Fault positioning method and system of smart energy big data monitoring system

Abstract

The invention relates to the technical field of fault detection, in particular to a fault positioning method and system for a smart energy big data monitoring system, and the method comprises the following steps: monitoring power input and output, analyzing a loss rate, screening an abnormal region, extracting boundary equipment, calculating a tangential change rate, generating an abnormal set, comparing a threshold value, and eliminating invalid equipment. The method comprises the steps of monitoring power input and output, analyzing a power loss rate to recognize abnormal equipment, calculating a tangential change rate to generate an abnormal set, classifying by using a gradient lifting decision tree, detecting isolated equipment, calculating flow vector matching and merging, dividing a grid to generate a deviation sequence, and analyzing consistency to position fault equipment. The method comprises the following steps: monitoring power input and output, analyzing a power loss rate to recognize the abnormal equipment, calculating a tangential change rate to generate an abnormal set, and classifying by using a gradient lifting decision tree; isolated equipment is detected, flow direction vectors are matched, regional equipment is merged, energy deviation is captured in combination with multi-layer grids and wavelet transform, faults are positioned in real time, and the stability and safety of the system are improved.

Description

Technical Field

[0001] The present invention relates to the field of fault detection technology, and in particular to a fault locating method and system for a smart energy big data monitoring system. Background Art

[0002] The field of fault detection technology encompasses the identification, diagnosis, location, and early warning of faults in various systems, equipment, or processes. Core areas of expertise include fault diagnosis methods, fault detection algorithms, data acquisition and processing techniques, and the design and optimization of fault identification systems. Its goal is to promptly detect and accurately determine the location and cause of faults, avoid system losses caused by these failures, and ensure the safe, stable, and efficient operation of equipment and systems. Fault detection technology is widely used in industries such as industrial control, energy management, automotive electronics, aerospace, and manufacturing. It encompasses a wide range of technologies, from signal processing to big data analysis, supporting real-time monitoring, fault prediction, and maintenance planning.

[0003] Among them, the fault location method and system of the smart energy big data monitoring system refers to the real-time monitoring, diagnosis, and location of faults in the energy system through a monitoring system based on big data analysis. The technical matters involved include energy system fault detection, real-time collection of fault information, and data stream processing and analysis. Specifically, the system collects energy equipment operating data, processes and analyzes the collected data through big data analysis methods, locates faults in equipment operation in real time, and accurately locates them based on abnormal fluctuations or trends in the data. It mainly uses data fusion and fault pattern recognition methods to achieve rapid identification of the cause of the fault, and combines it with the diagnostic module of the monitoring system for automated judgment and positioning.

[0004] Existing technologies often use relatively basic fault diagnosis methods when dealing with fault detection and location, and for complex energy equipment systems, they often rely on traditional data acquisition and signal processing methods. When faced with the collection and processing of large amounts of data, existing technologies are prone to information overload and the accumulation of redundant data, resulting in inaccurate fault location. In addition, traditional methods lag behind in outlier removal and equipment matching, and are unable to quickly capture operational changes in equipment, resulting in extended positioning response time and the failure to achieve real-time fault diagnosis. In terms of fault pattern recognition, existing technologies still have certain blind spots in the case of large-scale systems. Especially in highly complex equipment environments, the occurrence of faults is often not promptly warned or accurately located, further affecting the stable operation of the system. These shortcomings make it difficult for existing technologies to meet the high real-time and high-precision requirements of actual operations when faced with sudden faults, greatly increasing the maintenance costs after the fault occurs and its impact on the normal operation of the system. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art, an embodiment of the present invention provides a fault location method for a smart energy big data monitoring system, comprising the following steps: In order to achieve the above objectives, the present invention adopts the following technical solution: a fault location method for a smart energy big data monitoring system, comprising the following steps: S1: Monitor the power input and output values ​​of power equipment and analyze the power loss rate. The areas exceeding the power judgment threshold are considered as compression areas. The device input gradient boosting decision tree at the area boundary is extracted to calculate the tangential change rate of the device power and generate a boundary compression set. S2: comparing the tangential change rate of the device power in the boundary compression with the upper and lower thresholds, removing devices that do not reach the lower threshold, inserting devices that exceed the upper threshold into the compression area, and detecting isolated devices based on the path connectivity from the device to the area center to generate an isolated device set; S3: Based on the isolated device set, calculate the power flow vector of the isolated device, perform similarity matching on the power flow vectors of the remaining devices, merge the matched isolated devices, and generate a new compressed region set; S4: Dividing multiple layers of grids based on the new compressed region set, comparing the multi-grid power input and output deviations at multiple resolutions, and generating an energy deviation sequence; S5: Call the energy deviation sequence to analyze the directional consistency of the deviation points at multiple resolutions. If the deviation exceeds the average deviation, mark the deviation concentration point, and calculate the cumulative value of the path power difference to locate the deep fault device and generate a fault location result.

[0006] As a further solution of the present invention, the boundary compression set specifically includes compression areas, boundary devices, and power tangential change rates; the isolated device set includes devices that have not reached the lower threshold, devices that have exceeded the upper threshold, and isolated device locations; the new compression area set includes isolated devices, similarity matching devices, and power flow vectors; the energy deviation sequence includes multi-layer grid power input deviations, multi-layer grid power output deviations, and energy deviation values; the fault location results specifically refer to deviation concentration points, power difference cumulative values, and deep fault devices.

[0007] As a further solution of the present invention, the specific steps of S1 are: S101: Monitor the power input and output values ​​of power equipment, match them, and compare them one by one. Calculate the power loss values ​​of multiple devices within a fixed time period based on the difference between the input and output powers, and summarize them to generate a power loss rate data set. S102: Calling the power loss rate dataset, identifying the device numbers and corresponding operating intervals where the loss rate exceeds the power judgment threshold, and based on the distribution of devices whose loss rates continuously exceed the threshold within the marked interval, demarcating the compression area boundary according to the topological structure to generate the device compression boundary range; S103: Based on the device compression boundary range, call the power change records of the boundary devices to construct an input feature set, split the input feature set and train the node judgment value through the gradient boosting decision tree, calculate the power tangential change rate of the multi-node devices and integrate them to generate a boundary compression set.

[0008] As a further solution of the present invention, the power judgment threshold is calculated by long-term monitoring of the input and output power of the device, and the power loss rate upper limit of the device is selected based on the device design requirements or industry standards.

[0009] As a further solution of the present invention, the specific steps of S2 are: S201: Calling the boundary compression set, removing devices whose tangential change rates do not reach the lower limits of the upper and lower thresholds, then marking devices whose tangential change rates exceed the upper limits as devices entering the compression area, and recording the corresponding change rate data to generate a compression area device data set; S202: Based on the compressed regional device dataset, calculate the path connectivity between each device and the regional center, determine whether the device is an isolated device, filter out isolated devices that cannot be connected through other devices or regions, and generate an isolated device connectivity calculation result; S203: combining the isolated device connectivity calculation result, recording the device number and region information of each isolated device, and summarizing all the information to generate an isolated device set.

[0010] As a further solution of the present invention, the upper and lower thresholds are set by the performance standards of the equipment or the maximum fluctuation range in the operating data. The upper threshold is set based on the performance standards of the equipment or the maximum fluctuation range in the operating data, and the lower threshold is set by the optimal standard value of the normal fluctuation of the equipment.

[0011] As a further solution of the present invention, the specific steps of S3 are: S301: Constructing a power conduction direction map based on the isolated device set, obtaining power flow direction data for each device under multiple connection channels, analyzing power flow vectors by extracting instantaneous power transmission differences between connection points and physical direction vectors of corresponding channels, and obtaining a device power flow vector set; S302: Calling the device power flow vector set, calculating the similarity score between the power flow vector corresponding to each isolated device and the power flow vectors of other non-isolated devices one by one, selecting device pairs that meet a basic predetermined similarity matching threshold, and obtaining a similarity vector matching pair table; S303: merging isolated devices with matching relationships according to the similarity vector matching pair table, marking the merging identification numbers to which the matching devices belong, and aggregating isolated devices with the same identification number into the same region block to generate a new compressed region set; The basic limited similarity matching threshold is set by the similarity test result between the device power flow vectors.

[0012] As a further solution of the present invention, the specific steps of S4 are: S401: Based on each region in the new compressed region set, apply grid division schemes of multiple resolutions to each region, extract power flow data of each grid, and obtain multi-layer grid power input and output results; S402: Calling the multi-layer grid power input and output results, calculating the corresponding power input and output deviation at each grid level, analyzing the energy deviation at multiple resolutions by comparing the power flow differences of multiple levels of grids, and obtaining the multi-resolution grid deviation; S403: Comparing deviation data at multiple resolution levels based on the multi-resolution grid deviation, calibrating the energy deviation range of each level of grid, and generating an energy deviation sequence.

[0013] As a further solution of the present invention, the specific steps of S5 are: S501: calling the energy deviation sequence, analyzing the direction vectors corresponding to the deviation points in the sequence at multiple resolutions, counting the direction projection differences of each deviation point at multiple resolutions, and calculating the average fluctuation degree of the direction angle as a criterion for judging direction consistency, thereby generating a direction consistency interval; S502: Based on the directional consistency interval, filter the deviation points whose directional offset values ​​are greater than the interval mean and extract the position index. Calculate the power offset increment and directional drift amplitude at the grid resolution for each deviation point, and compare the degree of deviation with the remaining deviation points to obtain the position of the deviation concentration point. S503: Based on the position of the deviation concentration point, the power transmission value differences on the adjacent node paths are calculated and accumulated, and the corresponding devices on each path are compared and located according to the trend of the accumulated value changes. The devices with abnormal power transmission mutation are extracted as deep fault nodes to generate fault location results.

[0014] The fault location system of the smart energy big data monitoring system includes: The power monitoring module monitors the power input and output values ​​of power equipment and analyzes the power loss rate. It selects areas exceeding the power judgment threshold as compression areas, extracts devices at the area boundary, calculates the power tangential change rate through the gradient boosting decision tree, generates a boundary anomaly set, and transmits it to the boundary detection module. The boundary detection module compares the power tangential change rate in the boundary anomaly with the upper and lower thresholds, removes devices that do not reach the lower threshold, inserts devices that exceed the upper threshold into the compression area, and detects isolated devices based on the path connectivity between the device and the area center. The isolated device set is generated and passed to the isolated device matching module; An isolated device matching module calculates the power flow vector of the isolated device based on the isolated device set, performs similarity matching on the power flow vectors of the remaining devices in the region, merges the matched isolated devices, generates a new compressed region set, and transmits it to the multi-grid analysis module; a multi-grid analysis module, which divides the multi-grid into multiple layers based on the new compressed region set, compares the multi-grid power input and output differences at multiple resolutions, generates an energy deviation sequence through wavelet transform, and transmits the energy deviation sequence to the fault location module; The fault location module calls the energy deviation sequence to analyze the directional consistency of the deviation points at multiple resolutions. If the deviation exceeds the average deviation, the abnormal concentration point is marked, and the cumulative value of the path power difference is calculated to locate the deep fault device and generate the fault location result.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by monitoring the power input and output values ​​and analyzing the power loss rate, abnormal areas are effectively identified and devices that do not meet the threshold are further eliminated. By calculating the power tangential change rate, not only is the generation of boundary anomaly sets achieved, but also the gradient boosting decision tree method is used to finely divide the devices, so that problem devices can be quickly identified and classified. This strategy accurately merges devices in multiple areas by detecting isolated devices and matching the similarity of power flow vectors, ensuring that the correlation between devices is optimized. With the application of multi-layer grid division and wavelet transform, energy deviation points can be effectively captured and accurately located at multiple resolutions. This improvement in accuracy not only makes fault location more real-time, but also more effectively discovers potential fault points in the context of processing big data and complex equipment systems. Through these technical innovations, the overall solution can accurately locate faulty devices in real time, ensure the stability and efficiency of the system, avoid losses caused by faults, and maximize the safety of the energy system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1Schematic diagram of the steps of the present invention; Figure 2 This is a schematic diagram of the refinement of S1 of the present invention; Figure 3 This is a schematic diagram of the refinement of S2 of the present invention; Figure 4 This is a schematic diagram of the refinement of S3 of the present invention; Figure 5 This is a schematic diagram of the refinement of S4 of the present invention; Figure 6 This is a schematic diagram of the refinement of S5 of the present invention; Figure 7 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0019] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0020] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0021] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0022] See also Figure 1 , an embodiment of the present invention provides a fault location method for a smart energy big data monitoring system, comprising the following steps: S1: Monitor the power input and output values ​​of power equipment and analyze the power loss rate. The areas exceeding the power judgment threshold are considered as compression areas. The device input gradient boosting decision tree at the area boundary is extracted to calculate the tangential change rate of the device power and generate a boundary compression set. S2: Compare the tangential change rate of device power during boundary compression with upper and lower thresholds, remove devices that do not reach the lower threshold, insert devices that exceed the upper threshold into the compression area, and detect isolated devices based on the path connectivity from the device to the area center to generate an isolated device set. S3: Based on the isolated device set, the power flow vector of the isolated device is calculated, the power flow vectors of the remaining devices are matched for similarity, and the matched isolated devices are merged to generate a new compressed region set; S4: Divide the multi-layer grid based on the new compressed region set, compare the multi-grid power input and output deviations at multiple resolutions, and generate an energy deviation sequence; S5: Call the energy deviation sequence to analyze the directional consistency of the deviation points at multiple resolutions. If the deviation exceeds the average deviation, mark the deviation concentration point, and calculate the cumulative value of the path power difference to locate the deep fault device and generate the fault location result.

[0023] The boundary compression set specifically includes the compression area, boundary devices, and power tangential change rate; the isolated device set includes devices that have not reached the lower threshold, devices that have exceeded the upper threshold, and the location of isolated devices; the new compression area set includes isolated devices, similarity matching devices, and power flow vectors; the energy deviation sequence includes multi-layer grid power input deviation, multi-layer grid power output deviation, and energy deviation value; the fault location result specifically refers to the deviation concentration point, the accumulated power difference value, and the deep fault device.

[0024] See also Figure 2 , the specific steps of S1 are: S101: Monitor the power input and output values ​​of power equipment, match them, and compare them one by one. Calculate the power loss values ​​of multiple devices within a fixed time period based on the difference between the input and output powers, and summarize them to generate a power loss rate data set. The power input and output values ​​of the monitoring power equipment are matched and compared one by one. The power data of 10 transformers (numbered A001-A010) in the substation are collected from 00:00 to 24:00 on August 4, 2025. The sampling interval is 1 hour, and there are 24 time points in total. The input power is performed at each time point. and output power Subtraction operation to calculate the power loss value , taking device A001 as an example, at 00:00 , ,but , 01:00 , ,but , the subsequent 22 time points change according to a linear law: , (t is the hour number, 0≤t≤23), after point-by-point calculation, sum up for a fixed time (24 hours): total energy loss ,in , total energy input , 24 Sum of values ,but , 24 Sum of values ,but , power loss rate , and calculate other devices in the same way, such as A002 , ,but =5%, summarize the number, date, The values ​​form a data set.

[0025] S102: Calling the power loss rate data set, identifying the device number and corresponding operating range whose loss rate exceeds the power judgment threshold, and based on the distribution of devices whose loss rate continuously exceeds the threshold in the marked interval, demarcating the compression area boundary according to the topological structure, and generating the device compression boundary range; Call the power loss rate data set (including hourly loss rate), set the power judgment threshold to 6% (according to the upper limit of the transformer rated load loss rate specified in GB / T6451-2015), and identify devices with loss rates exceeding the threshold. The loss rate sequence of device A003 from 10:00 to 15:00 is 6.1%, 6.2%, 6.3%, 6.4%, 6.2%, and 6.1%. Comparing it point by point with the threshold of 6%, the results show that 6.1>6, 6.2>6, indicating that the threshold has been exceeded for 6 consecutive hours. The sequence of device A006 from 14:00 to 17:00 is 7.1%, 7.2%, 7.0%, 7.3%, are all greater than 6%, and it is determined that the threshold value has been exceeded for 4 consecutive hours. The sequence of device A009 from 13:00 to 16:00 is 6.6%, 6.7%, 6.5%, and 6.4%. Compared with the threshold value: 6.6>6, 6.7>6, 6.5>6, 6.4>6, it is determined that the threshold value has been exceeded for 4 consecutive hours. Based on the topological coordinates: A003 (10, 20), A006 (15, 25), A009 (12, 30), take the minimum x-coordinate value of 10 and the maximum value of 15, the minimum y-coordinate value of 20 and the maximum value of 30, and define the compression boundary range as , .

[0026] S103: Based on the device compression boundary range, the power change records of the boundary devices are called to construct an input feature set. The input feature set is split and node judgment values ​​are trained using a gradient boosting decision tree. The power tangential change rates of multiple node devices are calculated and integrated to generate a boundary compression set. Based on device compression boundary range , , determine the devices A003, A004, A005, A006, and A009 within the boundary, and call their power records in the exceeding threshold interval 14:00-17:00 (time interval ), N: the total number of device nodes, calculated by compressing the boundary range, based on the defined boundary range , , traverse the topological coordinates and count the number of devices within the range, devices A003 (10, 20), A004 (11, 22), A005 (13, 25), A006 (15, 25), A009 (12, 30) → N = 5, : The judgment value of node n, quantifying the degree of device power fluctuation, collecting the power change value sequence of device n at multiple time points during the analysis period (such as hourly data from 14:00 to 17:00), Sequence: 10kW (14:00 → 15:00), -5kW (15:00 → 16:00), 8kW (16:00 → 17:00), average =4.33kW, variance =43.96, the largest variance among all devices (A006) → , : represents the power change value of node n (unit: kW), calls the device power record, and calculates the difference between adjacent time points: , Calculation example (A003@15:00): , → , : Time interval (unit: hour), determined by the data sampling frequency, data is collected once every hour → , : Boundary device power extreme value (unit: kW). Traverse the power values ​​of all devices within the boundary during the analysis period (14:00-17:00) and take the global maximum / minimum value. A006@17:00: 1067kW→ ,A004@14:00:990kW→ , extreme difference: , (15:00 time point) Substitute the parameter value: Calculate the power tangential change rate , The results show that the power tangential change rate H at 15:00 is 0.4268, which is directly used as the element of the boundary compression set (corresponding to the 15:00 time point). Similarly, the H values ​​at 14:00, 16:00, and 17:00 are calculated (such as 14:00: H=0.35, 16:00: H=0.48) to generate the compression set. .

[0027] Table 1: 15:00 time formula parameter table

[0028] See also Figure 3 , the specific steps of S2 are: S201: Calling the boundary compression set, removing devices whose tangential change rate does not reach the lower limit of the upper and lower thresholds, then marking devices whose tangential change rate exceeds the upper threshold as devices entering the compression area, and recording the corresponding change rate data to generate a compression area device data set; Call bounds compression set (generated , corresponding to the power tangential change rate from 14:00 to 17:00), set the lower threshold (Based on: analyzing 100 sets of normal data and taking the 5% quantile), upper threshold (95% quantile), traverse the tangential change rate of multiple time points in the compression set, and compare the value with For example, if 0.35 is less than 0.4 at 14:00, remove all device data at that time and compare the value with , 0.4268 at 15:00 is less than 0.45, so the device is not marked. 0.48 at 16:00 is greater than 0.45, so the device group (A003, A004, A005, A006, and A009) at this moment is marked as devices entering the compression area. The device number, time point 16:00, and tangential change rate of 0.48 are recorded and stored in the dataset. For example, the record of A003 is {"device":"A003","time":"16:00","change rate":0.48}, and a compression area device dataset (containing 5 records) is generated.

[0029] S202: Based on the compressed regional device dataset, calculate the path connectivity between each device and the regional center, determine whether the device is an isolated device, filter out isolated devices that cannot be connected through other devices or regions, and generate isolated device connectivity calculation results; Based on the compressed regional device data set (e.g., 5 devices at 16:00), the coordinates of the regional center are calculated: devices A003 (10, 20), A004 (11, 22), A005 (13, 25), A006 (15, 25), and A009 (12, 30). The average x-coordinate value is (10+11+13+15+12) / 5=12.2, and the average y-coordinate value is (20+22+25+25+30) / 5=24.4. The regional center is , : The connectivity characteristic value of device i (dimensionless), quantifying the degree of device isolation, : Euclidean distance from device i to the center of the region (unit: meter), calculate the coordinates of the region center : Take the average x-coordinate and y-coordinate of all devices and calculate the Euclidean distance : , Equipment A005): Regional Center , device coordinates , , : The average distance from all devices to the center of the area (unit: meter), calculate all The arithmetic mean of , example: , , , , , , : The maximum distance between all devices and the area center (unit: meter), compare all The value takes the maximum value, , : The number of devices adjacent to device i (dimensionless), set the proximity radius R = 5m (based on: the average distance between devices is 4-6 meters), calculate the Euclidean distance between device i and all other devices, and count the number of devices with a distance ≤ R (excluding itself). Example (device A004): Distance to A003 , distance from A005 , distance from other devices>5→ , T: Total number of devices in the compression area (dimensionless), count the number of devices in the compression area device data set, example: T=5 (devices A003, A004, A005, A006, A009), substitute the parameter values: , step-by-step calculation: Numerator: 1.0-3.414=-2.414, relative deviation: , absolute value: , neighboring proportion: , , isolated threshold, randomly select 10 groups of normal device data to calculate : , maximum value 0.45<0.6→threshold Reasonable, the results show that the connectivity characteristic value of device A005 , determined to be an isolated device.

[0030] Table 2: Device connectivity parameters (16:00)

[0031] S203: Based on the isolated device connectivity calculation results, record the device number and region information of each isolated device, and summarize all the information to generate an isolated device set; Combined with the isolated device connectivity calculation results (such as A005 ), set the isolation threshold (Based on: Analysis of 50 sets of fault data, 98% isolated equipment ), traverse multiple devices value, The connectivity eigenvalue of device i is compared with 0.6. A003's 0.4686 is less than 0.6, indicating it is not isolated. A004's 0.5304 is less than 0.6, indicating it is not isolated. A005's 0.8308 is greater than 0.6, indicating it is an isolated device. A006's 0.2981 is less than 0.6, indicating it is not isolated. A009's 0.3908 is less than 0.6, indicating it is not isolated. The isolated device number A005 is recorded, and its region is the compressed region ID (Region01). The isolated device set is generated with the structure: {"Device":"A005","Region":"Region01"}.

[0032] See also Figure 4 , the specific steps of S3 are: S301: Construct a power conduction direction map based on the isolated device set, obtain power flow direction data for each device under multiple connection channels, analyze the power flow vector by combining the instantaneous power transmission difference between the extracted connection points with the physical direction vector of the corresponding channel, and obtain a device power flow vector set; Based on the isolated device set (such as the generated {"device":"A005","region":"Region01"}), call the power system topology map to obtain the device connection channel data. Taking device A005 as an example, there are 3 connection channels: channel 1 (connected to device B101), channel 2 (connected to device B102), and channel 3 (connected to device B103). The instantaneous power transmission value at 15:00 is collected, and the input power of channel 1 is , output power , calculate the transmission difference , channel direction vector (unit vector), analyzing the power flow vector: , similarly calculate channel 2: , , , channel 3: , , , superimpose multi-channel vectors: , generating device A005 power flow vector , and process other isolated devices in the same way to form a device power flow vector set (such as A005: , A008: ).

[0033] S302: Calling the device power flow vector set, calculating the similarity score between the power flow vector corresponding to each isolated device and the power flow vectors of other non-isolated devices one by one, selecting device pairs that meet the basic predetermined similarity matching threshold, and obtaining a similarity vector matching pair table; Call device power flow vector set (including 2 isolated devices: A005 , A008 , 3 non-isolated devices: B101 , B102 , B103 ), calculate the cosine similarity: , represents the similarity score of the power flow vectors between device i and device j, Represents the power flow vector of device i (unit: kW), obtained by vector superposition, represents the power flow vector of device j (unit: kW), : vector modulus, calculate the square root of the sum of squares of multiple components, and set the basic limit similarity matching threshold =0.85 (Based on: Analysis of 100 sets of normal data, equipment pairs with similarity > 0.85 belong to the same functional area), calculated pair by pair: A005 and B101: , A005 and B102: (mismatch), A005 and B103: , A008 and B101: ,Filter matching pairs: A005-B101, A005-B103, and generate similarity vector matching pair table (structure: {"device pair":"A005-B101","similarity":1.0}, {"device pair":"A005-B103","similarity":1.0}).

[0034] S303: Based on the similarity vector matching pair table, isolated devices with matching relationships are merged, the matching devices are marked with their merge identification numbers, and isolated devices with the same identification number are aggregated into the same region block to generate a new compressed region set; The basic limited similarity matching threshold is set by the similarity test results between the power flow vectors of the devices; According to the similarity vector matching pair table (including A005-B101, A005-B103), the merged devices: A005 has a matching relationship with B101 and B103, and are assigned the merged identification number Group01. The isolated device A008 has no matching pair and is assigned Group02. The devices with the same identification number are aggregated: Group01 includes {A005, B101, B103}, and Group02 includes {A008}. A new compressed region set is generated (structure: {"region":"Group01","device":["A005","B101","B103"]}, {"region":"Group02","device":["A008"]}).

[0035] See also Figure 5 , the specific steps of S4 are: S401: Based on each region in the newly compressed region set, apply grid division schemes of multiple resolutions to each region, extract power flow data of each grid, and obtain multi-layer grid power input and output results; Based on the new compressed region set, the generated Group01 region is as follows: devices A005 (13, 25), B101 (11, 22), and B103 (14, 28). Three grid resolutions are set: level 1 (100m×100m, 1 grid for the entire region), level 2 (50m×50m, divided into 4 grids), and level 3 (25m×25m, divided into 16 grids). The device power data at 15:00 is collected. Level 1 grid 1: input power = device A005 input 1050kW + B101 input 1040kW + B103 input 1025kW = 3115kW, output power = 1040+1030+1020 = 3090kW. Level 2 grid division: grid 1 (x:10-12.5, y:20-25) contains B101, output Input is 1040kW and output is 1030kW. Grid 2 (x:12.5-15, y:20-25) contains A005. Input is 1050kW and output is 1040kW. Grid 3 (x:10-12.5, y:25-30) has no equipment. Grid 4 (x:12.5-15, y:25-30) contains B103. Input is 1025kW and output is 1020kW. Level 3 selects a grid with equipment (such as grid (3, 3): B101). Input is 1040kW and output is 1030kW, forming a multi-layer grid power input and output result (level 1: [3115, 3090], level 2: [1040, 1030], [1050, 1040], [0, 0], [1025, 1020], level 3: 16 groups of data including [1040, 1030]).

[0036] Table 3: Level 2 grid power data

[0037] As shown in Table 3, Table 3 lists the power data at a resolution of 50m×50m.

[0038] S402: Calling the multi-layer grid power input and output results, calculating the corresponding power input and output deviation at each grid level, analyzing the energy deviation at multiple resolutions by comparing the power flow differences of multiple levels of grids, and obtaining the multi-resolution grid deviation; Call the multi-grid power input and output results and calculate the multi-grid deviation: Level 1 Grid 1: , level 2 grid 1: , level 2 grid 2: , level 2 grid 4: , calculate the level energy deviation (mean): , , , generate a multiresolution grid deviation: .

[0039] S403: Based on the multi-resolution grid deviation, the deviation data at multiple resolution levels are compared to calibrate the energy deviation range of each level grid and generate an energy deviation sequence; According to the multi-resolution grid deviation value ( ), generate energy deviation sequence: Level 1 deviation range: (Single value), Level 2 deviation range: Minimum grid deviation ,maximum → , Level 3 deviation range: (empty grid), → , forming an energy deviation sequence .

[0040] See also Figure 6 , the specific steps of S5 are: S501: Calling the energy deviation sequence, analyzing the direction vectors corresponding to the deviation points in the sequence at multiple resolutions, counting the direction projection differences of each deviation point at multiple resolutions, and calculating the average fluctuation degree of the direction angle as a criterion for judging direction consistency, thereby generating a direction consistency interval; Call the energy deviation sequence (such as the generated level 1 deviation range [0.803%, 0.803%], level 2 [0%, 0.962%], level 3 [0%, 0.962%]), extract level 2 grid 2 (coordinate range x: 12.5-15m, y: 20-25m) as the deviation point, obtain the power direction vector of the grid at three resolutions: level 1 (entire area) vector (5, 12), level 2 (this grid) vector (5, 12), level 3 (subgrid) vector (5, 12), calculate the direction projection difference: level 1 to level 2 projection The shadow difference = |(5×5+12×12) / 13-13| = |(25+144) / 13-13| = |13-13| = 0, the projection difference from level 2 to level 3 = 0, calculate the direction angle fluctuation: the angle between level 1 and level 2 is cosθ = (5×5+12×12) / (13×13) = 169 / 169 = 1 → θ = 0°, the angle between level 2 and level 3 is 0°, the average fluctuation = (0+0) / 2 = 0°, and the same method is used to calculate other deviation points. For example, the average fluctuation of grid 4 is 8.2°, generating a direction consistency interval [0, 8.2°] (the upper limit is the maximum fluctuation value).

[0041] Table 4: Direction vector and angle table

[0042] As shown in Table 4, the table lists the direction vector and angle fluctuation data of the deviation point.

[0043] S502: Based on the directional consistency interval, filter out deviation points whose directional offset values ​​are greater than the interval mean and extract their position indexes. Calculate the power offset increment and directional drift amplitude at the grid resolution for each deviation point, and compare the degree of deviation with the remaining deviation points to obtain the position of the deviation concentration point. Based on the directional consistency interval [0, 8.2°] and the interval mean of 4.1°, we screened for deviation points with an offset value greater than 4.1°: grid 4 (fluctuation 8.2°). We extracted the location index (x:12.5-15m, y:25-30m), and calculated the power offset increment: the deviation value change from level 1 to level 3 was |0.488%-0.803%| = 0.315%, and the directional drift amplitude was 8.2°. In contrast, for example, grid 2 had an increment of 0 and an amplitude of 0. The comprehensive deviation value was increment × amplitude = 0.315 × 8.2 = 2.583, and the grid 2 value was 0. We then obtained the location of the deviation concentration point: grid 4 coordinate range (x:12.5-15m, y:25-30m).

[0044] S503: Based on the location of the deviation concentration point, the power transmission value differences on adjacent node paths are calculated and accumulated. The corresponding devices on each path are compared and located based on the trend of the accumulated values. Devices with abnormal power transmission mutations are identified as deep fault nodes, and the fault location result is generated. Based on the deviation concentration point location (grid 4), the adjacent node path is identified: device B103 → device C201 (distance 3m). The power transmission value per second from 15:00 to 15:05 is collected. B103 output: 1020, 1019, 1018, 1017, 1015kW, C201 input: 1018, 1017, 1016, 1015, 1012kW. The transmission difference per second is calculated: |102 0-1018|=2, |1019-1017|=2, |1018-1016|=2, |1017-1015|=2, |1015-1012|=3kW, cumulative difference = 2+2+2+2+3=11kW. Trend analysis shows a sudden increase to 3kW in the 5th second (2kW in the first 4 seconds). The abnormal device B103 is located (the output suddenly drops by 5kW). The fault location result is generated: B103 is abnormal.

[0045] See also Figure 7 , the fault location system of the smart energy big data monitoring system includes: The power monitoring module monitors the power input and output values ​​of power equipment and analyzes the power loss rate. It selects areas exceeding the power judgment threshold as compression areas, extracts devices at the area boundary, calculates the power tangential change rate through the gradient boosting decision tree, generates a boundary anomaly set, and transmits it to the boundary detection module. The boundary detection module compares the power tangential change rate in the boundary anomaly with the upper and lower thresholds, removes devices that do not reach the lower threshold, inserts devices that exceed the upper threshold into the compression area, and detects isolated devices based on the path connectivity between the device and the area center. It generates an isolated device set and passes it to the isolated device matching module; The isolated device matching module calculates the power flow vector of the isolated device based on the isolated device set, performs similarity matching on the power flow vectors of the remaining devices in the region, merges the matched isolated devices, generates a new compressed region set, and passes it to the multi-grid analysis module; The multi-grid analysis module divides the multi-layer grid based on the new compressed region set, compares the multi-grid power input and output differences at multiple resolutions, generates an energy deviation sequence through wavelet transform, and transmits it to the fault location module; The fault location module uses the energy deviation sequence to analyze the directional consistency of the deviation points at multiple resolutions. If the deviation exceeds the average deviation, the abnormal concentration point is marked, and the cumulative value of the path power difference is calculated to locate the deep fault device and generate the fault location result.

[0046] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A fault location method for a smart energy big data monitoring system, characterized in that: The following steps are involved: S1: Monitor the power input and output values ​​of power equipment and analyze the power loss rate. The areas exceeding the power judgment threshold are considered as compression areas. The device input gradient boosting decision tree at the area boundary is extracted to calculate the tangential change rate of the device power and generate a boundary compression set. S2: Compare the power tangential change rate of the devices in the boundary compression set with the upper and lower thresholds, remove devices that do not reach the lower threshold, insert devices that exceed the upper threshold into the compression area, and detect isolated devices based on the path connectivity between the devices and the area center to generate an isolated device set; S3: Based on the isolated device set, calculate the power flow vector of the isolated device, perform similarity matching on the power flow vectors of the remaining devices, merge the matched isolated devices, and generate a new compressed region set; S4: Dividing multiple layers of grids based on the new compressed region set, comparing the multi-grid power input and output deviations at multiple resolutions, and generating an energy deviation sequence; S5: Call the energy deviation sequence to analyze the directional consistency of the deviation points at multiple resolutions. If the deviation exceeds the average deviation, mark the deviation concentration point, and calculate the cumulative value of the path power difference to locate the deep fault device and generate a fault location result.

2. The fault location method of the smart energy big data monitoring system according to claim 1 is characterized in that: The boundary compression set specifically includes the compression area, boundary devices, and power tangential change rate; the isolated device set includes devices that do not reach the lower threshold, devices that exceed the upper threshold, and the location of isolated devices; the new compression area set includes isolated devices, similarity matching devices, and power flow vectors; the energy deviation sequence includes multi-layer grid power input deviation, multi-layer grid power output deviation, and energy deviation value; the fault location result specifically refers to the deviation concentration point, power difference cumulative value, and deep fault device.

3. The fault location method of the smart energy big data monitoring system according to claim 1 is characterized in that: The specific steps of S1 are: S101: Monitor the power input and output values ​​of power equipment, match them, and compare them one by one. Calculate the power loss values ​​of multiple devices within a fixed time period based on the difference between the input and output powers, and summarize them to generate a power loss rate data set. S102: Calling the power loss rate data set, identifying the device number and corresponding operating range whose loss rate exceeds the power judgment threshold, and based on the distribution position of devices whose loss rate continuously exceeds the threshold in the marked interval range, demarcating the compression area boundary according to the topological structure to generate the device compression boundary range; S103: Based on the device compression boundary range, call the power change records of the boundary devices to construct an input feature set, split the input feature set and train the node judgment value through the gradient boosting decision tree, calculate the power tangential change rate of the multi-node devices and integrate them to generate a boundary compression set.

4. The fault location method of the smart energy big data monitoring system according to claim 3 is characterized in that: The power judgment threshold is calculated by long-term monitoring of the input and output power of the device, and the upper limit of the power loss rate of the device is selected based on the device design requirements or industry standards.

5. The fault location method of the smart energy big data monitoring system according to claim 1 is characterized in that: The specific steps of S2 are: S201: Calling the boundary compression set, removing devices whose tangential change rates do not reach the lower limits of the upper and lower thresholds, then marking devices whose tangential change rates exceed the upper limits as devices entering the compression area, and recording the corresponding change rate data to generate a compression area device data set; S202: Based on the compressed regional device dataset, calculate the path connectivity between each device and the regional center, determine whether the device is an isolated device, filter out isolated devices that cannot be connected through other devices or regions, and generate an isolated device connectivity calculation result; S203: combining the isolated device connectivity calculation result, recording the device number and region information of each isolated device, and summarizing all the information to generate an isolated device set.

6. The fault location method of the smart energy big data monitoring system according to claim 5 is characterized in that: The upper and lower thresholds are set by the performance standards of the equipment or the maximum fluctuation range in the operating data. The upper threshold is set based on the performance standards of the equipment or the maximum fluctuation range in the operating data, and the lower threshold is set by the optimal standard value of the normal fluctuation of the equipment.

7. The fault location method of the smart energy big data monitoring system according to claim 1 is characterized in that: The specific steps of S3 are: S301: Constructing a power conduction direction map based on the isolated device set, obtaining power flow direction data for each device under multiple connection channels, analyzing power flow vectors by extracting instantaneous power transmission differences between connection points and physical direction vectors of corresponding channels, and obtaining a device power flow vector set; S302: Calling the device power flow vector set, calculating the similarity score between the power flow vector corresponding to each isolated device and the power flow vectors of other non-isolated devices one by one, selecting device pairs that meet a basic predetermined similarity matching threshold, and obtaining a similarity vector matching pair table; S303: merging isolated devices with matching relationships according to the similarity vector matching pair table, marking the merging identification numbers to which the matching devices belong, and aggregating isolated devices with the same identification number into the same region block to generate a new compressed region set; The basic limited similarity matching threshold is set by the similarity test result between the device power flow vectors.

8. The fault location method of the smart energy big data monitoring system according to claim 1 is characterized in that: The specific steps of S4 are: S401: Based on each region in the new compressed region set, apply grid division schemes of multiple resolutions to each region, extract power flow data of each grid, and obtain multi-layer grid power input and output results; S402: Calling the multi-layer grid power input and output results, calculating the corresponding power input and output deviation at each grid level, analyzing the energy deviation at multiple resolutions by comparing the power flow differences of multiple levels of grids, and obtaining the multi-resolution grid deviation; S403: Comparing deviation data at multiple resolution levels based on the multi-resolution grid deviation, calibrating the energy deviation range of each level of grid, and generating an energy deviation sequence.

9. The fault location method of the smart energy big data monitoring system according to claim 1 is characterized in that: The specific steps of S5 are: S501: calling the energy deviation sequence, analyzing the direction vectors corresponding to the deviation points in the sequence at multiple resolutions, counting the direction projection differences of each deviation point at multiple resolutions, and calculating the average fluctuation degree of the direction angle as a criterion for judging direction consistency, thereby generating a direction consistency interval; S502: Based on the directional consistency interval, filter the deviation points whose directional offset values ​​are greater than the interval mean and extract the position index. Calculate the power offset increment and directional drift amplitude at the grid resolution for each deviation point, and compare the degree of deviation with the remaining deviation points to obtain the position of the deviation concentration point. S503: Based on the position of the deviation concentration point, the power transmission value differences on the adjacent node paths are calculated and accumulated, and the corresponding devices on each path are compared and located according to the trend of the accumulated value changes. The devices with abnormal power transmission mutation are extracted as deep fault nodes to generate fault location results.

10. The fault location system of the smart energy big data monitoring system is characterized by: The system is used to implement the fault location method of the smart energy big data monitoring system according to any one of claims 1 to 9, and the system includes: The power monitoring module monitors the power input and output values ​​of power equipment and analyzes the power loss rate. It selects areas exceeding the power judgment threshold as compression areas, extracts devices at the area boundary, calculates the power tangential change rate through the gradient boosting decision tree, generates a boundary anomaly set, and transmits it to the boundary detection module. The boundary detection module compares the power tangential change rate in the boundary anomaly with the upper and lower thresholds, removes devices that do not reach the lower threshold, inserts devices that exceed the upper threshold into the compression area, and detects isolated devices based on the path connectivity between the device and the area center. The generated isolated device set is passed to the isolated device matching module; An isolated device matching module calculates the power flow vector of the isolated device based on the isolated device set, performs similarity matching on the power flow vectors of the remaining devices in the region, merges the matched isolated devices, generates a new compressed region set, and transmits it to the multi-grid analysis module; a multi-grid analysis module, which divides the multi-grid into multiple layers based on the new compressed region set, compares the multi-grid power input and output differences at multiple resolutions, generates an energy deviation sequence through wavelet transform, and transmits the energy deviation sequence to the fault location module; The fault location module calls the energy deviation sequence to analyze the directional consistency of the deviation point at multiple resolutions. If the deviation exceeds the average deviation, the abnormal concentration point is marked, and the cumulative value of the path power difference is calculated to locate the deep fault device and generate the fault location result.

Images