Intelligent ring main unit power distribution system of primary-secondary fusion ring main unit ring main unit

Abstract

The application discloses a primary and secondary fusion complete ring net box intelligent ring net box power distribution system and belongs to the technical field of ring net box monitoring, which comprises a data processing module, which acquires monitoring parameters of each functional unit of the ring net box; a fault analysis module, which analyzes based on the monitoring parameters and judges whether each functional unit has a fault; an alarm response module, which generates corresponding alarm responses for the functional units with faults based on the judgment result of the fault analysis module; and an execution module, which dynamically executes alarm response scheduling based on each alarm response. The application can simultaneously have explicit fault and implicit fault detection capabilities, can not only timely find real-time faults that have occurred, but also can identify potential hazards of equipment in advance, eliminates the faults in the budding state, and effectively reduces the occurrence rate of power failure accidents.

Description

Technical Field

[0001] This invention belongs to the field of ring network box monitoring technology, specifically relating to an intelligent ring network box power distribution system that integrates primary and secondary ring network boxes. Background Technology

[0002] Integrated primary and secondary ring main units are core infrastructure equipment in distribution network automation systems. They deeply integrate primary switching equipment with secondary intelligent terminals, realizing integrated functions of distribution line segmentation, interconnection, protection, measurement, and monitoring. They are a key support for improving the reliability and intelligence level of distribution network power supply. With the continuous deepening of distribution network construction and transformation in my country, integrated primary and secondary ring main units have been widely used in urban and rural distribution networks. Their operational status directly affects the safe and stable operation of the entire distribution network and the electricity user experience.

[0003] Currently, fault monitoring and alarm scheduling technologies for integrated primary and secondary ring main units still have many shortcomings. In terms of fault detection, traditional monitoring systems mainly rely on fixed threshold alarm mechanisms, which can only detect explicit faults that have already occurred, such as overcurrent, overvoltage, short circuit, and over-temperature. For latent faults that gradually develop during long-term operation, such as insulation aging, contact wear, mechanical jamming, and sealing failure, traditional systems cannot identify them because they do not cause significant parameter exceedances in their early stages. Industry statistics show that over 60% of sudden power outages in distribution networks originate from latent faults that were not detected in time, which has become a major factor affecting the reliability of power supply in distribution networks. Even those few studies that attempt to detect latent faults mostly rely on independent parameter analysis of single functional units, completely ignoring the complex electrical topology and physical spatial coupling relationships between the functional units within the ring main unit. This results in low accuracy and high false alarm rates for latent fault detection, failing to meet the needs of practical engineering applications.

[0004] In terms of alarm dispatching, existing systems generally employ rigid dispatching strategies such as fixed priority or first-come-first-served. Fixed priority strategies pre-set fixed priority levels for different types of faults, making dynamic adjustments impossible based on the actual severity and development of the faults. First-come-first-served strategies handle alarms strictly according to the order in which they occur. Neither strategy considers the causal relationships between faults or the risk of cascading failures. A seemingly low-priority fault may be the root cause of multiple high-priority faults; failure to address these faults first could trigger a domino effect, leading to widespread power outages. When multiple faults occur concurrently, traditional dispatching systems cannot scientifically assess the overall severity of each fault, leaving maintenance personnel overwhelmed by a large volume of alarm information. This often delays the handling of the most urgent and dangerous faults, causing unnecessary economic losses and social impact.

[0005] In view of this, the present invention proposes an intelligent ring network box power distribution system that integrates primary and secondary ring network boxes to solve the above problems. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent ring network box power distribution system that integrates primary and secondary ring network boxes, in order to solve the problems faced in the above-mentioned background art.

[0007] The objective of this invention can be achieved through the following technical solutions: The intelligent ring main unit power distribution system integrating primary and secondary integrated ring main units, wherein the monitoring system includes: The data processing module is used to acquire the monitoring parameters of each functional unit of the ring network box and to perform preprocessing operations on the monitoring parameters. The fault analysis module includes an explicit fault analysis unit and a implicit fault analysis unit. The explicit fault analysis unit analyzes the monitoring parameters of each functional unit to determine whether there is a real-time fault in each functional unit. The implicit fault analysis unit analyzes the monitoring parameters of each functional unit to determine whether there is a implicit fault in each functional unit. An alarm response module generates corresponding alarm responses for functional units with faults based on the judgment results of the fault analysis module. The execution module dynamically schedules alarm responses based on each alarm response.

[0008] Furthermore, the method for determining whether each functional unit has a real-time fault is as follows: The system acquires monitoring parameters of each functional unit in real time, takes the average value of the monitoring parameters within a preset short time period as the judgment parameter of the monitoring parameters, and each functional unit has a preset reasonable operating range for the corresponding monitoring parameters. If the judgment parameter is not within the corresponding reasonable operating range, the corresponding functional unit is judged to have a real-time fault. The difference between the judgment parameter and the reasonable operating range is normalized and used as the corresponding explicit fault index.

[0009] Furthermore, the method for determining whether there are latent faults in each functional unit is as follows: The ideal parameters are obtained by taking the median value of the reasonable operating range of the monitoring parameters; the absolute value of the difference between the judgment parameters of each functional unit and the ideal parameters is calculated and normalized to obtain the relative rate of change of the monitoring parameters. Obtain the relative rate of change of each functional unit, and based on the analysis of the relative rate of change, obtain the state deviation coefficient of each functional unit; Construct electrical topology association unit groups and physical space association unit groups for each functional unit, obtain the monitoring parameters of each functional unit within the electrical topology association unit groups and physical space association unit groups, and obtain the topology coupling coefficient of each functional unit based on the analysis of the monitoring parameters. Based on the state deviation coefficient and topological coupling coefficient of each functional unit, the latent fault index of each functional unit is obtained; if the latent fault index is greater than the preset latent fault index threshold, it is determined that the corresponding functional unit has a latent fault.

[0010] Furthermore, the method for obtaining the state deviation coefficients of each functional unit is as follows: The process involves: acquiring the relative rate of change of each functional unit within a preset time interval; calculating the average of the relative rates of change of all functional units to obtain the population mean rate of change; formulating a population rate of change function based on the population mean rate of change within the preset time interval; integrating the population rate of change function within the preset time interval to obtain the population cumulative value; acquiring the relative rate of change of an individual functional unit within a preset time interval and formulating an individual rate of change function within the preset time interval; integrating the individual rate of change function within the preset time interval to obtain the individual cumulative value; and calculating and normalizing the absolute value of the difference between the population cumulative value and the individual cumulative value to obtain the state deviation coefficient of each functional unit.

[0011] Furthermore, the method for obtaining the topological coupling coefficients of each functional unit is as follows: The functional unit to be calculated is denoted as the target functional unit. The electrical topology associated unit group and the physical space associated unit group are merged to obtain the merged unit group. After removing the duplicate functional units in the merged unit group, the comprehensive neighborhood functional unit set of each target functional unit is obtained. The monitoring parameter sequences of each functional unit are collected synchronously, and the monitoring parameter sequences are discretized into finite state symbol sequences to obtain the symbolic state sequences of each functional unit. Based on the symbolic state sequence of each functional unit, the probability of occurrence of each state symbol is statistically analyzed, and the univariate information entropy of each functional unit is calculated. The target functional unit is combined with each functional unit in the comprehensive neighborhood functional unit set to obtain multiple combination pairs. The mutual information of each combination pair is calculated based on the univariate information entropy. Calculate the forward and reverse transfer entropy of each pair of combinations, and calculate the mean transfer entropy by averaging the forward and reverse transfer entropy. After normalizing the mutual information, perform a fusion operation with the mean transfer entropy to obtain the topological coupling value of each pair of combinations. Finally, weighted summation of the topological coupling values ​​of all pairs of combinations yields the topological coupling coefficient of the target functional unit.

[0012] Furthermore, the method for generating corresponding alarm responses for faulty functional units is as follows: When a functional unit is found to have a real-time fault or a latent fault, the corresponding fault is treated as an alarm response, and an alarm response set is built based on all alarm responses. Specifically, alarm responses to real-time faults are recorded as real-time alarm responses, and alarm responses to latent faults are recorded as latent alarm responses.

[0013] Furthermore, the method for dynamically executing alarm response scheduling is as follows: When both real-time alarm responses and hidden alarm responses exist in the alarm response set, the real-time alarm response is selected first for alarm response scheduling. When there are multiple real-time alarm responses or multiple latent alarm responses, calculate the urgency coefficient of each alarm response, and dynamically schedule the response based on the urgency coefficient.

[0014] Furthermore, the method for dynamically scheduling each alarm response based on the urgency coefficient is as follows: The explicit fault index and the implicit fault index are collectively referred to as the fault index. All real-time alarm responses and all implicit alarm responses are uniformly marked as core fault alarm responses. A fault causal relationship graph is constructed. Other alarm responses that are connected by edges to each core fault alarm response are obtained in the fault causal relationship graph and are recorded as the corresponding associated alarm responses. Associated alarm responses with a comprehensive association strength greater than a preset association strength threshold are selected from the associated alarm responses and recorded as core associated alarm responses. The comprehensive correlation strength of each core associated alarm response is obtained. This comprehensive correlation strength is then multiplied by the fault index of the corresponding core fault alarm response to obtain the correlation fault index for each core associated alarm response. All correlation fault indices corresponding to each core fault alarm response are weighted and summed, then added to its own fault index and normalized to obtain the urgency coefficient of each core associated alarm response. Based on the urgency coefficient, the real-time alarm responses and the latent alarm responses are sorted in descending order to obtain the real-time alarm response scheduling table and the latent alarm response scheduling table. The alarm responses are dynamically scheduled based on the order of the scheduling tables.

[0015] Furthermore, the method for constructing a fault causal relationship graph is as follows: The multidimensional state variables of each functional unit in the ring network box are obtained, and a unified multidimensional state vector is constructed. Based on prior physical knowledge, the differential equations of each state variable are established. Kernel regression estimation is performed on the actual observed time series data of each state variable in the multidimensional state vector. Any two state variables in the multidimensional state vector are matched to obtain multiple directed matching pairs. The differential dependency of each directed matching pair is calculated. Directed matching pairs with differential dependency greater than a preset differential dependency threshold are denoted as differential causal edges. Substitute the real-time observation data of a certain state variable into the corresponding differential equation to obtain the corresponding physical residual; obtain the physical residual corresponding to each differential causal edge; if the physical residual is greater than the preset residual threshold, remove the corresponding differential causal edge to obtain the key differential causal edge. Structural causal model intervention verification is performed on each key differential causal edge to obtain valid causal edges that pass verification; the differential sensitivity, physical residual confidence, and intervention response amplitude corresponding to each valid causal edge are obtained, and weighted summation is performed to obtain the comprehensive correlation strength of each valid causal edge; Using the fault type of each functional unit as nodes, each effective causal edge as a directed edge, and the comprehensive correlation strength as the edge weight, a fault causal correlation graph is constructed.

[0016] The beneficial effects of this invention are: This invention possesses both overt and covert fault detection capabilities. It can not only promptly detect real-time faults that have already occurred, but also identify potential equipment hazards in advance, eliminating faults in their nascent stage and effectively reducing the incidence of power outages. Furthermore, the covert fault detection employs a combination of state deviation coefficients and topological coupling coefficients, considering both the state changes of individual functional units and the electrical and physical coupling relationships between functional units, significantly improving the accuracy and reliability of covert fault judgment. Moreover, by using methods such as information entropy, mutual information, and transit entropy to mine hidden information in monitoring data, it can effectively capture minute changes in equipment state, improving the sensitivity of fault detection. The fault causal correlation graph constructed in this invention ensures that fault correlations conform to the physical laws of equipment operation from the source. It thoroughly eliminates false causal edges through physical residual verification and do-intervention operations. Simultaneously, it integrates multi-dimensional indicators to quantify the comprehensive correlation strength between faults, accurately depicting the causal transmission direction and impact of faults, and clearly identifying high-risk propagation links, providing a reliable foundation for cascading fault prediction and scientific scheduling decisions. Furthermore, based on the fault causal correlation graph, it calculates the urgency coefficient of the fault's severity and the risk of cascading spread. Through a two-layer rule combining real-time alarm absolute priority with alarms of the same type sorted by urgency coefficient, it achieves dynamic scheduling. Priority is updated in real-time according to the fault status, ensuring priority handling of major explicit faults and scientifically addressing concurrent multi-fault scenarios. This significantly optimizes the allocation of operation and maintenance resources, reduces the average fault repair time, and effectively improves the reliability of power distribution networks.

[0017] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a block diagram of the system of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] In one embodiment, an intelligent ring main unit power distribution system integrating primary and secondary ring main units is disclosed, such as... Figure 1 As shown, the monitoring system mainly includes: The data processing module is used to acquire the monitoring parameters of each functional unit of the ring network box and to perform preprocessing operations on the monitoring parameters. The ring main unit's functional units include, but are not limited to, circuit breaker units, load switch units, disconnector units, voltage transformer units, current transformer units, surge arrester units, cable termination units, intelligent terminal units (DTU / FTU), and power supply units. The specific division of functional units is determined based on actual conditions. Monitoring parameters include, but are not limited to, voltage, current, power, temperature, humidity, and insulation resistance of each functional unit, specifically acquired through sensors installed on each functional unit. Preprocessing is performed on the monitoring parameters to improve data quality and provide a reliable data foundation for subsequent fault analysis. Data processing includes data cleaning, such as using the 3σ principle to remove outliers from the original data (i.e., if the difference between a monitoring parameter value and its mean value over the past hour is greater than three times the standard deviation, it is considered an outlier and removed); for missing values, linear interpolation is used to complete them; data filtering, such as using a moving average filter to smooth the cleaned data, with the moving average window size set to 5 sampling points, and the filtered data being the arithmetic mean of the data within the window to eliminate the influence of random noise; and data standardization, such as converting the filtered data into dimensionless standard values ​​to facilitate subsequent analysis and calculation.

[0022] The fault analysis module includes an explicit fault analysis unit and a implicit fault analysis unit. The explicit fault analysis unit analyzes the monitoring parameters of each functional unit to determine whether there is a real-time fault in each functional unit. The implicit fault analysis unit analyzes the monitoring parameters of each functional unit to determine whether there is a implicit fault in each functional unit.

[0023] The working method of the explicit fault analysis unit is as follows: real-time acquisition of monitoring parameters of each functional unit, taking the average value of the monitoring parameters within a preset short time period as the judgment parameter of the monitoring parameters, each functional unit has a preset reasonable operating range for the corresponding monitoring parameters, if the judgment parameter is not within the corresponding reasonable operating range, then the corresponding functional unit is judged to have a real-time fault; and the difference between the judgment parameter and the reasonable operating range is normalized and used as the corresponding explicit fault index.

[0024] Specifically, the preprocessed monitoring parameters of each functional unit are acquired in real time. The average value of the monitoring parameters within a preset short time period is taken as the judgment parameter for that parameter. The preset short time period is 10s-60s, and 30s is preferred in this embodiment. A corresponding reasonable operating range is set for each monitoring parameter of each functional unit in advance. The reasonable operating range is determined based on the equipment's factory technical parameters and historical normal operation data statistics. For example, the reasonable operating range for the circuit breaker contact temperature is [0, 80], and the reasonable operating range for the current transformer secondary current is [0, 5]. Here, the reasonable operating range is also a dimensionless standard value after processing. The judgment parameter is compared with the corresponding reasonable operating range. If the judgment parameter is not within the reasonable operating range, the corresponding functional unit is determined to have a real-time fault. The difference between the judgment parameter and the reasonable operating range is calculated and normalized to obtain the corresponding explicit fault index. For example, if the judgment parameter corresponding to the circuit breaker contact temperature is 90, the difference is 90-80=10. The value range of the explicit fault index is... The larger the value, the more severe the fault.

[0025] The method for determining whether there are hidden faults in each functional unit is as follows: take the median value of the reasonable operating range of the monitoring parameters to obtain the ideal parameters; calculate the absolute value of the difference between the judgment parameters of each functional unit and the ideal parameters and normalize it to obtain the relative change rate of the monitoring parameters. Obtain the relative rate of change of each functional unit, and based on the analysis of the relative rate of change, obtain the state deviation coefficient of each functional unit; Construct electrical topology association unit groups and physical space association unit groups for each functional unit, obtain the monitoring parameters of each functional unit within the electrical topology association unit groups and physical space association unit groups, and obtain the topology coupling coefficient of each functional unit based on the analysis of the monitoring parameters. Based on the state deviation coefficient and topological coupling coefficient of each functional unit, the latent fault index of each functional unit is obtained; if the latent fault index is greater than the preset latent fault index threshold, it is determined that the corresponding functional unit has a latent fault.

[0026] The method for obtaining the state deviation coefficients of each functional unit is as follows: Obtain the relative rate of change of each functional unit within a preset time interval, and calculate the average of the relative rates of change of all functional units to obtain the group average rate of change; based on the group average rate of change within the preset time interval, formulate a group rate of change function; perform an integral operation on the group rate of change function within the preset time interval to obtain the group cumulative value; obtain the relative rate of change of an individual functional unit within a preset time interval, and formulate an individual rate of change function within the preset time interval; perform an integral operation on the individual rate of change function within a preset time interval to obtain the individual cumulative value; calculate the absolute value of the difference between the group cumulative value and the individual cumulative value and normalize it to obtain the state deviation coefficient of each functional unit.

[0027] The method for obtaining the topological coupling coefficient of each functional unit is as follows: the functional unit to be calculated is denoted as the target functional unit, the electrical topology associated unit group and the physical space associated unit group are merged to obtain the merged unit group, and after removing the duplicate functional units in the merged unit group, the comprehensive neighborhood functional unit set of each target functional unit is obtained. The monitoring parameter sequences of each functional unit are collected synchronously, and the monitoring parameter sequences are discretized into finite state symbol sequences to obtain the symbolic state sequences of each functional unit. Based on the symbolic state sequences of each functional unit, the occurrence probability of each state symbol is calculated, and the univariate information entropy of each functional unit is calculated. The target functional unit is combined with each functional unit in the comprehensive neighborhood functional unit set to obtain multiple combination pairs, and the mutual information of each combination pair is calculated based on the univariate information entropy. Calculate the forward and reverse transfer entropy of each pair of combinations, and calculate the mean transfer entropy by averaging the forward and reverse transfer entropy. After normalizing the mutual information, perform a fusion operation with the mean transfer entropy to obtain the topological coupling value of each pair of combinations. Finally, weighted summation of the topological coupling values ​​of all pairs of combinations yields the topological coupling coefficient of the target functional unit.

[0028] Specifically, the median value of the reasonable operating range of the monitoring parameters is taken to obtain the ideal parameters. The absolute value of the difference between the judgment parameters of each functional unit and the ideal parameters is calculated and normalized to obtain the relative change rate of the monitoring parameters. The relative change rate R ranges from [0,1]. The larger the value, the further the parameter deviates from the ideal state.

[0029] Then, the state deviation coefficient of each functional unit is calculated to obtain the relative change rate sequence of each functional unit within a preset time interval. The preset time interval ranges from 1h to 12h, and in this embodiment, 6h is preferred. The arithmetic mean of the relative change rates of all functional units at the same time is calculated to obtain the population mean change rate sequence. Based on the population mean change rate sequence within the preset time interval, the least squares method is used to fit the population change rate function. The population change rate function is integrated within the preset time interval to obtain the population cumulative value. The population cumulative value represents the cumulative amount of the overall operating state change trend of all functional units in the entire ring network box within the preset time interval, reflecting the operating health benchmark of the ring network box system. Similarly, the state deviation coefficient of a single functional unit within the preset time interval is obtained. The relative rate of change sequence within the interval is used to formulate the individual rate of change function within the preset time interval. The individual rate of change function is integrated within the preset time interval to obtain the individual cumulative value. The individual cumulative value represents the cumulative amount of the change trend of the operating state of a specific functional unit within the preset time interval, reflecting the independent health status evolution process of the unit. The absolute value of the difference between the group cumulative value and the individual cumulative value is calculated and normalized to obtain the state deviation coefficient of each functional unit. The state deviation coefficient measures the degree of deviation of the operating state of an individual functional unit from the average operating state of all functional units in the entire ring network box. The larger the value, the greater the difference between the operating state of the functional unit and other normal functional units, and the higher the possibility of hidden faults.

[0030] Next, calculate the topology coupling coefficient of each functional unit, and denote the functional unit to be calculated as the target functional unit. Construct electrical topology association unit groups and physical space association unit groups for each functional unit. Electrical topology association unit groups refer to functional units that have a direct electrical connection with the target functional unit. For example, the electrical topology association units of a circuit breaker unit include the incoming current transformer, the outgoing load switch, and the bus unit. Physical space association unit groups refer to functional units that are physically adjacent to the target functional unit, such as circuit breakers and disconnectors in the same bay. Merge the electrical topology association unit groups and physical space association unit groups, remove duplicate functional units, and obtain the comprehensive neighborhood functional unit set of the target functional unit. Synchronously collect data from the target functional unit. The monitoring parameter sequence of each unit in the Yuanhe integrated neighborhood functional unit set is sampled at a frequency of 5Hz. The monitoring parameter sequence is discretized into a finite state symbol sequence. An equidistant discretization method can be used to divide the value range of each monitoring parameter into 5 equidistant intervals, corresponding to state symbols S0 (extremely low), S1 (low), S2 (normal), S3 (high), and S4 (extremely high). For example, the contact temperature range [0,100] is divided into [0,20), [20,40), [40,60), [60,80), and [80,100], corresponding to S0 to S4 respectively. Based on the symbolized state sequence of each functional unit, the probability of occurrence of each state symbol is statistically analyzed, and the univariate information entropy of each functional unit is calculated. The expression is:

[0031] For the i-th state symbol, State symbol The probability of n states appearing in a symbolic state sequence, where n is the total number of state symbols. Let x be the univariate information entropy of the functional unit. The univariate information entropy is used to measure the uncertainty of the state of the functional unit. The target functional unit (X) is combined with each functional unit in the comprehensive neighborhood functional unit set (Y) to obtain multiple combination pairs. The mutual information of each combination pair is calculated based on the univariate information entropy, and the expression is as follows:

[0032] in, Let be the mutual information between the target unit X and the neighboring functional unit Y, and j be the discrete state index of the neighboring functional unit Y. Let j be the j-th state symbol of the neighboring functional unit Y. State symbol The probability of appearing in the symbolic state sequence for State symbol And Y is in the state symbol The joint probability; mutual information is used to measure the correlation between the states of two functional units; Calculate the forward and reverse transfer entropy for each pair of combinations. Transfer entropy measures the direction and intensity of causal information transfer between two functional units. The forward transfer entropy TE(X→Y) represents the amount of information transferred from X to Y, and the reverse transfer entropy TE(Y→X) represents the amount of information transferred from Y to X. The expression is as follows:

[0033]

[0034] Where t is the time step, This represents the state symbol of the neighborhood functional unit Y at time t. This represents the state symbol of target unit X at time t. Let Y be at time t+1 Y at time t is X at time t is The probability of them occurring simultaneously, also Let X be at time t+1. X at time t is Y at time t is The probability of them happening simultaneously as well as These are the corresponding second-order conditional probabilities. as well as These represent the corresponding first-order Markov conditional probabilities. Transmission entropy measures the direction and intensity of causal information transmission between two functional units. The mean transmission entropy is calculated by averaging the forward and reverse transmission entropies. After normalizing the mutual information, this mean is fused with the mean transmission entropy to obtain the topological coupling value for each pair. The expression is:

[0035] in, For mutual information after normalization, To transmit the entropy mean, the topological coupling value indicates the overall coupling health index of the combined pair. The larger the value, the more abnormal the overall coupling relationship between the two units is, the more abnormal the linear correlation or causal transmission between the two is, or the higher the possibility that at least one of the units has a hidden fault. Therefore, the topological coupling values ​​of all combination pairs are weighted and summed to obtain the topological coupling coefficient of the target functional unit. Here, the weight coefficient of each combination pair is independently determined based on the experience of those in the field. The value range of the topological coupling coefficient is [0,1]. When the topological coupling coefficient is larger, it indicates that the overall coupling relationship between the functional unit and all neighboring units is more abnormal, and the possibility of hidden faults in the functional unit is higher.

[0036] Finally, the state deviation coefficients and topological coupling coefficients of each functional unit are weighted and accumulated to obtain the latent fault index of each functional unit. If the latent fault index is greater than the preset latent fault index threshold, the corresponding functional unit is judged to have a latent fault. The weighting coefficients and corresponding latent fault index thresholds can be preset in advance based on the experience and professional knowledge of those in the field, and will not be described in detail here. Therefore, this method can simultaneously possess the ability to detect both explicit and latent faults. It can not only promptly detect real-time faults that have already occurred, but also identify potential hidden dangers in equipment in advance, eliminating faults in their infancy and effectively reducing the incidence of power outages. At the same time, the latent fault detection method combines the state deviation coefficient and the topological coupling coefficient, which considers both the state changes of individual functional units and the electrical and physical coupling relationships between functional units, significantly improving the accuracy and reliability of latent fault judgment. Moreover, by using methods such as information entropy, mutual information, and transfer entropy to mine the implicit information in the monitoring data, it can effectively capture subtle changes in the equipment state and improve the sensitivity of fault detection.

[0037] The alarm response module generates corresponding alarm responses for functional units with faults based on the judgment results of the fault analysis module.

[0038] When a functional unit is determined to have a real-time fault or a latent fault, the corresponding fault is treated as an alarm response, and an alarm response set is constructed based on all alarm responses. The alarm response corresponding to the real-time fault is recorded as the real-time alarm response, and the alarm response corresponding to the latent fault is recorded as the latent alarm response.

[0039] Specifically, when the explicit fault analysis unit of the fault analysis module determines that a functional unit has a real-time fault, or the implicit fault analysis unit determines that a implicit fault exists, the alarm response generation unit automatically triggers the alarm response generation process. The specific steps are as follows: First, standardize the alarm object encapsulation: encapsulate the fault information into a unified format alarm response object, including the following required fields: Functional unit unique identifier ID, used to accurately locate the faulty equipment, such as "BRK-01-01" representing circuit breaker No. 1 in bay No. 1; Fault type code, used to distinguish between manifest faults (code prefix E) and latent faults (code prefix W), such as E-001 representing a circuit breaker overheating manifest fault, and W-003 representing a disconnector mechanism jamming latent fault; Fault index, corresponding to the manifest or latent fault index; Related parameter list, containing the specific monitoring parameter names and values ​​that triggered the fault; Raw data fragments: relevant monitoring data curves for 5 minutes before and after the fault occurred; Then, alarm type marking is performed: alarm responses corresponding to real-time faults are marked as real-time alarm responses, and alarm responses corresponding to latent faults are marked as latent alarm responses. Finally, maintain the alarm response collection: store all generated alarm responses in the alarm response collection, and use a sliding window deduplication mechanism: when the same type of alarm in the same functional unit is triggered repeatedly within 10 minutes, only the latest one is retained, and the fault index and occurrence time are updated.

[0040] The execution module dynamically schedules alarm responses based on each alarm response.

[0041] When both real-time alarm responses and hidden alarm responses exist in the alarm response set, the real-time alarm response is selected first for alarm response scheduling. When there are multiple real-time alarm responses or multiple latent alarm responses, calculate the urgency coefficient of each alarm response, and dynamically schedule the response based on the urgency coefficient.

[0042] Specifically, this includes the absolute priority principle and the dynamic sorting principle of the same type. The absolute priority principle is that the scheduling priority of all real-time alarm responses is absolutely higher than that of all hidden alarm responses. That is, when both real-time alarms and hidden alarms exist in the alarm response set, the system will prioritize processing all real-time alarm responses. After all real-time alarms have been processed or automatically downgraded, the processing flow of hidden alarm responses will be started. The dynamic sorting principle of the same type is that when there are multiple real-time alarm responses or multiple hidden alarm responses, the system will not simply sort them by the time the alarms occurred, but will calculate the urgency coefficient of each alarm response and dynamically schedule them from high to low according to the urgency coefficient.

[0043] The method for dynamically scheduling alarm responses based on urgency coefficients is as follows: Explicit fault indices and implicit fault indices are collectively referred to as fault indices. All real-time alarm responses and all implicit alarm responses are uniformly marked as core fault alarm responses. A fault causal relationship graph is constructed. In the fault causal relationship graph, other alarm responses connected by edges to each core fault alarm response are obtained and recorded as the corresponding associated alarm responses. Associated alarm responses with a comprehensive association strength greater than a preset association strength threshold are selected from the associated alarm responses and recorded as core associated alarm responses. The comprehensive association strength of each core associated alarm response is obtained, and the comprehensive association strength is multiplied by the fault index of the corresponding core fault alarm response to obtain the associated fault index of each core associated alarm response. All associated fault indices corresponding to each core fault alarm response are weighted and accumulated, then added to their own fault index and normalized to obtain the urgency coefficient of each core associated alarm response. Based on the urgency coefficients, each real-time alarm response and implicit alarm response are sorted in descending order to obtain a real-time alarm response scheduling table and an implicit alarm response scheduling table. The alarm responses are dynamically scheduled based on the order of the scheduling tables. Specifically, all real-time alarm responses and all latent alarm responses are uniformly marked as core fault alarm responses, each core fault alarm response corresponding to a fault index. Based on a pre-constructed fault causal relationship graph, for each core fault alarm response, all other alarm responses connected to it by directed edges are searched in the graph, and these are recorded as the associated alarm response set. A preset association strength threshold (which can be adjusted according to actual operating experience) is used to filter associated alarm responses from the associated alarm response set whose comprehensive association strength is greater than the preset association strength threshold, and these are recorded as the core associated alarm response set. For each core associated alarm response in the core associated alarm response set, the comprehensive association strength between it and the core fault alarm response is obtained from the graph. The comprehensive association strength is multiplied by the fault index of the corresponding core fault alarm response to obtain the associated fault index of each core associated alarm response. The associated fault index reflects the degree of risk of the core fault being transmitted to associated faults through causal relationships. The urgency coefficient of each core-related alarm response is obtained by weighting and summing all associated fault indices, adding them to its own fault index, and then normalizing the sum. The urgency coefficient comprehensively considers the severity of the fault itself and the risk of cascading faults that the fault may cause, and can more scientifically reflect the urgency of the alarm. The value range is [0,1]. The closer the value is to 1, the more urgent the alarm is and the more priority it needs to be handled. Therefore, all real-time alarm responses are sorted in descending order of urgency coefficient to generate a real-time alarm response scheduling table, and all hidden alarm responses are sorted in descending order of urgency coefficient to generate a hidden alarm response scheduling table. The system processes alarms in the order of the scheduling table. After each alarm is automatically handled or manually confirmed, the next alarm is automatically processed. When a new alarm is added, an existing alarm is cleared, or the fault index is updated, the system immediately recalculates the urgency coefficient of all alarms, updates the scheduling table, and adjusts the current processing order to complete the dynamic scheduling response of each alarm response.

[0044] The method for constructing a fault causal correlation graph is as follows: Obtain the multidimensional state variables of each functional unit within the ring network enclosure and construct a unified multidimensional state vector. Based on prior physical knowledge, establish differential equations for each state variable. Perform kernel regression estimation on the actual observed time-series data of each state variable in the multidimensional state vector. Match any two state variables in the multidimensional state vector to obtain multiple directed matching pairs. Calculate the differential dependency of each directed matching pair. Dedicated matching pairs with differential dependencies greater than a preset differential dependency threshold are recorded as differential causal edges. Substitute the real-time observation data of a certain state variable into the corresponding differential equation. Obtain the corresponding physical residuals; obtain the physical residuals corresponding to each differential causal edge. If the physical residual is greater than the preset residual threshold, the corresponding differential causal edge is removed to obtain the key differential causal edge; perform structural causal model intervention verification on each key differential causal edge to obtain the valid causal edge that passes the verification; obtain the differential sensitivity, physical residual confidence, and intervention response amplitude corresponding to each valid causal edge, and weight them to obtain the comprehensive correlation strength of each valid causal edge; construct a fault causal correlation graph with the fault type of each functional unit as the node, each valid causal edge as the directed edge, and the comprehensive correlation strength as the edge weight.

[0045] Specifically, key monitoring parameters of all functional units within the ring main unit are extracted to construct a unified multi-dimensional state vector. State variables include circuit breaker contact temperature, three-phase current, three-phase voltage, current transformer secondary current, voltage transformer secondary voltage, surge arrester leakage current, disconnector switch contact temperature, energy storage pressure, ambient temperature and humidity, and SF6 gas density within the cabinet. Based on prior knowledge of power system electromagnetics, thermodynamics, and mechanics, dynamic differential equations are established between these state variables to describe their intrinsic physical relationships. This is not described in detail in the existing technology. For example, the dynamic equation for circuit breaker contact temperature is:

[0046] Where I is the load current, R is the contact resistance, h is the preset convective heat transfer coefficient, A is the heat dissipation area, m is the contact mass, and c is the specific heat capacity. For ambient temperature, For example, the circuit breaker contact temperature; and the dynamic equation for bus voltage is:

[0047] in, This refers to the busbar-to-ground capacitance. For incoming line current, For outgoing current, This refers to the bus voltage. Then, historical time-series data of each state variable over the past 3 months were collected at a sampling frequency of 1Hz. Gaussian kernel regression was used to smooth the time-series data, and the time derivative was calculated to eliminate the influence of noise on the derivative calculation. Gaussian kernel regression is an existing technique and will not be described in detail here. Multiple directed matching pairs were obtained by matching any two state variables in the multidimensional state vector. The differential dependency of each directed matching pair was calculated, and directed matching pairs with differential dependencies greater than a preset differential dependency threshold were denoted as differential causal edges. For example, for a directed matching pair (… Differential dependencies between them ;in, The partial correlation coefficient, State variables The first time derivative, For state variables The first time derivative, To remove and The set of all other state variables besides the one mentioned above; based on experience, a differential dependency threshold is preset, and directed matching pairs with differential dependencies greater than the preset differential dependency threshold are recorded as differential causal edges, thus initially screening out state variable pairs that may have causal relationships; To eliminate spurious differential causal edges that do not conform to physical laws, a physical residual verification mechanism is introduced. For each differential causal edge ( ),Will Historical observation data was substituted into the initial setup. The differential equation, calculate Theoretical prediction value And calculate the average physical residual of the causal edge over the verification period (1 month). ,in To verify the number of sampling points within the period, a residual threshold (e.g., 0.15) is preset based on experience. If the average physical residual is greater than the preset residual threshold, it indicates that the differential causal edge does not match the actual physical process, and the corresponding differential causal edge is removed; otherwise, it is retained and recorded as a key differential causal edge. To further confirm the authenticity of the causal relationship and eliminate spurious correlations caused by confounding variables, a structural causal model intervention verification was performed on each key differential causal edge to obtain valid causal edges that passed the verification, specifically: For each critical differential causal edge ( Construct a structural causal model: Where f is a causal function, Let u be the set of confusing variables, and u be the noise term. Perform do-intervention procedures: fixation The value is a, that is, do( =a), observe The change in probability distribution; The Kolmogorov-Smirnov (KS) test was used to compare the results before and after the intervention. The distribution difference; if the p-value < 0.05, it indicates intervention. Significant changes The distribution of the edges confirms the existence of a causal relationship and is recorded as a valid causal edge; otherwise, it is discarded. The causal function mentioned above can be obtained by fitting historical time-series data. Commonly used regression algorithms such as linear regression, multinomial regression, gradient boosting trees, or shallow neural networks can be used to fit the input variables. With output variables The mapping relationships between them, the set of confounding variables combined with prior knowledge of power system physics (such as known electrical coupling relationships) and existing causal discovery algorithms (such as PC algorithm and FCI algorithm) are used to identify the simultaneously influencing variables from all state variables. and The third variable, the noise term, represents random fluctuations that the model cannot explain. It is usually assumed to follow a normal distribution with a mean of 0. Its distribution parameters are obtained through statistical analysis of the fitting residuals, which is a standard processing step in regression modeling. The Kolmogorov-Smirnov (KS) test is a recognized existing technique in the field of statistics. The p-value is the standard output result of the Kolmogorov-Smirnov (KS) test, which is automatically calculated by existing statistical test algorithms and is used to quantify the significance level of the difference in the distribution of the two groups of data before and after the intervention. All of these are existing techniques and will not be described in detail here. The differential sensitivity, physical residual confidence, and intervention response amplitude corresponding to each effective causal edge are obtained, and weighted summation is used to obtain the comprehensive correlation strength of each effective causal edge. Here, the differential sensitivity is the corresponding differential dependency after normalization, reflecting the sensitivity of the causal relationship. The physical residual confidence is 1 minus the normalized average physical residual. The physical residual confidence is inversely proportional to the average physical residual, reflecting the degree to which the causal relationship conforms to physical laws. The intervention response amplitude is the maximum change amplitude of the observed result variable when performing the do-intervention operation, obtained after normalization, reflecting the actual impact of the causal relationship. The weight coefficients are determined autonomously based on experience, and finally weighted summation is used to obtain the comprehensive correlation strength of each effective causal edge. The greater the comprehensive correlation strength, the closer the causal relationship and the higher the degree of influence of the corresponding fault propagation. Finally, a fault causal correlation graph is constructed with the fault type of each functional unit as the node, each effective causal edge as the directed edge, and the comprehensive correlation strength as the edge weight.

[0048] In this way, the fault causal correlation graph constructed by this invention ensures from the source that fault correlations conform to the physical laws of equipment operation. It thoroughly eliminates false causal edges through physical residual verification and do-intervention operations, while simultaneously integrating multi-dimensional indicators to quantify the comprehensive correlation strength between faults. This not only accurately depicts the causal transmission direction and impact of faults but also clearly identifies high-risk propagation links, providing a reliable foundation for cascading fault prediction and scientific scheduling decisions. Furthermore, based on the fault causal correlation graph, it calculates the urgency coefficient of the fault's own severity and the risk of cascading spread. Dynamic scheduling is achieved through a two-layer rule combining real-time alarm absolute priority with alarms of the same type sorted by urgency coefficient. The priority is updated in real-time according to the fault status, ensuring priority handling of major explicit faults and scientifically addressing concurrent multi-fault scenarios. This significantly optimizes the allocation of operation and maintenance resources, reduces the average fault repair time, and effectively improves the reliability of power distribution networks.

[0049] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0050] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0051] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. An intelligent ring main unit power distribution system integrating primary and secondary integrated ring main units, characterized in that, The monitoring system includes: The data processing module is used to acquire the monitoring parameters of each functional unit of the ring network box and to perform preprocessing operations on the monitoring parameters. The fault analysis module includes an explicit fault analysis unit and a implicit fault analysis unit. The explicit fault analysis unit analyzes the monitoring parameters of each functional unit to determine whether there is a real-time fault in each functional unit. The implicit fault analysis unit analyzes the monitoring parameters of each functional unit to determine whether there is a implicit fault in each functional unit. An alarm response module generates corresponding alarm responses for functional units with faults based on the judgment results of the fault analysis module. The execution module dynamically schedules alarm responses based on each alarm response.

2. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 1, characterized in that, The method for determining whether each functional unit has a real-time fault is as follows: The system acquires monitoring parameters of each functional unit in real time, takes the average value of the monitoring parameters within a preset short time period as the judgment parameter of the monitoring parameters, and each functional unit has a preset reasonable operating range for the corresponding monitoring parameters. If the judgment parameter is not within the corresponding reasonable operating range, the corresponding functional unit is judged to have a real-time fault. The difference between the judgment parameter and the reasonable operating range is normalized and used as the corresponding explicit fault index.

3. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 2, characterized in that, The method for determining whether there are latent faults in each functional unit is as follows: The ideal parameters are obtained by taking the median value of the reasonable operating range of the monitoring parameters; the absolute value of the difference between the judgment parameters of each functional unit and the ideal parameters is calculated and normalized to obtain the relative rate of change of the monitoring parameters. Obtain the relative rate of change of each functional unit, and based on the analysis of the relative rate of change, obtain the state deviation coefficient of each functional unit; Construct electrical topology association unit groups and physical space association unit groups for each functional unit, obtain the monitoring parameters of each functional unit within the electrical topology association unit groups and physical space association unit groups, and obtain the topology coupling coefficient of each functional unit based on the analysis of the monitoring parameters. Based on the state deviation coefficient and topological coupling coefficient of each functional unit, the latent fault index of each functional unit is obtained; if the latent fault index is greater than the preset latent fault index threshold, it is determined that the corresponding functional unit has a latent fault.

4. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 3, characterized in that, The method for obtaining the state deviation coefficients of each functional unit is as follows: The process involves: acquiring the relative rate of change of each functional unit within a preset time interval; calculating the average of the relative rates of change of all functional units to obtain the population mean rate of change; formulating a population rate of change function based on the population mean rate of change within the preset time interval; integrating the population rate of change function within the preset time interval to obtain the population cumulative value; acquiring the relative rate of change of an individual functional unit within a preset time interval and formulating an individual rate of change function within the preset time interval; integrating the individual rate of change function within the preset time interval to obtain the individual cumulative value; and calculating and normalizing the absolute value of the difference between the population cumulative value and the individual cumulative value to obtain the state deviation coefficient of each functional unit.

5. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 3, characterized in that, The method for obtaining the topological coupling coefficients of each functional unit is as follows: The functional unit to be calculated is denoted as the target functional unit. The electrical topology associated unit group and the physical space associated unit group are merged to obtain the merged unit group. After removing the duplicate functional units in the merged unit group, the comprehensive neighborhood functional unit set of each target functional unit is obtained. The monitoring parameter sequences of each functional unit are collected synchronously, and the monitoring parameter sequences are discretized into finite state symbol sequences to obtain the symbolic state sequences of each functional unit. Based on the symbolic state sequence of each functional unit, the probability of occurrence of each state symbol is statistically analyzed, and the univariate information entropy of each functional unit is calculated. The target functional unit is combined with each functional unit in the comprehensive neighborhood functional unit set to obtain multiple combination pairs. The mutual information of each combination pair is calculated based on the univariate information entropy. Calculate the forward and reverse propagation entropy for each pair of combinations. Calculate the average propagation entropy by taking the average of the forward and reverse propagation entropy. Then, normalize the mutual information and merge it with the average propagation entropy to obtain the topological coupling value of each pair of combinations. The topological coupling values ​​of all combination pairs are weighted and summed to obtain the topological coupling coefficient of the target functional unit.

6. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 3, characterized in that, The method for generating corresponding alarm responses for faulty functional units is as follows: When a functional unit is found to have a real-time fault or a latent fault, the corresponding fault is treated as an alarm response, and an alarm response set is built based on all alarm responses. Specifically, alarm responses to real-time faults are recorded as real-time alarm responses, and alarm responses to latent faults are recorded as latent alarm responses.

7. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 3, characterized in that, The method for dynamically executing alarm response scheduling is as follows: When both real-time alarm responses and hidden alarm responses exist in the alarm response set, the real-time alarm response is selected first for alarm response scheduling. When there are multiple real-time alarm responses or multiple latent alarm responses, calculate the urgency coefficient of each alarm response, and dynamically schedule the response based on the urgency coefficient.

8. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 7, characterized in that, The method for dynamically scheduling alarm responses based on urgency coefficients is as follows: The explicit fault index and the implicit fault index are collectively referred to as the fault index. All real-time alarm responses and all implicit alarm responses are uniformly marked as core fault alarm responses. A fault causal relationship graph is constructed. Other alarm responses that are connected by edges to each core fault alarm response are obtained in the fault causal relationship graph and are recorded as the corresponding associated alarm responses. Associated alarm responses with a comprehensive association strength greater than a preset association strength threshold are selected from the associated alarm responses and recorded as core associated alarm responses. The comprehensive correlation strength of each core associated alarm response is obtained. This comprehensive correlation strength is then multiplied by the fault index of the corresponding core fault alarm response to obtain the correlation fault index for each core associated alarm response. All correlation fault indices corresponding to each core fault alarm response are weighted and summed, then added to its own fault index and normalized to obtain the urgency coefficient of each core associated alarm response. Based on the urgency coefficient, the real-time alarm responses and the latent alarm responses are sorted in descending order to obtain the real-time alarm response scheduling table and the latent alarm response scheduling table. The alarm responses are dynamically scheduled based on the order of the scheduling tables.

9. The intelligent ring main unit power distribution system with integrated primary and secondary ring main units according to claim 8, characterized in that, The method for constructing a fault causal relationship graph is as follows: The multidimensional state variables of each functional unit in the ring network box are obtained, and a unified multidimensional state vector is constructed. Based on prior physical knowledge, the differential equations of each state variable are established. Kernel regression estimation is performed on the actual observed time series data of each state variable in the multidimensional state vector. Any two state variables in the multidimensional state vector are matched to obtain multiple directed matching pairs. The differential dependency of each directed matching pair is calculated. Directed matching pairs with differential dependency greater than a preset differential dependency threshold are denoted as differential causal edges. Substitute the real-time observation data of a certain state variable into the corresponding differential equation to obtain the corresponding physical residual; obtain the physical residual corresponding to each differential causal edge; if the physical residual is greater than the preset residual threshold, remove the corresponding differential causal edge to obtain the key differential causal edge. Structural causal model intervention verification is performed on each key differential causal edge to obtain valid causal edges that pass verification; the differential sensitivity, physical residual confidence, and intervention response amplitude corresponding to each valid causal edge are obtained, and weighted summation is performed to obtain the comprehensive correlation strength of each valid causal edge; Using the fault type of each functional unit as nodes, each effective causal edge as a directed edge, and the comprehensive correlation strength as the edge weight, a fault causal correlation graph is constructed.

Images