A method for monitoring and managing power safety

Abstract

This application discloses a power safety monitoring and management method, relating to the field of power monitoring technology, comprising: acquiring a basic dataset, calculating the dependency weights between devices and obtaining a set of associated devices, and constructing an initial network graph based on the set of associated devices; identifying abnormal devices and their corresponding abnormal time nodes; obtaining a first association group and a second association group based on the set of associated devices; overlapping the first association group and the second association group with the initial network graph to obtain a main path and a secondary path, respectively; calculating the risk intensity value of each device in the main path, obtaining the handling efficiency and evaluation indicators at different time nodes, and obtaining the optimal intervention time; calculating the risk delay growth curve of the secondary path, identifying the inflection point of risk growth, and determining the optimal prevention time; and by quantifying electrical dependence and physical distance, breaking through the limitation of traditional topology networks that only focus on physical connections, accurately locking the critical path of fault propagation, and quantifying cross-device chain reactions.

Description

Technical Field

[0001] This invention relates to the field of power monitoring technology, and in particular to a method for power safety monitoring and management. Background Technology

[0002] In power systems, substations are a crucial component. In the current field of substation safety monitoring and management, traditional methods mainly rely on manual inspections and periodic maintenance, which suffer from incomplete equipment monitoring and delayed safety risk warnings. Particularly noteworthy is the insufficient consideration of inter-equipment risks by current technology. When one piece of equipment fails, it is often impossible to prevent the fault from spreading to related equipment in a timely manner, leading to an expansion of the fault range and impacting the overall safe operation of the substation.

[0003] Current fault analysis and assessment methods are mostly based on single devices, failing to fully consider the cascading effects of device failures on other devices. Furthermore, the static and fixed network topology makes it difficult to dynamically adapt to device damage or line reconfiguration. At the same time, ignoring the delayed burst characteristics of faults in the time dimension further limits the effectiveness and timeliness of fault handling.

[0004] Therefore, developing a power safety monitoring and management method that can comprehensively monitor equipment, accurately predict risks, dynamically adapt to changes, and consider spatiotemporal continuity has become an urgent need to improve the safety of substations. Summary of the Invention

[0005] This application provides a power safety monitoring and management method that solves the problems of insufficient monitoring and low safety of substation equipment in the prior art. It achieves the technical effect of fully analyzing the potential correlation between multiple devices, realizing comprehensive monitoring, timely early warning and fault elimination.

[0006] This application provides a power safety monitoring and management method, the method comprising:

[0007] S1: Obtain the basic dataset, calculate the dependency weights between devices and obtain the set of associated devices, and construct the initial network graph based on the set of associated devices;

[0008] S2: Monitor device status in real time, identify abnormal devices and corresponding abnormal time points; obtain the first association group and the second association group based on the associated device set; overlap the first association group and the second association group with the initial network graph to obtain the main path and the secondary path, respectively.

[0009] S3: Calculate the risk intensity value of each device in the main path, obtain the handling efficiency and evaluation indicators at different time points, and obtain the optimal intervention time; calculate the risk delay growth curve of the secondary path, identify the inflection point of risk growth, and determine the optimal prevention time;

[0010] S4: If the risk intensity value is not greater than the high threshold, between the optimal intervention time and the optimal prevention time, identify alternative equipment based on the status of abnormal equipment and alternative indicators, construct a complementary characteristic matrix, and determine alternative transmission paths.

[0011] Furthermore, the method also includes: S5: acquiring historical intervention data records, generating high-dimensional state vectors and defining a set of intervention actions, and constructing a digital twin model; based on the historical intervention data records and the digital twin model, simulating the response process under different fault scenarios, introducing random perturbations, and obtaining the best intervention actions and corresponding intervention times under different abnormal states according to the reward function index;

[0012] S6: Obtain the target area and several substations within the target area, and determine a central station; obtain real-time data of adjacent stations, determine cross-station dependency weights based on the adjacent station tie line parameters, and divide the collaborative defense area; when abnormal equipment is detected, identify the abnormal state, obtain the optimal intervention action and corresponding intervention time of the central station according to step S5, and determine the cross-station propagation probability and resource requirements of the central station, and generate a global management plan based on the collaborative defense area to determine global risk constraints.

[0013] Furthermore, the basic dataset includes the electrical parameters, physical location, and control logic of the power equipment;

[0014] The dependency weight is a parameter value used to quantify the energy transfer dependency between devices;

[0015] Based on the set of associated devices, obtain the specific device nodes, establish associated edges, and mark the dependency weights to form an initial network graph;

[0016] The associated device set includes a direct association set and an indirect association set; based on the direct association set, devices associated with the abnormal device are obtained and form a first association group; based on the indirect association set, devices associated with the abnormal device are obtained and form a second association group.

[0017] Furthermore, the first association group and the second association group are overlapped with the initial network graph to obtain the main path and the secondary path, respectively, including: determining the specific location of the abnormal device in the initial network graph, extracting the device nodes in the first association group, and connecting the device nodes to form the main path; extracting the device nodes in the second association group and connecting the device nodes to form the secondary path.

[0018] Furthermore, the risk intensity value refers to the degree of dynamic failure risk of the equipment under specific spatiotemporal coordinates; the specific spatiotemporal coordinates are the spatial coordinates and time axis coordinates of the equipment, the spatial coordinates are used to locate the physical location of the fault propagation, and the time axis coordinates are used to quantify the dynamic evolution process of the fault over time.

[0019] Based on the current waveform data of abnormal devices obtained from the dataset, the accumulated fault energy is calculated, and evaluation indicators are set to assess and verify the handling efficiency at different time points to obtain the optimal intervention time. The accumulated fault energy refers to the fault energy generated over time, which is used to measure the destructive capability of the fault. The evaluation indicators include the energy blocking rate and the energy consumption reduction rate per unit time.

[0020] Furthermore, the risk delay growth curve of the secondary path is calculated, the inflection point of risk growth is identified, and the optimal prevention time is determined. This includes: obtaining the device nodes in the secondary path and their corresponding initial dependency weights; calculating the risk secondary value corresponding to the time change of each device node; generating the risk delay growth curve based on the risk secondary value and the time node; calculating the risk change rate based on the risk secondary value; identifying the inflection point of risk growth; and determining the optimal prevention time.

[0021] The initial risk sub-value for each device is 0, and the risk sub-value at the corresponding time is calculated based on the initial dependency weight, environmental acceleration factor, and line degradation rate.

[0022] Furthermore, the high threshold is preset and used to measure the degree of abnormality of the abnormal device;

[0023] The substitution metrics include functional compensation degree, capacity matching rate, and switching delay. The functional compensation degree measures the percentage by which the substitution device covers the critical capabilities of the faulty device. The capacity matching rate is the ratio of the capacity of the substitution device to the needs of the faulty device. The switching delay is the time required to switch from the faulty device to the substitution device, including mechanical action time, grid synchronization time, and system stabilization time.

[0024] Furthermore, a complementary characteristic matrix is ​​constructed based on the state of the abnormal equipment, including: the complementary characteristic matrix is ​​a three-dimensional matrix, including candidate replacement equipment, abnormal state value and comprehensive value; the abnormal state of the abnormal equipment is analyzed to determine the functional loss rate, capacity gap and risk diffusion rate, and the abnormal state value is obtained by weighted summation;

[0025] The comprehensive value is the difference between the substitution index value and the abnormal state value of each candidate alternative device. The larger the difference, the greater the substitution capability of the candidate alternative device.

[0026] Determining alternative transmission paths includes: identifying several candidate paths based on the complementary characteristic matrix; determining the initial alternative device for each candidate path based on the connection relationship between devices; retaining candidate paths corresponding to the initial alternative devices with a dependency weight greater than 0.6 for the abnormal device; arranging the dependency weights of subsequent alternative devices in descending order; and selecting them sequentially to form alternative transmission paths.

[0027] Furthermore, the historical intervention data record includes fault type, occurrence time, intervention action, loss result, and time tag. The time tag includes the theoretical optimal intervention time and the actual optimal time. The actual optimal time is the time obtained by expert analysis after the anomaly occurs.

[0028] The high-dimensional state vector includes equipment status, risk assessment indicators, and environmental parameters; the risk assessment indicators include the trend of risk intensity value change, the propagation speed of the main path, and the delay change speed of the secondary path.

[0029] The set of intervention actions includes early intervention, delayed intervention, and action granularity. The action granularity is a fixed time step, and the specific number of steps to delay or advance is intelligently selected according to the time step discretization. The reward function indicators include disaster reduction benefits, operational costs, and system stability. The random disturbances include fluctuations in new energy output, sudden load changes, and sensor noise.

[0030] Furthermore, based on the basic dataset of each substation within the target area, a central site and its defense radius are determined; the collaborative range is determined based on the capacity of the central site to form the defense radius.

[0031] The parameters of the adjacent station tie lines include voltage level ratio and power transmission ratio; the collaborative defense area is divided according to the cross-station dependency weight and the minimum defense radius of the other stations, and a clustering method is adopted. The optimization objective is to maximize the connection density within the area and minimize the connection between areas. The collaborative defense area includes strongly coupled area, medium coupled area and weakly coupled area.

[0032] The cross-site propagation probability is the rate of change of the risk intensity value after the central site's intervention action within a unit of time. The unit of time is dynamically set according to the range of the target area, and its value is less than the optimal prevention time of the central site.

[0033] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0034] By quantifying electrical dependence and physical distance, this approach overcomes the limitations of traditional topology networks that only focus on physical connections. It accurately identifies critical paths for fault propagation, dynamically superimposes risk intensity values ​​across multiple devices, quantifies cross-device chain reactions, and solves the problem of misjudging risks in remote devices. It predicts the optimal intervention time, addressing the shortcomings of static risk models in handling delayed faults. It identifies the inflection point of risk growth on secondary paths, providing a final time window for preventative maintenance and avoiding secondary faults. By calculating and determining alternative transmission paths between the optimal intervention time and the optimal prevention time, it protects currently abnormal devices, further reduces the occurrence of faults, and prevents secondary propagation of faults. Attached Figure Description

[0035] Figure 1This is a schematic diagram of a power safety monitoring and management method according to an embodiment of the present invention. Detailed Implementation

[0036] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0038] Example 1: As Figure 1 As shown, a power safety monitoring and management method includes:

[0039] S1: Obtain the basic dataset, calculate the dependency weights between devices and obtain the set of associated devices, and construct the initial network graph based on the set of associated devices.

[0040] In some embodiments, the basic dataset includes electrical parameters, physical location, and control logic of the power equipment; the electrical parameters include basic electrical quantities such as standard voltage, current, and resistance of the equipment and their related characteristic parameters; the physical location is the actual coordinate of the equipment; and the control logic refers to the linkage mechanism between devices implemented through preset rules, which is the core decision to prevent fault propagation. For example, when device A malfunctions, the control logic automatically triggers preset actions of devices B / C to form a fault isolation barrier and prevent fault propagation. When the transformer temperature exceeds the limit, the upstream circuit breaker is immediately tripped while the automatic switching of the downstream capacitor is blocked (to prevent overvoltage breakdown of downstream equipment). Preferably, when acquiring the real-time status of the equipment, dual-mode sensors deployed on key nodes such as transformers and circuit breakers are used to capture microsecond-level current surges and mechanical deformations.

[0041] In some embodiments, the dependency weight is a parameter value used to quantify the energy transfer dependency between devices. Calculating the dependency weight between any two devices requires a weighted sum based on electrical dependency and distance attenuation factor, as shown in the following formula:

[0042]

[0043] in, It depends on weight. Electrical dependence refers to the proportion of the input current of downstream equipment to the output current of upstream equipment, reflecting the dominant relationship of energy flow. It is the distance attenuation factor, which is calculated based on the actual distance between devices and is used to measure the degree of influence of distance on the interdependence between devices.

[0044] In some embodiments, the electrical dependence is between [0, 1], where a larger electrical dependence indicates a tighter energy transfer dependency. The calculation formula is as follows:

[0045]

[0046] in, It is the operating current of downstream equipment (such as the circuit breaker coil current). This refers to the output current of the upstream equipment (such as the secondary current of a transformer). The formula for calculating the distance attenuation factor is:

[0047]

[0048] in, It is the distance decay factor; This refers to the actual distance between devices, in meters. Weighted summation of electrical dependence and actual distance between devices can resolve distortion scenarios where there is strong electrical dependence but a long physical distance. For example, if the electrical dependence between a transformer and a remote circuit breaker is 0.9, but the distance is 20m, the corresponding distance attenuation factor is 0.018, and the dependence weight is 0.635, corresponding to a medium risk level.

[0049] In some embodiments, a threshold range [0.3, 0.8] is preset based on historical experimental data. If the dependency weight is less than 0.3, it indicates weak dependency, downstream equipment can operate independently, and its cascading failure risk level is low risk. If the dependency weight is within the threshold range, it indicates moderate dependency, upstream failure causes partial failure of downstream equipment, and its cascading failure risk level is medium risk. If the dependency weight is greater than 0.8, it indicates heavy dependency, downstream equipment heavily relies on upstream power supply, upstream failure will lead to downstream paralysis, and its cascading failure risk level is high risk.

[0050] In this embodiment, a central device is used, and related devices are added to a set of associated devices, with one associated device per device. Specifically, devices with moderate and heavy dependencies are added to the set of directly associated devices, while devices with weak dependencies are added to the set of indirectly associated devices. Based on the associated device sets, specific device nodes are obtained, association edges are established, and dependency weights are marked to form an initial network graph.

[0051] In some embodiments, devices in the directly related set have close dependencies on each other. If one device malfunctions, directly related devices are highly likely to be affected quickly, with a high probability of failure. Devices in the indirectly related set, however, have relatively weaker dependencies, and the impact of a device malfunction is buffered and delayed. By distinguishing between these two sets, the propagation boundaries of faults on devices with different levels of dependency can be clearly defined, allowing for rapid and accurate determination of the range of devices potentially affected by the fault, and clarifying the direction of fault handling and resource allocation. In practical applications, based on the directly and indirectly related sets, differentiated monitoring frequencies and maintenance plans can be developed for devices in different sets. For devices in the directly related set, due to their higher risk and faster rate of failure, higher-frequency real-time monitoring is required to promptly identify potential problems and take corrective action. For devices in the indirectly related set, the monitoring frequency can be appropriately reduced, but still requires continued attention to ensure timely detection when faults show a tendency to spread. This allows for the rational allocation of monitoring resources, improved overall monitoring efficiency, and reduced operation and maintenance costs.

[0052] In this embodiment, the dependency weight value is calculated based on electrical dependence and physical distance, which breaks through the limitation of traditional topology networks that only focus on physical connections. By quantifying the dominance of energy flow, the critical path of fault propagation is accurately located, enabling cascading and further improving fault blocking efficiency.

[0053] S2: Monitor device status in real time, identify abnormal devices and corresponding abnormal time nodes; obtain the first association group and the second association group based on the associated device set; overlap the first association group and the second association group with the initial network graph to obtain the main path and the secondary path.

[0054] The associated device set includes a direct association set and an indirect association set; based on the direct association set, devices associated with the abnormal device are obtained and form a first association group; based on the indirect association set, devices associated with the abnormal device are obtained and form a second association group.

[0055] In some embodiments, the device status includes two states: normal and abnormal. Abnormal refers to the detection of an abnormal state, which is not a fault state. The abnormal device and its corresponding abnormal time point are determined. Here, the abnormal time point is the time point when the abnormality is detected. In reality, the actual occurrence of the abnormality and the time point when the abnormality is detected are not the same, depending on the frequency and accuracy of the monitoring equipment. In subsequent calculations, the actual occurrence time point of the abnormality is inferred from the data indicators of the detected abnormality. If the accuracy of the monitoring equipment is guaranteed, the time difference can be ignored, and this application does not impose specific limitations here.

[0056] In some embodiments, the first association group and the second association group are overlapped with the initial network graph to obtain the main path and the secondary path, respectively. This includes: determining the specific location of the abnormal device in the initial network graph; extracting device nodes from the first association group; connecting the device nodes to form the main path; and extracting device nodes from the second association group; connecting the device nodes to form the secondary path. The device node refers to the corresponding device, with one device node corresponding to one device.

[0057] S3: Calculate the risk intensity value of each device in the main path, obtain the handling efficiency and evaluation indicators at different time points, and obtain the optimal intervention time; calculate the risk delay growth curve of the secondary path, identify the inflection point of risk growth, and determine the optimal prevention time.

[0058] The risk intensity value refers to the dynamic failure risk level of the equipment under specific spatiotemporal coordinates, and the calculation formula is as follows:

[0059]

[0060] in, It is a risk intensity value; It is the dependency weight of the primary path device i; This is the propagation attenuation coefficient; it is set to 0.05 for high-voltage equipment and 0.02 for low-voltage equipment. It is the current time. Abnormal time points of abnormal equipment Time difference; It is the total distance attenuation. It is the distance between device i and the abnormal device; the risk intensity value uses the superposition of risk values ​​of multiple devices to reflect the chain transmission characteristics of the fault along the electrical path and quantify the dynamic evolution of the fault in the spatiotemporal continuum.

[0061] The specific spatiotemporal coordinates are the spatial coordinates and time axis coordinates of the equipment. The spatial coordinates are used to locate the physical location of the fault propagation, calculate the total distance attenuation, determine the fault impact range based on the characteristic of exponential attenuation with distance, and then accurately delineate the isolation boundary. The time axis coordinates are used to quantify the dynamic evolution process of the fault over time. The risk intensity value changes non-uniformly with time. Based on the time axis coordinates, the optimal intervention time can be quantitatively predicted to obtain the highest handling efficiency.

[0062] In some embodiments, the handling efficiency and evaluation indicators at different time points are obtained to determine the optimal intervention time. Specifically, based on the current waveform data of the abnormal device obtained from the dataset, the accumulated fault energy is calculated, and evaluation indicators are set to assess and verify the handling efficiency at different time points to determine the optimal intervention time. The accumulated fault energy refers to the fault energy generated over time, used to measure the destructive capacity of the fault.

[0063] In some embodiments, fault current modeling is performed based on current waveform data, using the following formula:

[0064]

[0065] in, It was through The accumulation of fault energy over time quantifies the destructive force of the fault and determines the degree of equipment damage; the energy accumulation rate decreases non-linearly over time. It is the equivalent fault resistance, which reflects the arc characteristics at the fault point. The smaller the value, the more severe the short circuit. It is set to 0.2 ohms. It is the initial fault current peak value, which determines the energy release intensity and is positively correlated with the short-circuit capacity; It is the attenuation coefficient, which controls the rate of current decay and reflects the damping characteristics of the system; it is set to 0.15. The angular frequency determines the oscillation frequency. Set to 314 radians per second, it accurately simulates the oscillation decay characteristics of the fault current. Actual measurement data shows that… For every 0.1 increase in the value, the fault energy decreases by 37%. It is the integral time variable, representing the time infinitesimal element of the energy accumulation process, where, This represents the time difference.

[0066] Set the processing efficiency optimization function as follows:

[0067]

[0068] in, It is the time-based disaster mitigation efficiency, used to obtain the optimal intervention window corresponding to the peak efficiency point; It reflects the theoretical disaster reduction effect of intervention measures. It is the fault energy accumulation corresponding to 10ms, used to control the upper limit of the prediction time; This represents the cost of quantifying the handling action (such as the energy consumption of the circuit breaker mechanism), which, based on actual experimental measurements, can be set to 0.3. By taking the derivative and setting it to zero, we can obtain the extreme points.

[0069] The evaluation indicators include energy blocking rate and energy consumption reduction rate per unit time. Efficiency is highest when the energy blocking rate equals the average disaster reduction efficiency. Evaluation results are obtained at different time points, and the optimal intervention time, i.e., the efficiency peak point, is determined based on the evaluation indicator data. This indicates that the optimal intervention time, i.e. the efficiency peak point, is reached 2 milliseconds (ms) after the abnormal time point. In this application, ms refers to milliseconds.

[0070] In this embodiment, the risk intensity value is obtained by superimposing the risk values ​​of multiple devices based on the obtained dependency weights, quantifying the chain reaction across devices, and accurately calculating the total risk of the fault chain. This solves the problem of misjudging the risk of remote devices caused by using only electrical weights in traditional models. At the same time, a time axis coordinate is added to predict the optimal intervention time node, which solves the problem that traditional methods using static risk values ​​cannot predict delay faults in circuits.

[0071] In some embodiments, calculating the risk delay growth curve of the secondary path includes: obtaining the device nodes in the secondary path and their corresponding initial dependency weights; calculating the risk secondary value corresponding to the time change of each device node; and generating a risk delay growth curve based on the risk secondary value and the time node. Specifically, the initial value of the risk secondary value for each device is 0. The risk secondary value at the corresponding time is calculated based on the initial dependency weight, environmental acceleration factor, and line degradation rate. The calculation formula is as follows:

[0072]

[0073] in, It was through The risk sub-value at any given moment differs from that of the main path in that the time interval used here is in hours. It is the initial dependency weight of the secondary path device y; It is the environmental acceleration factor, which is determined based on the impact of abnormal equipment on the environmental conditions of subsequent equipment. It can be determined based on historical experimental data and the specific analysis of the abnormal situation. For example, for abnormalities after lightning strikes, the environmental acceleration factor is set to 1.5. This application does not impose any specific restrictions on this. The rate of line degradation is determined by the deterioration of the line caused by abnormal equipment. This can be identified through oil chromatography analysis and the content of specific gases. The unit is the degree of degradation per hour, for example, set to 0.02. This application does not impose specific limitations on this. For the analysis of equipment on secondary paths, the time interval can be extended according to the actual situation to improve the data analysis effect. Based on the risk sub-value at different time intervals, the risk change rate is calculated by taking the third derivative of the risk sub-value to identify the risk growth inflection point (including the critical point where the risk growth rate changes from increasing to decreasing or from decreasing to increasing), and the optimal prevention time is determined. The inflection point condition is inversely proportional to the line degradation rate. If the line degradation rate is 0.02, the corresponding time is 50 hours. Less than 50 hours indicates an accelerated risk accumulation stage, and more than 50 hours indicates a slowing risk accumulation stage. Therefore, 50 hours after the anomaly is identified is the optimal prevention time and the last time window for preventive maintenance, effectively avoiding secondary failures. After obtaining the optimal intervention time and optimal prevention time, preventive intervention is carried out on the main path. After the main path action occurs, the parameter values ​​of the associated secondary paths are updated, the optimal prevention time is calculated, and the linkage optimization of the main and secondary paths is realized.

[0074] S4: If the risk intensity value is not greater than the high threshold, between the optimal intervention time and the optimal prevention time, identify alternative equipment based on the status of the abnormal equipment and alternative indicators, and construct a complementary characteristic matrix to determine alternative transmission paths; the alternative indicators include functional compensation degree, capacity matching rate and switching delay.

[0075] The high threshold is preset to measure the severity of anomalies in abnormal equipment. It needs to be dynamically adjusted based on historical experimental data. For example, setting the high threshold to 0.6 means that if the risk intensity value is not greater than 0.6, the current anomaly is not considered extremely urgent; if the risk intensity value is greater than 0.6, the current anomaly is extremely urgent and highly likely to cause serious losses, requiring immediate measures such as equipment or line replacement. For cases where the risk intensity value is not greater than the high threshold, further adjustments should be made after the initial intervention (intervention measures implemented at the optimal intervention time). The timing of these adjustments is between the optimal intervention time and the optimal prevention time, ensuring rapid reduction of losses while further eliminating the probability of anomalies occurring.

[0076] In some embodiments, the functional compensation degree is a measure of the percentage by which the alternative equipment functionally covers the critical capabilities of the faulty equipment, ensuring the functional integrity of the system after replacement (e.g., whether the backup transformer can perform the same voltage conversion function in the event of a transformer failure). This is achieved by extracting a list of core functions (e.g., voltage transformation, reactive power compensation) from the equipment technical manual and verifying functional availability using real-time monitoring data (e.g., the results of the standby equipment's no-load test). The capacity matching rate is the ratio of the alternative equipment's capacity to the faulty equipment's requirements, preventing overload risks (e.g., verifying whether the standby equipment can handle a 480MVA load in the event of a 500kV transformer failure). The switching delay is the time required to switch from the faulty equipment to the alternative equipment, including mechanical action time (circuit breaker opening and closing), grid synchronization time, and system stabilization time, quantifying the impact of the switching process on system continuity.

[0077] In some embodiments, constructing a complementary characteristic matrix based on the state of the abnormal equipment includes: the complementary characteristic matrix is ​​a three-dimensional matrix, including candidate replacement equipment, abnormal state values, and a comprehensive value; analyzing the abnormal state of the abnormal equipment to determine the functional loss rate, capacity gap, and risk diffusion rate, and performing a weighted summation (e.g., weights set to 0.4 for the functional loss rate, 0.4 for the capacity gap, and 0.2 for the risk diffusion rate) to obtain the abnormal state value. The comprehensive value is the difference between the replacement index value and the abnormal state value of each candidate replacement equipment. The larger the difference, the greater the replacement capability of the candidate replacement equipment. If the difference is less than 0, it cannot be replaced, and in the complementary characteristic matrix, all candidate replacement equipment are greater than 0.

[0078] The functional loss rate refers to the proportion of critical functions lost by abnormal equipment to its original total number of functions. It measures the degree of system function degradation caused by the fault. For example, a transformer originally has five major functions: voltage transformation, reactive power compensation, harmonic suppression, insulation isolation, and overload protection. After an anomaly is discovered, only the voltage transformation and harmonic suppression functions can be used normally. The functional loss rate is 60%, and backup equipment must be activated to compensate for the missing functions. The capacity gap is the ratio of the difference between the current load demand of the abnormal equipment and the maximum carrying capacity of the replacement equipment. It is used for overload risk warning and safety warning. For example, if a faulty transformer has a load demand of 480 MVA (megavolt-amperes, a unit of power) and the maximum capacity of the backup transformer is 420 MVA, then 60 MVA of non-critical load needs to be cut off. The risk diffusion rate is the rate of change of the main path risk intensity value over time.

[0079] In some embodiments, determining an alternative transmission path includes: determining several candidate paths based on a complementary characteristic matrix; determining the initial replacement device for each candidate path based on the connection relationship between devices; retaining candidate paths corresponding to the initial replacement devices with a dependency weight greater than 0.6 for the abnormal device; and arranging the dependency weights of subsequent replacement devices in descending order and selecting them sequentially to form an alternative transmission path. Specifically, based on the dependency weights of the initial replacement devices, after selecting a portion of qualified candidate paths, the abnormal device (the initial replacement device most closely connected to the abnormal device) is determined as the first to be replaced. For subsequent replacement devices, they are arranged in descending order based on their dependency weights and selected sequentially. Only one device is selected for each line node, ultimately forming an alternative transmission path. The transmission process of the current power line is adjusted according to the formed alternative transmission path to further ensure power safety. For example, the oil temperature monitoring data of the main transformer (equipment number T1) in a 500kV substation is abnormal: real-time temperature: 95℃, temperature rise rate: 8℃ / min. Sustained high temperature may cause the insulating oil to decompose and trigger a deflagration. Construct a topology network, calculate the dependency weights of related devices, and form direct and indirect association sets; locate abnormal time nodes, including the alarm time of the monitoring system and inferring the actual start of the abnormality; calculate the risk intensity value of each device in the main path to obtain the optimal intervention time and solve the problem of fire; calculate the risk delay growth curve of the secondary path, identify the inflection point of risk growth, determine the optimal prevention time, and then determine the alternative transmission path to eliminate the risk of fire in the future.

[0080] In this embodiment, by quantifying electrical dependence and physical distance, it overcomes the limitation of traditional topology networks that only focus on physical connections, accurately pinpointing critical paths for fault propagation. Dynamic risk intensity values ​​are superimposed on the risks of multiple devices, quantifying cross-device chain reactions and resolving the problem of misjudging risks of remote devices. It predicts the optimal intervention time, addressing the inability of static risk models to handle delayed faults; the identification of the inflection point of secondary path risk growth provides a final time window for preventative maintenance, avoiding secondary faults. By calculating and determining alternative transmission paths between the optimal intervention time and the optimal prevention time, it protects currently abnormal equipment, further reducing the occurrence of faults, preventing secondary propagation of faults, and eliminating greater power losses. For anomaly detection in power systems, especially transformers in substations, it achieves the effects of accurate monitoring, timely early warning, and anomaly handling.

[0081] Example 2: The above example calculates the optimal intervention time using a fault energy accumulation model and the derivative extreme point. However, power system faults have strong nonlinear characteristics, which leads to a deviation between the theoretical optimal time and the actual effect. This example makes further improvements based on the above content.

[0082] Continue to refer to Figure 1The method further includes S5: acquiring historical intervention data records, generating high-dimensional state vectors and defining a set of intervention actions, and constructing a digital twin model; based on the historical intervention data records and the digital twin model, simulating the response process under different fault scenarios, and introducing random disturbances, obtaining the best intervention actions and corresponding intervention times under different abnormal states according to the reward function index; the random disturbances include new energy output fluctuations, load mutations and sensor noise, etc., to enhance the generalization ability of the intelligent agent.

[0083] The historical intervention data records include fault type, occurrence time, intervention action, loss result, and time tag. The time tag includes the theoretical optimal intervention time and the actual optimal time. The actual optimal time is the time obtained by expert analysis after the anomaly occurs.

[0084] The high-dimensional state vector includes equipment status, risk assessment indicators, and environmental parameters. The equipment status can be obtained from the basic dataset and includes at least basic parameters such as transformer temperature, line current, and voltage deviation. The risk assessment indicators include the trend of risk intensity value change, the propagation speed of the main path, and the delay change speed of the secondary path. The environmental parameters include the proportion of new energy output (wind power / photovoltaic), system frequency fluctuations, and basic environmental parameters (temperature, humidity, irradiance, etc.).

[0085] The set of intervention actions includes early intervention, delayed intervention (extending observation before intervention), and action granularity. The action granularity is a fixed time step, which is discretized according to the time step (e.g., 1ms) to intelligently select the specific number of steps to delay or advance.

[0086] The reward function indicators include disaster reduction benefits (proportion of reduced failure losses), operational costs (number of circuit breaker trips, equipment downtime) and system stability (voltage fluctuations, frequency deviations).

[0087] In some embodiments, for example, fault records of the substation over the past three years are collected, including fault type, occurrence time, intervention action, and loss outcome; two types of time labels are used: theoretical optimal intervention time and actual optimal time; a digital twin model is constructed: replicating the real substation topology (transformer / circuit breaker / transmission line), and setting the topology and parameters (such as impedance and capacity); fault scenarios are preset, covering 80% of common fault types (such as single-phase grounding and phase-to-phase short circuit), and three severity levels (mild, medium, and severe) are set for each scenario, for example, setting a "lightning strike + short circuit" combined fault with an initial risk value of 0.5. A Deep Q-Network (DQN) is selected as the reinforcement learning algorithm framework, and the neural network weights are initialized (randomly or pre-trained). The capacity of the experience replay pool is set to store historical interaction data. At each time step, the agent observes the current state vector (such as transformer temperature 85°C, line current 1.2 times the rated value, and risk intensity value 0.4). Actions are selected using an ε-greedy strategy, a classic exploration-balancing method in reinforcement learning. This strategy involves randomly exploring the action space with probability ε and selecting the currently known optimal action (i.e., the action with the highest estimated evaluation value based on prior experience) with probability 1-ε, thus consolidating the learning outcomes. The selected intervention action (e.g., tripping 20ms in advance) is executed, and the system response is simulated in a simulated environment (e.g., fault current is cut off, load is transferred to a backup line). The immediate reward and next state returned by the environment are observed. Positive rewards are given for several pre-successful attempts (fault loss reduction >30%); negative rewards (i.e., penalties) are given for several pre-failure attempts (fault propagation); and negative rewards are given if maintaining the status quo leads to fault deterioration. The system state is updated to the next time step (e.g., temperature drops to 80℃, current returns to normal), and the state-action-reward-new state quadruple is recorded in the experience replay pool. A batch of samples (e.g., 64 samples) is randomly sampled from the experience replay pool, and the target evaluation value is calculated. The target network parameters are periodically (e.g., every 1000 steps) synchronized to the main network to stabilize the training process. The trained agent is run on an independent test set (fault scenarios not included in the training), and the intervention success rate (the proportion of cases successfully preventing fault propagation) and average disaster reduction benefits are statistically analyzed. Comparison with benchmark methods: performance is compared with the derivative extreme point method and fixed-time intervention method (e.g., unified tripping 100ms after a fault) in Example 1. If the test results are unsatisfactory (e.g., intervention success rate <90%), the reward function weights are adjusted (e.g., increasing the disaster reduction benefit ratio to 0.7) or the number of training rounds is increased. High-reward samples are prioritized for sampling to accelerate learning in key scenarios. The trained agent is embedded into the substation monitoring system to receive equipment status data in real time and output intervention suggestions. Network parameters are continuously updated based on actual intervention effects to adapt to equipment aging or topology changes. The above is merely an example of setting the reward function and training process; using reinforcement learning for training is a well-known technique, and this application will not elaborate further.

[0088] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0089] This application establishes a digital twin model to accurately capture the dynamic characteristics of the fault process by simulating random disturbances, avoiding misjudgments caused by static assumptions in traditional models; it comprehensively represents the operating state of the power system, balances disaster reduction benefits with operational costs, and dynamically determines the optimal intervention actions and intervention times for different abnormal states of abnormal equipment; it dynamically adjusts intervention actions based on real-time state vectors to adapt to complex scenarios such as load fluctuations and changes in renewable energy output; and it fully leverages expert experience from historical intervention records through an experience playback pool and target network synchronization mechanism to improve training efficiency and strategy stability.

[0090] Example 3: The above examples optimize the dynamic intervention time of equipment in a single substation through reinforcement learning. However, modern power systems are characterized by multi-station interconnection and regional coordination. The defense strategy of a single station may fail due to ignoring cross-station associated risks (such as power flow transfer and voltage collapse caused by faults in adjacent substations).

[0091] Continue to refer to Figure 1 The method further includes S6: acquiring the target area and several substations within the target area, and determining a central station; acquiring real-time data of adjacent stations, determining cross-station dependency weights based on adjacent station tie line parameters, and dividing the collaborative defense area; when abnormal equipment is detected, identifying the abnormal state, acquiring the optimal intervention action and corresponding intervention time of the central station according to step S5, determining the cross-station propagation probability and resource requirements of the central station, and generating a global management plan based on the collaborative defense area to determine global risk constraints.

[0092] In some embodiments, a central site and its defense radius are determined based on the basic dataset of each substation in the target area. Specifically, the structural importance coefficient of each substation is calculated based on the power grid topology data, real-time load data of the substation, and historical fault records (the three data are normalized and weighted to obtain the structural importance coefficient), and the optimal central site is determined. The real-time data are the operating status parameters of each substation.

[0093] The defense radius is formed by determining the collaborative scope based on the capacity of the central substation. In a power collaborative defense system, the defense radius refers to the dynamic range of action defined by the central substation, based on its power supply capacity, grid topology, and equipment interrelationships. This radius determines which adjacent substations need to be included in the same collaborative defense area, enabling cross-site resource scheduling, joint risk prevention, and unified decision-making. Based on geographical distance (e.g., 30km, 20km) or electrical distance (e.g., equivalent impedance value), and dynamically adjusted according to the central substation capacity, this application does not impose specific limitations. Based on a power grid geographic information system (GIS), substation clusters with shared power supply areas are identified. For example, a region may contain five 220kV substations (A / B / C / D / E). GIS analysis determines that their power supply range covers urban areas and suburbs.

[0094] The adjacent station tie line parameters include voltage level ratio and power transmission ratio. Cross-station dependency weights are calculated based on the adjacent station tie line parameters and the dependency weight calculation method in Example 1. Cooperative defense zones are divided based on the cross-station dependency weights and the minimum defense radius of the remaining stations. A clustering method is used, with the optimization objective being to maximize the connection density within each zone and minimize inter-zone connections. These cooperative defense zones include strongly coupled, moderately coupled, and weakly coupled zones. For example, five substations are divided into two defense zones: {A, B, C} and {D, E}. Cross-zone power flow at zone boundaries is monitored in real time, and cross-zone transmission paths are automatically isolated in case of anomalies. The risk resistance index for each zone is calculated as: redundant line ratio within the zone × 0.4 + critical equipment reserve rate × 0.3 + resource self-sufficiency rate (resources available within the zone / total demand) × 0.3.

[0095] In some embodiments, determining the cross-site propagation probability and resource requirements of the central site, and generating a global management plan based on global risk constraints determined by the collaborative defense area, includes: the cross-site propagation probability is the rate of change of the risk intensity value after the central site's intervention action within a unit time, wherein the unit time is dynamically set according to the scope of the target area, and its value is less than the optimal prevention time of the central site. The resource requirements include, but are not limited to, emergency repair resources, backup equipment, and human resources. The global risk constraints are risk propagation constraints, fairness constraints, and timeliness constraints. Specifically, the risk propagation constraint means that any resource allocation must not cause the risk value of adjacent areas to increase by more than 15%; the fairness constraint means that the resource allocation is fair and avoids resource accumulation; the timeliness constraint means that the resource allocation time should be minimized.

[0096] In some embodiments, a global management scheme is generated based on global risk constraints determined by the collaborative defense zone. The power system is then monitored, managed, and anomaly handled according to this scheme, with real-time effects tracked and key indicators monitored: fault propagation range (whether it breaches the defense zone), system recovery time (from anomaly occurrence to stable operation), and resource utilization rate (actual / planned usage ratio). Based on the hierarchical collaborative architecture of the dynamic center and adaptive defense zones, spatiotemporal collaborative optimization of defense strategies is achieved, constructing a cross-site defense system supporting high-proportion energy access to adapt to the needs of new power systems.

[0097] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0098] This application clarifies fault propagation paths through dynamic dependency weighting and propagation probability calculations, avoiding defense blind spots caused by neglecting cross-site correlations in traditional methods. Based on quantified resource requirements and global constraints, it achieves on-demand resource scheduling, avoiding local resource accumulation or insufficiency. Through regional division and risk constraint mechanisms, it ensures that single-site defense actions do not trigger risk increases at other sites, achieving a shift from local optima to global optima. It constructs an adaptive collaborative defense region, breaking through the limitations of traditional fixed topology models, and proposes three-dimensional constraint optimization to achieve a balance between fairness and efficiency in cross-site resource allocation, further improving the accuracy of power safety monitoring and the efficiency of anomaly handling.

[0099] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A power safety monitoring and management method, characterized in that, include: S1: Obtain the basic dataset, calculate the dependency weights between devices and obtain the set of associated devices, and construct the initial network graph based on the set of associated devices; The basic dataset includes the electrical parameters, physical location, and control logic of the power equipment; The dependency weight is a parameter value used to quantify the energy transfer dependency between devices; Based on the set of associated devices, obtain the specific device nodes, establish associated edges, and mark the dependency weights to form an initial network graph; The associated device set includes a direct association set and an indirect association set; based on the direct association set, devices associated with the abnormal device are obtained and form a first association group; based on the indirect association set, devices associated with the abnormal device are obtained and form a second association group. S2: Monitor device status in real time, identify abnormal devices and corresponding abnormal time points; obtain the first association group and the second association group based on the associated device set; overlap the first association group and the second association group with the initial network graph to obtain the main path and the secondary path, respectively. S3: Calculate the risk intensity value of each device in the main path, obtain the handling efficiency and evaluation indicators at different time points, obtain the current waveform data of abnormal devices based on the basic dataset, calculate the accumulated fault energy, set evaluation indicators to evaluate and verify the handling efficiency at different time points, and obtain the optimal intervention time. Calculate the risk delay growth curve of the secondary path, identify the inflection point of risk growth, and determine the optimal prevention time; including: obtaining the device nodes in the secondary path and their corresponding initial dependency weights, calculating the risk secondary value corresponding to the time change of each device node, generating the risk delay growth curve based on the risk secondary value and time node; calculating the risk change rate based on the risk secondary value, identifying the inflection point of risk growth, and determining the optimal prevention time; S4: If the risk intensity value is not greater than the high threshold, between the optimal intervention time and the optimal prevention time, identify alternative equipment based on the status of abnormal equipment and alternative indicators, construct a complementary characteristic matrix, and determine alternative transmission paths. The complementary characteristic matrix is ​​a three-dimensional matrix, including candidate replacement devices, abnormal state values, and a comprehensive value. The abnormal state of the abnormal devices is analyzed to determine the functional loss rate, capacity gap, and risk diffusion rate, and the abnormal state value is obtained by weighted summation. The comprehensive value is the difference between the replacement index value and the abnormal state value of each candidate replacement device. The larger the difference, the greater the replacement capability of the candidate replacement device. Determining alternative transmission paths includes: identifying several candidate paths based on the complementary characteristic matrix; determining the initial alternative device for each candidate path based on the connection relationship between devices; retaining candidate paths corresponding to initial alternative devices with a dependency weight greater than 0.6 for abnormal devices; arranging the dependency weights of subsequent alternative devices in descending order; and selecting them sequentially to form alternative transmission paths. The risk intensity value refers to the dynamic fault risk level of the equipment under specific spatiotemporal coordinates; the specific spatiotemporal coordinates are the spatial coordinates and time axis coordinates of the equipment. The spatial coordinates are used to locate the physical location of fault propagation, and the time axis coordinates are used to quantify the dynamic evolution process of the fault over time; the fault energy accumulation refers to the fault energy generated over time, which is used to measure the fault's destructive capability; the evaluation indicators include energy blocking rate and energy consumption reduction rate per unit time. The initial risk sub-value for each device is 0. The risk sub-value at the corresponding time is calculated based on the initial dependency weight, environmental acceleration factor, and line degradation rate.

2. The power safety monitoring and management method as described in claim 1, characterized in that, The method further includes: S5: acquiring historical intervention data records, generating high-dimensional state vectors and defining a set of intervention actions, and constructing a digital twin model; based on historical intervention data records and the digital twin model, simulating the response process under different fault scenarios, introducing random perturbations, and obtaining the best intervention actions and corresponding intervention times under different abnormal states according to the reward function index; S6: Obtain the target area and several substations within the target area, and determine a central station; obtain real-time data of adjacent stations, determine cross-station dependency weights based on the adjacent station tie line parameters, and divide the collaborative defense area; when abnormal equipment is detected, identify the abnormal state, obtain the optimal intervention action and corresponding intervention time of the central station according to step S5, and determine the cross-station propagation probability and resource requirements of the central station, and generate a global management plan based on the collaborative defense area to determine global risk constraints.

3. The power safety monitoring and management method as described in claim 1, characterized in that, The first association group and the second association group are respectively overlapped with the initial network graph to obtain the main path and the secondary path, including: determining the specific location of the abnormal device in the initial network graph, extracting the device nodes in the first association group, and connecting the device nodes to form the main path; extracting the device nodes in the second association group and connecting the device nodes to form the secondary path.

4. The power safety monitoring and management method as described in claim 1, characterized in that, The high threshold is preset and is used to measure the degree of abnormality of the abnormal device; The substitution metrics include functional compensation degree, capacity matching rate, and switching delay. The functional compensation degree measures the percentage by which the substitution device covers the critical capabilities of the faulty device. The capacity matching rate is the ratio of the capacity of the substitution device to the needs of the faulty device. The switching delay is the time required to switch from the faulty device to the substitution device, including mechanical action time, grid synchronization time, and system stabilization time.

5. The power safety monitoring and management method as described in claim 2, characterized in that, The historical intervention data records include fault type, occurrence time, intervention action, loss result, and time tag. The time tag includes the theoretical optimal intervention time and the actual optimal time. The actual optimal time is the time obtained by expert analysis after the anomaly occurs. The high-dimensional state vector includes equipment status, risk assessment indicators, and environmental parameters; the risk assessment indicators include the trend of risk intensity value change, the propagation speed of the main path, and the delay change speed of the secondary path. The set of intervention actions includes early intervention, delayed intervention, and action granularity. The action granularity is a fixed time step, and the specific number of steps to delay or advance is intelligently selected according to the time step discretization. The reward function indicators include disaster reduction benefits, operational costs, and system stability. The random disturbances include fluctuations in new energy output, sudden load changes, and sensor noise.

6. The power safety monitoring and management method as described in claim 2, characterized in that, Based on the basic dataset of each substation in the target area, a central site and its defense radius are determined; the collaborative range is determined based on the capacity of the central site to form the defense radius. The parameters of the adjacent station tie lines include voltage level ratio and power transmission ratio; the collaborative defense area is divided according to the cross-station dependency weight and the minimum defense radius of the other stations, and a clustering method is adopted. The optimization objective is to maximize the connection density within the area and minimize the connection between areas. The collaborative defense area includes strongly coupled area, medium coupled area and weakly coupled area. The cross-site propagation probability is the rate of change of the risk intensity value after the central site's intervention action within a unit of time. The unit of time is dynamically set according to the range of the target area, and its value is less than the optimal prevention time of the central site.

Images