to form the strategy generation knowledge graph; The strategy generates a knowledge graph for strategy reasoning, resulting in a candidate stream of emergency strategies. The candidate emergency strategies are collaboratively verified to generate emergency strategies for power distribution network disasters.

2. The method as described in claim 1, characterized in that, The step of performing time-series correlation analysis on the equipment state perception sequence and the environmental disturbance perception sequence to generate equipment state degradation correlation features includes: Extract the device parameter change event points from the device status sensing sequence, and extract the environmental disturbance event points from the environmental disturbance sensing sequence; The device parameter change event points and the environmental disturbance event points are timestamped to obtain time synchronization information. The time synchronization information includes the overlap ratio information of the device parameter change event and the environmental disturbance event on the time axis. A dependency analysis is performed on the changing trend of the equipment parameter change event points and the disturbance trend of the environmental disturbance event points to obtain the dependency information of the changing trend. The dependency information of the changing trend includes the covariance information of the increasing trend of equipment parameters and the increasing trend of environmental disturbance. The timing synchronization information and the dependence information of the change trend are combined to form the device state degradation correlation feature.

3. The method as described in claim 1, characterized in that, The step of performing spatial coupling analysis on the environmental disturbance sensing sequence and the load response sensing sequence to generate environmental disturbance coupling features includes: Extract the disturbance area distribution information from the environmental disturbance sensing sequence. The disturbance area distribution information includes the geographical location information of high wind speed areas, the coverage information of heavy precipitation areas, and the boundary coordinate information of high frequency thunderstorm areas. Extract the load fluctuation region distribution information from the load response sensing sequence. The load fluctuation region distribution information includes the geographical location information of the high load growth region, the coverage information of the load drop region, and the boundary coordinate information of the load oscillation region. Spatial overlap calculation is performed on the distribution information of the disturbance area and the distribution information of the load fluctuation area to obtain spatial overlap information, which includes the intersection area information of the high wind speed area and the high load growth area. Extract the disturbance intensity gradient information from the environmental disturbance sensing sequence. The disturbance intensity gradient information includes the wind speed attenuation gradient, precipitation decrease gradient, and thunderstorm frequency decrease gradient from the disturbance center to the boundary. Extract the load fluctuation amplitude gradient information from the load response sensing sequence. The load fluctuation amplitude gradient information includes the load increase decreasing gradient and the load decrease increasing gradient from the load center to the boundary. Gradient correspondence analysis is performed on the disturbance intensity gradient information and the load fluctuation amplitude gradient information to obtain gradient correspondence information, which includes the linear fitting relationship information between the wind speed attenuation gradient and the load growth decrease gradient. The spatial overlap information and the gradient correspondence information are combined to form the environmental perturbation coupling feature.

4. The method as described in claim 1, characterized in that, The step of performing causal dependency analysis on the load response sensing sequence and the device state sensing sequence to generate load response sensitive features includes: Extract the load change event chain from the load response sensing sequence. The load change event chain includes a load increase trigger event, a load stability maintenance event, and a load decrease termination event. Extract the equipment state change event chain from the equipment state perception sequence. The equipment state change event chain includes normal equipment operation events, equipment abnormality warning events, and equipment failure shutdown events. Causal relationship mining is performed on the load change event chain and the device state change event chain to obtain causal chain information. The causal chain information includes direct causal relationship information from the load increase trigger event to the device abnormal warning event and indirect causal relationship information from the load stability maintenance event to the device normal operation event. Extract the load change information from the load response sensing sequence. The load change information includes the absolute value of the load increase, the absolute value of the load decrease, and the relative proportion of the load change. Extract the equipment status impact information from the equipment status perception sequence. The equipment status impact information includes the degree of impact of load changes on equipment voltage, the degree of impact on equipment current, and the degree of impact on equipment temperature. The load change information and the equipment status impact information are analyzed to obtain impact transmission path information, which includes the path information of the load increase being transmitted to the equipment temperature through the current parameter. The causal chain information and the influence propagation path information are combined to form the load response sensitivity feature.

5. The method as described in claim 1, characterized in that, The methods for establishing the first association, the second association, and the third association include: Regarding the first association: when the overlap ratio of the temporal synchronization information of the first type of disaster characteristic entities is greater than the first association threshold, an association is established between the first type of disaster characteristic entities and the equipment preventive maintenance strategy; when the covariance value of the dependency information of the changing trend of the first type of disaster characteristic entities is greater than the second association threshold, an association is established between the first type of disaster characteristic entities and the equipment emergency isolation strategy; the equipment preventive maintenance strategy includes maintenance scope determination rules and maintenance time window selection rules, and the equipment emergency isolation strategy includes isolation operation priority rules and system reconstruction principles after isolation; Regarding the second association: when the intersection area of ​​the spatial overlap information of the second type of disaster characteristic entities is greater than the third association threshold, an association is established between the second type of disaster characteristic entities and the regional joint protection strategy; when the linear fit goodness value of the gradient correspondence information of the second type of disaster characteristic entities is greater than the fourth association threshold, an association is established between the second type of disaster characteristic entities and the gradient-level protection strategy; the regional joint protection strategy includes rules for the coordinated allocation of protection resources and rules for the synchronous execution of protection measures, and the gradient-level protection strategy includes rules for adjusting the protection intensity as the gradient changes and rules for phased verification of protection effects; Regarding the third association: when the direct causal relationship of the causal chain information of the third type of disaster characteristic entity is clearer than the fifth association threshold, an association is established between the third type of disaster characteristic entity and the load dynamic transfer strategy; when the path clarity of the impact transmission path information of the third type of disaster characteristic entity is greater than the sixth association threshold, an association is established between the third type of disaster characteristic entity and the load classification and control strategy; the load dynamic transfer strategy includes rules for filtering the type of transferred load and rules for selecting the target line for transfer, and the load classification and control strategy includes rules for classifying load control levels and rules for ensuring the supply of important loads.

6. The method as described in claim 1, characterized in that, The step of generating a knowledge graph through the strategy to perform strategy reasoning and generate an emergency strategy candidate stream includes: Extract the equipment status degradation associated feature instance at the current moment from the disaster evolution feature stream, query the disaster feature entity that matches the equipment status degradation associated feature instance in the strategy generation knowledge graph, obtain the equipment intervention strategy entity associated with the disaster feature entity, and generate the equipment intervention strategy unit. Extract the environmental disturbance coupling feature instance at the current moment from the disaster evolution feature stream, query the disaster feature entity that matches the environmental disturbance coupling feature instance in the strategy generation knowledge graph, obtain the environmental buffer strategy entity associated with the disaster feature entity, and generate an environmental buffer strategy unit. Extract the load response sensitive feature instance at the current moment from the disaster evolution feature stream, query the disaster feature entity that matches the load response sensitive feature instance in the strategy generation knowledge graph, obtain the load adjustment strategy entity associated with the disaster feature entity, and generate a load adjustment strategy unit. The device intervention strategy unit, the environment buffer strategy unit, and the load adjustment strategy unit are arranged in timestamp order to generate an emergency strategy candidate stream.

7. The method as described in claim 6, characterized in that, The step of querying the knowledge graph generated by the strategy to find the disaster feature entity that matches the equipment state degradation associated feature instance, obtaining the equipment intervention strategy entity associated with the disaster feature entity, and generating the equipment intervention strategy unit includes: The temporal synchronization information of the equipment state deterioration associated feature instance is compared with the temporal synchronization information of the disaster feature entity in the strategy-generated knowledge graph to obtain a first similarity value; The second similarity value is obtained by comparing the dependency information of the changing trend of the equipment condition deterioration associated feature instance with the dependency information of the changing trend of the disaster feature entity in the knowledge graph generated by the strategy. The disaster feature entity with the largest sum of the first similarity value and the second similarity value is selected as the matching entity; Obtain the device intervention strategy entity associated with the matching entity, wherein the device intervention strategy entity includes device preventive maintenance strategy and device emergency isolation strategy; Based on the severity information of the device state degradation associated feature instances, the highest priority strategy is selected from the device intervention strategy entities as the device intervention strategy unit. The severity information includes the urgency and scope of impact of the device parameter changes.

8. A computer system, characterized in that, include: A memory, wherein a computer program is stored; A processor is configured to load the computer program to implement the emergency strategy generation method for power distribution network disasters as described in any one of claims 1-7.

Images