Method and system for evaluating power supply reliability of power distribution network based on data analysis

Abstract

The invention discloses a power distribution network power supply reliability evaluation method and system based on data analysis, and relates to the technical field of power distribution network detection, a meteorological station, a GIS and an equipment monitoring terminal collect temperature, rainfall, wind speed, load and fault record data, and after time sequence alignment and interpolation correction, a unified environment data set is generated to provide an accurate basis for model construction; secondly, using historical and online data, calling a nonlinear mapping function and an equipment function to train a meteorological-fault rate model, outputting the equipment fault probability in real time, and then calculating a system risk index and a line risk value in combination with confidence-weighted weather forecast and equipment load coupling operation to accurately identify a high-risk area; priority scheduling, load transfer and graded load shedding are carried out on the high-risk area, fault information is transmitted back, closed-loop control is achieved, and the operation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network detection, and specifically to an evaluation method and system for the power supply reliability of a distribution network based on data analysis. Background Technique

[0002] In modern distribution systems, with the large-scale access of distributed energy and the continuous increase in user electricity demand, the network operation environment has become increasingly complex, showing characteristics such as an increased peak-valley difference in load and a variable power flow direction. At the same time, the frequent occurrence of extreme meteorological events (such as typhoons, blizzards, high temperatures, sandstorms, etc.) has significantly increased the power outage risk and potential faults of distribution equipment under multiple external disturbances. To ensure the stable operation and power supply quality of the distribution network in diverse scenarios, it is urgent to effectively integrate and efficiently analyze environmental data (meteorology, geographical information, air pollution index, etc.) and equipment self-state data (fault records, health status, load level, etc.), and achieve proactive intervention and dynamic operation and maintenance based on refined prediction means. However, the diversity and heterogeneity of data sources, combined with factors such as the easy failure of sensors and the easy interruption of communication under extreme weather conditions, make the traditional evaluation methods relying on static experience or single data sources difficult to accurately reflect the potential risks in the actual operation of the distribution network.

[0003] In the Chinese invention patent with the application publication number CN118213983A, a reliability evaluation method for an active distribution network is disclosed, including: initializing the distribution system parameters; S2, inputting the first-layer indicators; S3, initializing the parameters; S4, calculating the equivalent load; S5, obtaining the load historical data; S6, determining whether the equivalent load is greater than 0, if less than 0, then jump to step S9; S7, determining whether the load point is normal, if not normal, then jump to step S9; S8, recording the interruption and the interruption duration; S9, determining whether the interruption duration T is greater than the maximum interruption duration Tmax, if not greater than Tmax, then T = T + 1 and jump to step S6; S10, traversing each equivalent load in the node, calculating the average interruption frequency AIF and the average interruption duration AID of each equivalent load; S11, calculating the second-layer indicators of each system.

[0004] Combined with the above actual application scenarios and existing technologies: The current technology mainly has problems such as insufficient fusion processing of multi-source data of the environment and equipment, inaccurate dynamic failure rate modeling, and difficulty in providing real-time reliability assessment and proactive operation and maintenance decisions in the face of high load or extreme weather. Specifically, on the one hand, multi-source data often exhibits problems such as inconsistent sampling frequencies, different format standards, a single interpolation method that cannot take into account spatio-temporal characteristics, etc., resulting in the quality of the fused data being difficult to meet the requirements of high-precision prediction. On the other hand, in terms of failure rate modeling, there is a lack of dynamic correlation description of the rapidly changing environmental factors and equipment health under extreme weather, resulting in lagging prediction results or large deviations from the on-site situation. In addition, in the face of potential large-scale power outage risks, traditional operation and maintenance methods generally adopt after-fact repair or static load shedding strategies, making it difficult to allocate emergency resources in a timely manner or implement hierarchical load management, thus unable to effectively reduce the overload risk of key equipment in the early stage of a disaster and improve the overall resilience and sustainable operation ability of the distribution network.

[0005] Therefore, the present invention provides an evaluation method and system for the power supply reliability of a distribution network based on data analysis. Summary of the Invention

[0006] (I) Technical problems to be solved Aiming at the deficiencies of the prior art, the present invention provides an evaluation method and system for the power supply reliability of a distribution network based on data analysis. By collecting temperature, precipitation, wind speed, load and fault record data from a meteorological station, GIS and equipment monitoring terminals, and through time series alignment and interpolation correction, a unified environmental data set is generated to provide an accurate basis for model construction. Secondly, using historical and online data, a non-linear mapping function and an equipment function are called to train a meteorology-failure rate model, and the equipment failure probability is output immediately. Thirdly, by combining confidence-weighted weather forecasts and equipment load coupling operations, the system risk index and line risk values are calculated to accurately identify high-risk areas. Finally, priority scheduling, load transfer, and hierarchical load shedding are implemented for high-risk areas, and fault information is transmitted back to achieve closed-loop control and improve operation efficiency; thus, the technical problems recorded in the background art are solved.

[0007] (II) Technical solutions To achieve the above objectives, the present invention is realized through the following technical solutions: An evaluation method for the power supply reliability of a distribution network based on data analysis, including: When a multi-source data trigger synchronization signal from a meteorological station, a geographic information system and a distribution equipment monitoring terminal is detected, a time series alignment and multiple imputation algorithm is called to process the original data (such as temperature, precipitation, wind speed, load and fault records, denoted as the environmental vector and the correction sequence ) Perform format standardization and confidence weighting processing to generate a comprehensive environmental dataset with unified spatiotemporal index, complete data and high precision, providing a solid data foundation for subsequent model construction; When the environment vector When a preset completeness threshold is reached, a nonlinear feature extraction function is used With learnable functions , extract features and dynamically train the cleaned historical samples and online incremental data, and output real-time fault probability , realize adaptive prediction and online correction of equipment health status and environmental impact; When real-time environment sequences are acquired for the next several hours to days After completing the confidence correction, combined with the equipment operation status Coupling factor with equipment load , the short-term failure risk index is calculated through nonlinear mapping and integral weighted operation Line Risk Value , locate the high-risk area collection and dynamically characterize the risks; When the high-risk area triggers the prevention mechanism, the scheduling priority coefficient An optimized scheduling algorithm is used to prioritize the allocation of emergency vehicles and spare parts resources. At the same time, load transfer and graded load shedding strategies are implemented, and temporary power supply is connected to key equipment to achieve real-time intervention and closed-loop control of fault risks. The actual fault information is transmitted back to the data platform for online correction.

[0008] Preferably, the data in discrete format is uniformly converted and timestamped. Arrange the data for the core index; remove data segments with too many null values, mark the extreme values ​​that are obviously unreasonable, and output a standardized data set; When the difference between adjacent observation times is greater than expected, a higher-order function interpolation or piecewise polynomial strategy is used to obtain a corrected sequence for the standardized data set. ; Preferably, a composite interpolation function is used to perform short-term inference on the missing interval and remove suspected invalid data; Introduce the fusion formula to obtain the comprehensive output environment vector :

[0009] in, is the total number of data sources; Indicates Data sources at time Reliability assessment factors on ; For the The sensitivity coefficient of each data source; Indicates exponential operation; is the sequence value of the th data source obtained after the previous interpolation; Preferably, the environmental vector is specifically used for the input on the model environment side, and a running feature vector is defined for each device such that the running feature vector is strictly aligned with the environmental vector in the time dimension; Select a mapping function to characterize the correlation between environmental factors and the device failure probability: where,

[0010] represents a specific time point represents the failure probability or failure rate of the device at time ; represents a non - linear feature extraction function, and is a learnable function for the device ; different devices can each adapt their running features to different learnable functions ; Preferably, historical samples are selected to initially train the parameters of the non - linear feature extraction function and the learnable function to obtain an initial set of model parameters ; for the new failure events and newly added environmental data that continuously arrive during subsequent actual operation, time stamps are attached to them and they are merged into the incremental data set, and online learning is performed based on the following correction formula:

[0011] where, represents the set of model parameters obtained after the th correction, is the parameter after the previous iteration; includes the newly added failure labels and corresponding environment - device data at this time point; is the correction mapping function; the final iteration obtains a stable and convergent set of parameters to form the final dynamic model; Preferably, the current time and the environmental sequence are obtained from a meteorological monitoring station or a satellite observation platform, where is the prediction duration; and Convert or match the model to the environmental vector Corresponding dimension; use the following device failure probability mapping method to calculate the future time The failure probability under :

[0012] Where: Is replaced by the prediction sequence Let ; Still represents the operating state of the device at Time; By the environmental factor inputs at each time Execute the above mapping to obtain the corresponding time , forming a short-term failure probability sequence of the device within the prediction interval , where Is the estimated maximum prediction duration; Preferably, introduce the following device load coupling factor To characterize the load magnitude and importance of the device :

[0013] Among them, Represents the real-time load value carried by the device , Is a positive parameter for regulating load sensitivity, ; Is the importance coefficient of the device, ); Is an exponential function; Based on the device load coupling factor And the device failure probability , construct a short-term failure risk index at the system level :

[0014] In the formula: is the time The failure probability of the th device at Is a non-linear function, Is a time weighting function; Preferably, group the lines or feeder clusters to which the device belongs at the network topology level, and summarize their corresponding failure probability distributions and device load coupling factors , for each line or feeder , define the line risk value : ​

[0015] Among them, is the contribution ratio coefficient of device j to line in its area; Set as the risk threshold, , then it is determined that line has a fault threat at time ; If the line risk values in multiple consecutive time periods during the prediction period all exceed the risk threshold , the corresponding area is determined as a high-risk area and constitutes a high-risk area set; resource allocation and emergency arrangements are made for the lines or feeders whose line risk values exceed the risk threshold; Preferably, select from the line risk values the lines or device sets that exceed the risk threshold multiple times during the prediction period ; introduce a scheduling priority coefficient for high-risk areas, and arrange them in descending order according to the magnitude of the scheduling priority coefficient to determine that the repair personnel and emergency supplies are preferentially dispatched to the area with the highest value; The construction formula of the scheduling priority coefficient

[0016] is as follows: Among them represents the set of lines or devices in the high-risk area; is the device load coupling factor of device ; is the value after non-linearly transforming the failure probability of device ; represents the total number of devices in area ; is the total value of key loads in area , is the load reference value; represents the average duration since the last maintenance of the devices in area , is the maintenance reference period , , are weight coefficients, all with values greater than 0, and are used to adjust the relative influence of each factor in the overall scheduling priority; If some devices If the probability of failure in the future period is higher than the preset threshold, the corresponding spare parts or temporary replacement parts can be pre-placed in a targeted manner; Preferably, for the located high-risk lines or equipment, if the actual load is about to enter the predicted medium- and high-risk sections, the transferable load can be redistributed based on the adaptive load transfer formula:

[0017] in, Indicates the amount of load transfer Next, equipment Load buffer factor; For equipment The load translation reference; is a positive regulatory factor; When the load buffer factor When the buffer threshold is exceeded, the device If the load transfer operation brings too much cost, a load shedding solution should be adopted. For loads that cannot be safely transferred or equipment that has reached a critical state of failure, graded load shedding measures are automatically triggered; among which: Based on the equipment load importance factor Sort by equipment and keep the most critical loads; if the failure probability is still high, gradually expand the load shedding range until the equipment The operating status of the equipment falls back to a safe range or the emergency repair team arrives to carry out offline disposal; Preferably, the actual fault information and the final load transfer record are transmitted back to the established data platform in real time; such new data will be automatically marked with the corresponding timestamp and device identification, and the record includes: Actual fault occurrence time and position, and the amount of load transfer ultimately performed And the corresponding time period, whether there are false alarms or missed alarms, and the reasons.

[0018] The evaluation method of distribution network power supply reliability based on data analysis includes: The data acquisition fusion module, when detecting the multi-source data trigger synchronization signal, calls the timing alignment and multiple interpolation algorithms to perform format standardization and confidence weighting processing on the original data to generate a comprehensive environmental data set; Failure rate modeling module: When the environmental vector reaches the preset integrity threshold, it uses nonlinear feature extraction functions and learnable functions to extract features and dynamically train the cleaned historical samples and online incremental data, outputs real-time failure probability, and adaptively predicts and corrects the health status of the equipment and environmental impact online; Short-term assessment module: After obtaining the real-time environmental sequence for the next few hours to days and completing the confidence correction, it combines the device operation status and the device load coupling factor, and calculates the short-term fault risk index and the line risk value through non-linear mapping and integral weighted operation to locate the high-risk area set; Operation and maintenance decision-making module: When the high-risk area set triggers the prevention mechanism, it preferentially processes according to the scheduling priority coefficient using the optimized scheduling algorithm, and at the same time implements the load transfer and hierarchical load shedding strategies, and transmits the actual fault information back to the data platform for online correction.

[0019] (3) Beneficial effects The present invention provides an evaluation method and system for the power supply reliability of a distribution network based on data analysis, having the following beneficial effects: By constructing a multi-source data set, unifying the spatio-temporal index and using interpolation and confidence factors for in-depth fusion, it can greatly improve the consistency and data quality of environmental factors and device operating conditions; Training and dynamically correcting the meteorological-failure rate model, using online updated parameters and the constructed non-linear mapping function , it can capture the real-time fluctuations of the device failure rate with meteorological changes, which is beneficial to providing accurate failure probabilities for real-time environmental factor prediction and short-term reliability assessment ; Combining the environmental sequence and the device load coupling factor to calculate the system risk index or the line-level risk value , it can prospectively identify high-risk areas and vulnerable links; Introducing the priority scheduling coefficient and implementing the adaptive load transfer and the hierarchical load shedding strategy to actively reduce the equipment overload risk before the fault critical point; all actual fault data and repair results are then dynamically transmitted back to the fusion platform and the online correction module mentioned above to form an end-to-end closed-loop control.

[0020] It can quickly locate potential fault devices during extreme weather, reduce the power outage scope and duration, and at the same time significantly improve the management efficiency of the entire life cycle of the distribution network, providing strong support for the distribution reliability assessment and risk prevention and control in complex scenarios. Brief description of the drawings

[0021] Figure 1 It is a schematic flow chart of the evaluation method for the power supply reliability of the distribution network of the present invention; Figure 2Schematic diagram of the evaluation system structure for the power supply reliability of the distribution network of the present invention. Detailed implementation manners

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] Please refer to Figure 1 , the present invention provides an evaluation method for the power supply reliability of a distribution network based on data analysis, including First step, when a multi-source data trigger synchronization signal from a meteorological station, a geographic information system, and a distribution equipment monitoring terminal is detected, the system calls a time series alignment and multiple imputation algorithm to perform format standardization and confidence weighting processing on the original data (such as temperature, precipitation, wind speed, load, and fault records, denoted as the environmental vector and the correction sequence ), so as to generate a comprehensive environmental data set with a unified spatio-temporal index, complete data, and high precision, providing a solid data foundation for subsequent model construction; The first step includes the following contents: Step 101, Multi-source data acquisition and preliminary standardization For multiple types of data sources including meteorological stations, geographic information systems GIS, environmental sensors, and distribution equipment monitoring systems, obtain multi-source basic data such as temperature, precipitation, wind speed, humidity, dust concentration, equipment operating conditions, and fault records; Define an independent index identifier for each data source , where is the data source number, , is the total number of data sources, and this index is referenced in subsequent steps to ensure that the data sources are not confused; perform unified conversion on data in discrete formats (such as XML, JSON, CSV) to ensure that field names, data types, and units are consistent (for example, temperature is unified to degrees Celsius, and wind speed is unified to meters per second), and perform preliminary arrangement of the data with the timestamp as the core index; Eliminate data segments with excessive null values, mark obviously unreasonable extreme values (such as fault records exceeding the physical limit of the equipment), and output the standardized data set; further, configure a set of acceptable physical range thresholds or confidence intervals, and when the monitored data (such as temperature, load, voltage value, etc.) exceeds this threshold, automatically mark it as abnormal and perform secondary verification. If it is confirmed to be invalid, it is not directly included in the imputation operation; During use, through multi-source data acquisition and preliminary standardization, unified naming and index management of data from different heterogeneous sources can be achieved, avoiding confusion during subsequent step references.

[0024] Step 102, Fine-grained data alignment, interpolation, and fusion Based on the standardized data set, further data alignment is performed at the time and space levels, and advanced interpolation processing is carried out on missing or abnormal segments to generate a comprehensive data set that can be used to construct a meteorology-failure rate dynamic model, where: According to the unified timestamp t, when the difference between adjacent observation times is greater than expected, a higher-order function interpolation or piecewise polynomial strategy is adopted instead of simple linear interpolation to obtain a smoother and physically consistent correction sequence ; If the sampling of some sensors is sparse and intermittent, after confirming that there are no serious deviations in their key data segments, a composite interpolation function can be used to perform short-term inference on the missing intervals, and suspected invalid data (such as continuous null values caused by device offline) can be excluded if necessary; After completing the interpolation and obtaining the correction sequences of each data source To take into account the differences in accuracy, reliability, and timeliness among different data sources, the following fusion formula is introduced to obtain the comprehensive output environmental vector :

[0025] where, is the total number of data sources; represents the reliability assessment factor of the th data source at time , and its value range is usually , and it can be specifically evaluated based on the equipment status integrity or sensor online rate of the source; is the sensitivity coefficient of the th data source, usually a positive value, used to control the response intensity to different ; represents the exponential operation; is the sequence value of the th data source obtained after the previous interpolation, which has been aligned with other sequences on the time axis; The above fusion process can ensure the flexible adjustment of the contribution degrees of each data source while making full use of the source reliability information, improving the stability and accuracy of the overall data set in subsequent predictions; finally, the obtained environmental vector and each correction sequence are synchronously output to form a comprehensive database; the output records need to retain , and other parameters; During use, through refined interpolation and fusion, while ensuring the basic authenticity of each data source, the differences in reliability and sensitivity can be fully considered, thereby improving the ability to correct missing data and outliers, and outputting a multi-source environmental vector with higher confidence 。

[0026] Step 2. When the environmental vector reaches the preset integrity threshold, use the non-linear feature extraction function and the learnable function to perform feature extraction and dynamic training on the cleaned historical samples and online incremental data, and output the real-time failure probability to achieve adaptive prediction and online correction of the equipment health status and environmental impact; The content of the said Step 2 includes the following: Step 201. Model structure design and data consistency verification For the environmental vector output from the first step and the operation feature sets of each device, such as the current load, electrical parameters, historical fault marks, etc. of the device , design a suitable failure probability mapping structure, and perform data consistency verification during this process. The core technical features and logic are as follows: Define the environmental vector specifically for the input on the environmental side of the model (where is the dimension of the fused environmental elements, such as key meteorological factors like temperature and wind speed); define the operation feature vector for each device , including the operation status, load level, and health index operation parameters of the device, so that the operation feature vector is strictly aligned with the environmental vector in the time dimension; Select a suitable mapping function to characterize the correlation between environmental elements and device failure probability. To improve scalability, the following form of mapping function is proposed in this step:

[0027] where, represents a specific time point (continuous or discrete), represents the failure probability or failure rate of the device at time ; represents a non-linear feature extraction function for extracting the high-order features of the environmental vector (such as non-linear combination or time series dependence), and A learnable function (which can be implemented by means of neural networks, random forests, etc.) to output a more targeted fault probability assessment while considering the state of the device itself ; Different devices can share the same non-linear feature extraction function for environmental side feature extraction and can adapt different learnable functions according to their respective operating characteristics ; Check the operating feature vector and the environmental vector to see if there are any missing values or time misalignments. If any anomalies are found, correct or recall the first step for imputation and revision; When in use, through the implemented model structure design and data consistency verification, the unity of the input and output of the meteorological-failure rate dynamic model in terms of time and object identification is ensured; at the same time, with the help of the non-linear feature extraction function and the customizable learnable function , a multi-dimensional non-linear mapping of environmental factors and device states can be realized.

[0028] Step 202, Model training and dynamic correction Select historical samples within a certain time range (including the operating feature vector and the environmental vector and the corresponding true fault labels), and use the gradient descent algorithm to initially train the parameters of the non-linear feature extraction function and the learnable function to obtain the initial model parameter set ; Add weight constraints to the pre-constructed objective loss function. If overfitting is found during the training process (such as overfitting to a small number of abnormal points), adjust it by means of regularization or updating the learning rate, etc.; For the new fault events and newly added environmental data (such as climate mutations, new records of extreme weather) that continuously arrive during subsequent actual operation, timestamp them and merge them into the incremental data set, and perform online learning based on the following correction formula:

[0029] where, represents the model parameter set obtained after the th correction, is the parameter after the previous iteration; includes the newly added fault labels and the corresponding environment-device data at this time point; is the correction mapping function (which can be implemented based on mini-batch updates or reinforcement learning strategies); finally, an iteratively obtained stable and convergent parameter set , a final dynamic model is formed, which can predict the failure probability of the device in real time; The obtained dynamic model can be used for each time point with the device to give the failure probability , and this probability value, together with the already trained non-linear feature extraction function and the learnable function parameters, are all output in the current step. The failure probability values of all devices are stored in the form of a time series ; When in use, through iterative training and dynamic correction, the meteorological-failure rate model can maintain a high prediction accuracy when facing changing environmental conditions and device states, and provide a direct input for subsequent short-term reliability assessment in the form of failure probability; the online update mechanism significantly improves the model's adaptability to sudden extreme weather, making the overall solution have the characteristics of continuous evolution.

[0030] Step 3: When obtaining the real-time environmental sequence for the next few hours to days and completing the confidence correction, combined with the device operating state and the device load coupling factor , calculate the short-term failure risk index and the line risk value through non-linear mapping and integral weighted operations, and locate and dynamically characterize the high-risk area set; The said Step 3 includes the following contents: Step 301: Obtain and interface the real-time meteorological data and short-term weather forecast Obtain the current time from a meteorological monitoring station or satellite observation platform and the environmental sequence for the next few hours to days , where is the prediction duration (which can be discretized into steps of several hours or days); convert or match the environmental sequence to the corresponding dimension of the model environmental vector ; if there are significant differences in the forecast accuracy corresponding to different time steps , then the confidence of the more distant forecast data can be weighted by combining the reliability factor or other weight mechanisms to reduce the impact of uncertainty on subsequent evaluations; Step 302: Short-term failure probability calculation and comprehensive evaluation based on the dynamic model Using the following device failure probability mapping method, calculate the failure probability at the future moment :

[0031] Wherein: is replaced by the prediction sequence , let ; still represents the operating state of the device at moment; By inputting the environmental factors at each moment to perform the above mapping, the corresponding to the corresponding moment can be obtained, forming a short-term fault probability sequence of the device within the prediction interval, where is the estimated maximum prediction duration; the environmental factor refers to the key meteorological factors included in the environmental vector ; so that the fault probability calculation process integrates real-time predicted environmental data and the real-time state of the device ; After obtaining the device fault probabilities of all devices, they can be combined with associated parameters such as device importance and load capacity, and the following device load coupling factor is introduced to characterize the load magnitude and importance of the device :

[0032] Wherein, represents the real-time load value (or maximum load ratio) borne by the device , is a positive parameter for regulating the load sensitivity, ; is the importance coefficient of the device (which can be set according to its criticality in the network structure or the nature of the power supply user, taking a positive value, and the larger the value, the more critical the device ); is an exponential function; Based on the device load coupling factor and the device fault probability , a short-term fault risk index at the system level is constructed:

[0033] In the formula: is the fault probability of the th device at time ; is a non-linear function used to appropriately amplify or suppress the fault probability ; For example: , where is a positive value used to adjust the amplification effect of the overall risk when the fault probability is relatively high; is a time-weighted function used to characterize the risk evolution trend from to . For example: , where is the attenuation factor, is the prediction duration, that is, the time within which the risk is desired to be evaluated after the current moment ; is the integration variable; by calculating the distribution of the short-term fault risk index within the prediction interval , the short-term risk evolution trend can be obtained; When in use, the verified fault probability model is combined with real-time predicted environmental data. On the basis of considering multiple factors such as equipment operation load and importance coefficient, a short-term risk assessment at the system level is formed. Through exponential amplification or non-linear mapping, the impact degrees of low-probability and high-probability fault devices on the overall network can be significantly distinguished, realizing more flexible and refined risk characterization.

[0034] Step 303, High-risk area identification and result output To clarify the high-risk area, the lines or feeder clusters to which the equipment belongs are grouped at the network topology level, and their corresponding fault probability distributions and equipment load coupling factors are summarized. For each line or feeder , the line risk value is defined as:

[0035] where is the contribution ratio coefficient of equipment j to line in its area (which can be combined with its load position, the number of branch nodes, etc.); by means of this weighting method, it is possible to quickly identify whether there are high-probability fault devices accumulated on a certain line; According to the operation and maintenance department or industry standards, can be set as the risk threshold. , then it is determined that line has an obvious potential fault threat at time ; If the line risk values of multiple consecutive time periods within the prediction period all exceed the risk threshold , then the risk in this area is continuously at a high level, and higher-priority handling should be given. The corresponding area is determined as a high-risk area and forms a high-risk area set; The obtained short-term fault probability sequence Short-term fault risk index and the line risk value are output together to form the overall evaluation result of the third step; For the line risk value lines or feeders with risk values exceeding the risk threshold are subject to resource allocation and emergency arrangements, and the real fault information is fed back to the dynamic model for online calibration ( update) and abnormal data correction; When in use, the risk quantification results at the equipment level are dimensionally elevated to the line or feeder area level to accurately locate potential vulnerable links in the actual distribution network topology, and a list of high-risk lines can be directly output for subsequent formulation of targeted operation and maintenance plans or contingency plans, which can effectively ensure the implementation and operability of model prediction in operation and maintenance practice.

[0036] Step Four: When the high-risk area set triggers the prevention mechanism, according to the dispatching priority coefficient the optimized dispatching algorithm is adopted to preferentially allocate emergency vehicles and spare parts resources, and at the same time, load transfer and hierarchical load shedding strategies are implemented, and temporary power supplies are connected to key equipment to achieve real-time intervention and closed-loop control of fault risks, and the actual fault information is transmitted back to the data platform for online calibration; ` The said Step Four includes the following contents: Step 401: High-risk area priority dispatching and emergency resource allocation Directly call the high-risk area set (including equipment or line identifiers) and the short-term fault probability sequence within the relevant interval ; Screen out the lines or equipment sets that exceed the risk threshold multiple times within the prediction period from the line risk value ; ; Introduce the dispatching priority coefficient for high-risk areas , and the construction of the dispatching priority coefficient is based on the comprehensive quantitative consideration of equipment risks, key loads and maintenance history within the area , and its formula is as follows:

[0037] In the formula: where represents the set of lines or equipment within the high-risk area; is the equipment load coupling factor of equipment ; is the value after non-linearly transforming the fault probability of equipment ; represents the total number of equipment within the area , and this item takes the mean of the risk contributions of all equipment; is the total value of critical loads within the area , is the load reference value for normalization processing; represents the average duration of the equipment within the area since the last maintenance; is the maintenance reference period , , are weight coefficients, all with values greater than 0, used to adjust the relative influence of each factor in the overall scheduling priority; According to the scheduling priority coefficient , arrange in descending order, and determine that the emergency repair personnel and emergency materials (such as spare transformers, cable assemblies) are preferentially dispatched to the area with the highest scheduling priority coefficient value; If some equipment has a very high probability of failure in the future period, corresponding spare parts or temporary replacement parts can be pre-positioned accordingly; if there are multiple high-risk areas at the same time, the resource allocation and the order of personnel deployment can be solved based on linear programming or heuristic algorithms, and try to cover the most critical or vulnerable points as much as possible under limited resources.

[0038] When in use, through the priority scheduling and emergency resource allocation logic, the operation and maintenance department can actively invest limited human and material resources into the areas most in need before extreme weather or high-failure-risk periods, significantly reducing the possible power outage scope and duration, and laying a decision-making basis for subsequent active emergency repair and load optimization.

[0039] Step 402, Load optimization and adaptive load shedding strategy For the located high-risk lines or equipment, if the actual load is about to enter the medium-high risk section in the prediction (for example, the load of equipment will break through its safety threshold in a short time), the transferable part of the load can be reallocated based on the following adaptive load transfer formula:

[0040] where represents the load buffer coefficient of equipment under the load transfer amount ; is the load translation benchmark of equipment , determined by combining its rated capacity and historical load characteristics; is a positive regulation factor, with a value greater than 0, used to balance the transfer cost and the benefit of risk reduction during load transfer, the larger it is, the more difficult it is to transfer the load, and higher incentives or more cautious operations are required;​ When the load buffer coefficient exceeds the buffer threshold, it means that the load transfer operation of the device brings too high a cost. Then, the next load shedding scheme should be adopted. For loads that cannot be safely transferred or devices that have reached the critical failure state, hierarchical load shedding measures are automatically triggered; among them: Sort according to the device load importance factor and first cut off a part of the load that has the least impact on system safety, and retain the most critical load; if the failure probability remains high, gradually expand the load shedding range until the operating state of the device drops back to the safe range or the repair team arrives on site to carry out off-line disposal; When in use, when the failure probability is relatively high but no actual failure has occurred, dynamically regulate the network load distribution by combining transfer + load shedding, reduce the time and probability of high-load operation of the device, thereby effectively reducing the failure probability caused by overheating or overcurrent. The adaptive algorithm provides a hierarchical response method, which can flexibly determine the load shedding scale and priority according to the actual network demand.

[0041] Step 403, Fault dynamic feedback and scheme iterative optimization Transmit the actual fault information (occurrence time, fault cause, repair duration, etc.) and the final load transfer record to the established data platform in real time; such new data will be automatically marked with corresponding timestamps and device identifiers, and the records include: The actual fault occurrence time and location, the final executed load transfer volume and the corresponding time period, whether false alarms, missed alarms occur and cause analysis (such as sensor anomalies); The transmitted data will trigger the second-step model online correction process (update as described in step 202 ), and perform new interpolation or denoising operations in the first-step data fusion link to improve the accuracy of the next-cycle prediction. The short-term fault risk index conducts a post-mortem review based on real fault data, and revises the value of the mapping function or the weighted kernel function; When in use, through continuous monitoring and closed-loop feedback, realize the dynamic iteration of the operation and maintenance strategy, so that this scheme can continuously adapt to new fault modes, new sensor data and grid structure changes. At the same time, the online correction will make the "weather - failure rate" model more practical, continuously approach the real device fault behavior, and improve the long-term reliability of the whole network.

[0042] Please refer to Figure 2 , the present invention provides an evaluation system for the power supply reliability of a distribution network based on data analysis, including, The data acquisition and fusion module, when detecting a multi-source data trigger synchronization signal from a meteorological station, a geographic information system, and a power distribution equipment monitoring terminal, the system calls the time series alignment and multiple imputation algorithm to perform format standardization and confidence weighting processing on the original data (such as temperature, precipitation, wind speed, load, and fault records, denoted as the environmental vector and the correction sequence ) to generate a comprehensive environmental dataset with unified spatio-temporal indexing, complete data, and high precision, providing a solid data foundation for subsequent model construction; The failure rate modeling module, when the environmental vector reaches the preset integrity threshold, uses the non-linear feature extraction function and the learnable function to perform feature extraction and dynamic training on the cleaned historical samples and online incremental data, and output the real-time failure probability to achieve adaptive prediction and online correction of the equipment health status and environmental impact; The short-term evaluation module, when obtaining the real-time environmental sequence for the next few hours to days and completing the confidence correction, combines the equipment operation status and the equipment load coupling factor to calculate the short-term failure risk index and the line risk value through non-linear mapping and integral weighting operations, and locates and dynamically characterizes the high-risk area set; The operation and maintenance decision-making module, when the high-risk area set triggers the prevention mechanism, according to the scheduling priority coefficient uses the optimized scheduling algorithm to preferentially allocate emergency vehicles and spare parts resources, and at the same time implements load transfer and hierarchical load shedding strategies, and connects temporary power supplies to key equipment to achieve real-time intervention and closed-loop control of the failure risk, and transmits the actual failure information back to the data platform for online correction.

[0043] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0044] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0045] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only for some logical function divisions. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0046] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0047] As mentioned above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for evaluating the reliability of power supply of a distribution network based on data analysis, characterized in that: include, When a multi-source data trigger synchronization signal is detected, the timing alignment and multiple interpolation algorithms are called to perform format standardization and confidence weighting processing on the original data to generate a comprehensive environmental data set; When the environmental vector reaches the preset integrity threshold, nonlinear feature extraction functions and learnable functions are used to extract features and dynamically train the cleaned historical samples and online incremental data, output real-time fault probability, and adaptively predict and correct the equipment health status and environmental impact online. After obtaining the real-time environmental sequence for the next few hours to days and completing the confidence correction, the short-term fault risk index and line risk value are calculated by combining the equipment operation status and equipment load coupling factor through nonlinear mapping and integral weighted operation to locate the high-risk area set; When a collection of high-risk areas triggers the prevention mechanism, an optimized scheduling algorithm is used to prioritize processing based on the scheduling priority coefficient, while load transfer and graded load shedding strategies are implemented, and actual fault information is transmitted back to the data platform for online correction.

2. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 1, characterized in that: Perform uniform conversion on discrete format data, remove data segments with too many null values, mark obviously unreasonable extreme values, and output a standardized data set; When the difference between adjacent observation times is greater than expected, a higher-order function interpolation or piecewise polynomial strategy is used to obtain a corrected sequence for the standardized data set.

3. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 2, characterized in that: The composite interpolation function is used to perform short-term inference on the missing interval, the suspected invalid data is eliminated, and the fusion formula is introduced to obtain the comprehensive output environment vector; The environmental vector is used as the input of the model environment side, and an operating feature vector is defined for each device so that the operating feature vector is aligned with the environmental vector; and a mapping function is selected to characterize the correlation between environmental factors and the probability of equipment failure.

4. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 3, characterized in that: Select historical samples to perform initial training on the parameters of the nonlinear feature extraction function and the learnable function to obtain an initial model parameter set; For new fault events and new environmental data that continue to arrive in subsequent actual operations, they are time-stamped and merged into the incremental data set. Online learning is performed based on the following correction formula, and finally a stable and convergent parameter set is obtained through iteration to form the final dynamic model.

5. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 4, characterized in that: Obtain the current time and environmental sequence from a meteorological monitoring station or satellite observation platform; convert or match the environmental sequence to the model-to-environment vector corresponding dimension; use the equipment failure probability mapping method to calculate the expected failure probability; By inputting the environmental factors at each moment and executing the above mapping, the failure probability at the corresponding moment can be obtained, forming a short-term failure probability sequence of the equipment within the prediction interval; the equipment load coupling factor is introduced to characterize the load magnitude and importance of the equipment, and based on the equipment load coupling factor and the equipment failure probability, a system-level short-term failure risk index is constructed.

6. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 5, characterized in that: At the network topology level, the line or feeder clusters to which the equipment belongs are grouped, and their corresponding fault probability distribution and equipment load coupling factors are summarized. The line risk value is defined for each line or feeder: if the line risk value is not less than the risk threshold, it is determined that the corresponding line has a fault threat at the predetermined time; If the line risk values ​​for multiple consecutive time periods during the forecast period exceed the risk threshold, the corresponding area will be judged as a high-risk area and constitute a high-risk area set.

7. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 6, characterized in that: Filter out the lines or equipment sets that exceed the risk threshold multiple times within the forecast period from the line risk values; Introduce a dispatch priority coefficient for high-risk areas, sort them in descending order according to their size, and prioritize repair personnel and emergency materials to the areas with the highest dispatch priority coefficients. If the probability of failure of certain equipment in the future period is higher than the preset threshold, pre-place corresponding spare parts or temporary replacement parts.

8. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 7, characterized in that: For the located high-risk lines or equipment, if the actual load is about to enter the predicted medium- and high-risk sections, the transferable load can be redistributed based on the adaptive load transfer formula; When the constructed load buffer coefficient exceeds the buffer threshold, the load shedding scheme is adopted. For loads that cannot be safely transferred or equipment that has reached a critical fault state, graded load shedding measures are automatically triggered; among which: Sort by equipment load importance factor and retain the most critical load; If the failure probability remains high, the load shedding range will be expanded step by step until the operating status of the equipment returns to a safe range or the emergency repair team arrives to implement offline disposal.

9. The method for evaluating the reliability of power supply of a distribution network based on data analysis according to claim 8, characterized in that: The actual fault information and final load transfer records are transmitted back to the established data platform in real time. Such new data will be automatically marked with the corresponding timestamp and equipment identification. The records include: the actual time and location of the fault, the final load transfer amount and the corresponding time period, whether there are false alarms, missed alarms, and cause analysis.

10. A distribution network power supply reliability evaluation system based on data analysis, characterized in that: include, The data acquisition fusion module, when detecting the multi-source data trigger synchronization signal, calls the timing alignment and multiple interpolation algorithms to perform format standardization and confidence weighting processing on the original data to generate a comprehensive environmental data set; Failure rate modeling module: When the environmental vector reaches the preset integrity threshold, it uses nonlinear feature extraction functions and learnable functions to extract features and dynamically train the cleaned historical samples and online incremental data, outputs real-time failure probability, and adaptively predicts and corrects the health status of the equipment and environmental impact online; The short-term assessment module, after obtaining the real-time environment sequence for the next few hours to days and completing the confidence correction, combines the equipment operation status and equipment load coupling factor, calculates the short-term fault risk index and line risk value through nonlinear mapping and integral weighted operation, and locates the high-risk area set; The operation and maintenance decision module uses an optimized scheduling algorithm to prioritize the high-risk areas when the prevention mechanism is triggered. It also implements load transfer and graded load shedding strategies, and transmits actual fault information back to the data platform for online correction.

Images