Fault prediction method and system for medium and low voltage distribution network

Through the combination of Apriori algorithm and neural network model, the problem of low timeliness and insufficient accuracy of power distribution network fault prediction is solved, more accurate and fast fault prediction is achieved, and the operational safety of the distribution network is improved.

CN119026723APending Publication Date: 2024-11-26XINZHOU POWER SUPPLY COMPANY STATE GRID SHANXI ELECTRIC POWER CORP
View PDF 0 Cites 4 Cited by

Patent Information

Application Number
CN202410875024.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-02
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The existing distribution network fault prediction problems are low timeliness and insufficient accuracy, especially in multi-factor correlation and disordered data environments, resulting in delays or errors in fault prediction.

Method used

The association rule mining technology based on the Apriori algorithm is used, combined with the neural network model, and the fault type of medium and low voltage distribution network is adjusted through the association rule set and the state of fault factors.

Benefits of technology

It improves the accuracy and rate of fault prediction, can more accurately identify fault patterns under the combined influence of multiple factors, reduces prediction delays and errors, and enhances the operational safety of the distribution network.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119026723A_ABST
    Figure CN119026723A_ABST
Patent Text Reader

Abstract

The invention discloses a fault prediction method and system for a medium and low voltage power distribution network, and belongs to the technical field of power distribution network fault monitoring. The method disclosed by the invention comprises the steps that a first fault factor state on each node in a power distribution network is collected, an association rule matched with the first fault factor state is obtained based on an association rule set, and the nodes in the power distribution network comprise a transformer substation, a ring main unit, a pole-mounted switch and a branch box; performing association coefficient adjustment on the first fault factor state of the current node according to an association rule to obtain a second fault factor state; and inputting the second fault factor state into the neural network model, and outputting and obtaining a fault prediction result of the current node. According to the method, potential historical fault factors are considered, and potential medium and low voltage power grid risks are predicted.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of power distribution network fault monitoring, and in particular relates to a fault prediction method and system for a medium and low voltage power distribution network. Background Art

[0002] With the rapid development of my country's economy, the proportion of electricity in energy supply has gradually increased, and the dependence of residential users and corporate factories on power safety production has increased, and the requirements for power supply reliability have become increasingly strong. The distribution network is an important link in the supply and distribution of electric energy in the power network, and is inseparable from the lives of the majority of residential users and the production of corporate factories. According to statistics, 80% of the power outages caused by user faults are caused by distribution network failures. Therefore, effectively reducing distribution network failures and improving power supply reliability play an important role in ensuring the daily production and life of the majority of users and the economic development of the whole society.

[0003] With the continuous deepening of the rural power grid transformation and upgrading project, the length of rural power distribution lines has been increasing, and the grid has been increasingly improved. However, due to the characteristics of long paths, many branches, and complex equipment in rural power distribution lines, the lines in some areas are complexly distributed and may be attacked by natural disasters or damaged by human forces at any time. These situations have put forward new requirements for ensuring the safe and reliable operation of the power grid and the management of distribution lines. In order to prevent the above-mentioned disasters from causing tripping and power outages on rural power lines and affecting people's production and life, power supply companies have been committed to the promotion and application of online monitoring systems for distribution network line faults and loads. At present, most of the online monitoring systems for distribution network faults and loads mainly use modern scientific and technological means such as solar power generation, wireless communication and information networks, which can monitor the power outage status and load current of the line in real time online, and send the collected fault alarm, power outage status and load current information back to the system master station; the master station performs data statistics, analysis, and topological calculation on the information to determine the fault area, thereby guiding the staff to quickly and accurately find the fault point. The application of these systems provides certain help for quickly finding line faults, improving the efficiency of relevant staff, and reducing labor intensity.

[0004] Distribution network fault prediction mainly targets trend faults with a long development cycle, faults with cumulative effects, and faults with certain statistical laws. It takes the current state of the equipment or system as the starting point, and combines the structural characteristics, operating parameters, meteorological factors, environmental conditions, and historical data of the known test object to predict its future state, so as to predict the risks of the distribution network, carry out maintenance or plans in advance, and ensure the stable and safe operation of the distribution network. Through the prediction of distribution network equipment faults, power supply companies can have a more comprehensive understanding of the full-cycle operating health of their own production equipment, thereby providing a scientific basis for the company's next stage of equipment technical transformation and maintenance and overhaul work, so as to reduce the subjectivity and blindness in the process of equipment technical transformation and maintenance and overhaul.

[0005] Under the current background of big data of distribution network, big data of distribution network provides opportunities and challenges for distribution network fault prediction. Due to the complexity of load and grid topology, and the statistical parameters include dynamic data such as current, voltage, temperature, humidity and wind speed in addition to static data such as geographic information, line information and equipment information, the factors causing the failure of distribution network are variable, and the factors causing the failure are not single. At this time, data mining technology is generally used to obtain multi-factor association rules, and the abnormal states of multiple associated factors are used to predict the possibility of a certain fault. However, these abnormal state data are not generated synchronously and are disordered, which will increase the time of fault prediction and diagnosis, especially when the occurrence of some distribution network faults is related to continuous bad working conditions, which has a cumulative effect, resulting in the fact that the collected time series feature data cannot be instantly accumulated to an abnormal state, causing the fault prediction to be delayed or errors due to early prediction. Since fault prediction has high requirements on the timeliness of the results, it is not only required that the prediction analysis model has good accuracy, but also that the rate of fault prediction should be improved. Summary of the invention

[0006] At least one aspect or advantage of the invention will be set forth in part in the following description, or may be obvious from the description, or may be learned by practice of the disclosed subject matter.

[0007] In order to overcome the above technical problems in the existing distribution network fault prediction, the present invention provides a medium and low voltage distribution network fault prediction method and system based on artificial intelligence. The present invention uses the Apriori algorithm to mine the potential correlation between fault factors, inputs the fault factor sequence with correlation into the neural network model, and the model makes the final fault prediction result output according to the training parameters.

[0008] According to a first aspect of the present invention, a method for predicting a fault in a medium and low voltage power distribution network comprises:

[0009] Collecting the first fault factor state on the nodes in the power distribution network, and obtaining the association rule matching the first fault factor state based on the association rule set, wherein the nodes in the power distribution network include a substation, a ring network cabinet, a pole switch and a branch box;

[0010] According to the correlation rule, the correlation coefficient of the first fault factor state of the current node is adjusted to obtain the second fault factor state;

[0011] Obtaining a fault prediction result of the current node according to the second fault factor state;

[0012] The process of obtaining the association rule set includes:

[0013] Input the fault factor status data into the association rule algorithm model to generate frequent item sets about the status data;

[0014] Obtain association rule sets based on frequent item sets;

[0015] The adjustment of the correlation coefficient includes completing the missing fault factors in the first fault factor based on the correlation rule according to the matching degree between the state of the first fault factor and the correlation rule.

[0016] According to one embodiment of the present invention, fault prediction is performed based on a neural network, and the acquisition process of the neural network model includes:

[0017] Collect statistics on distribution network fault sample data, including fault factor status data and fault results on fault nodes;

[0018] The neural network model is trained using the fault factor status data and fault results to obtain a neural network model for distribution network fault prediction.

[0019] According to one embodiment of the present invention, the fault factor states include vibration, gas concentration, smoke, voltage, current, frequency, temperature, humidity, wind speed and intrusion factors, and the fault prediction results include lightning short circuit fault, overload short circuit fault, equipment aging short circuit fault, tree-touching line short circuit fault, climate factor one-way grounding fault, animal-induced one-way grounding fault, equipment line break fault, and weather-induced line break fault.

[0020] According to an embodiment of the present invention, the association rule algorithm model is an April or FP-Tree algorithm model.

[0021] According to an embodiment of the present invention, the process of acquiring the association rule set includes:

[0022] Calculate the support of all single fault factors, select the fault factors that meet the threshold setting to establish frequent item sets, and in each subsequent traversal, use the frequent item set obtained in the previous traversal as the seed item set, substitute the support threshold, and generate a new frequent item set; among them, the newly generated frequent items have one more item than the seed item set;

[0023] Repeat the above steps until no new frequent item sets can be found;

[0024] All frequent item sets generated in this process are association rule sets.

[0025] According to an embodiment of the present invention, obtaining the rule associated with the first fault factor state based on the association rule set includes:

[0026] The fault factor status data monitored on the statistical node is formed into a fault factor set, and then the fault factor set is matched with the association rule set to find the most similar association rule; the matching algorithm used during matching is the cosine algorithm or the Pearson algorithm.

[0027] According to an embodiment of the present invention, the completion according to the matching degree between the first fault factor state and the associated rule includes:

[0028] determining a proximity based on the first fault factor state and the associated rule;

[0029] Determine the to-be-completed value of the to-be-completed fault factor based on the historical fault data and the first fault factor state associated by the association rule;

[0030] The completion value is determined based on the value to be completed and the approximation.

[0031] According to a second aspect of the present invention, a fault prediction system for a medium and low voltage power distribution network is characterized by comprising:

[0032] An associated rule acquisition unit is used to collect the first fault factor state on the nodes in the power distribution network, and obtain the associated rule matching the first fault factor state based on the associated rule set, wherein the nodes in the power distribution network include substations, ring network cabinets, pole switches and branch boxes;

[0033] A second fault factor state acquisition unit is used to adjust the correlation coefficient of the first fault factor state of the current node according to the correlation rule to obtain the second fault factor state;

[0034] A fault prediction unit, used to obtain a fault prediction result of the current node according to the second fault factor state;

[0035] The process of obtaining the association rule set includes:

[0036] Input the fault factor status data into the association rule algorithm model to generate frequent item sets about the status data;

[0037] Obtain association rule sets based on frequent item sets;

[0038] The adjustment of the correlation coefficient includes completing the missing fault factors in the first fault factor based on the correlation rule according to the matching degree between the state of the first fault factor and the correlation rule.

[0039] In one embodiment of the present invention, fault prediction is performed based on a neural network, and the acquisition process of the neural network model includes:

[0040] Collect statistics on distribution network fault sample data, including fault factor status data and fault results on fault nodes;

[0041] The neural network model is trained using the fault factor status data and fault results to obtain a neural network model for distribution network fault prediction.

[0042] According to one embodiment of the present invention, the fault factor states include vibration, gas concentration, smoke, voltage, current, frequency, temperature, humidity, wind speed and intrusion factors, and the fault prediction results include lightning short circuit fault, overload short circuit fault, equipment aging short circuit fault, tree-touching line short circuit fault, climate factor one-way grounding fault, animal-induced one-way grounding fault, equipment line break fault, and weather-induced line break fault.

[0043] According to an embodiment of the present invention, the association rule algorithm model is an April or FP-Tree algorithm model.

[0044] According to an embodiment of the present invention, obtaining the rule associated with the first fault factor state based on the association rule set includes:

[0045] The fault factor status data monitored on the statistical node is formed into a fault factor set, and then the fault factor set is matched with the association rule set to find the most similar association rule; the matching algorithm used during matching is the cosine algorithm or the Pearson algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 , a schematic flow chart of a fault prediction method for a power distribution network according to the present invention. DETAILED DESCRIPTION

[0047] The present disclosure will now be discussed with reference to several exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and thus implement the present disclosure, rather than implying any limitation on the scope of the present disclosure.

[0048] Please refer to Figure 1 According to one embodiment of the present invention, a fault prediction method for a medium and low voltage power distribution network includes:

[0049] Collecting the first fault factor state on the nodes in the power distribution network, and obtaining the association rule matching the first fault factor state based on the association rule set, wherein the nodes in the power distribution network include a substation, a ring network cabinet, a pole switch and a branch box;

[0050] According to the correlation rule, the correlation coefficient of the first fault factor state of the current node is adjusted to obtain the second fault factor state;

[0051] Obtaining a fault prediction result of the current node according to the second fault factor state;

[0052] The process of obtaining the association rule set includes:

[0053] Input the fault factor status data into the association rule algorithm model to generate frequent item sets about the status data;

[0054] Obtain association rule sets based on frequent item sets;

[0055] The adjustment of the correlation coefficient includes completing the missing fault factors in the first fault factor based on the correlation rule according to the matching degree between the state of the first fault factor and the correlation rule.

[0056] Furthermore, the fault prediction can be performed by a neural network model, and the acquisition process of the neural network model includes:

[0057] Collect statistics on distribution network fault sample data, including fault factor status data and fault results on fault nodes;

[0058] The neural network model is trained using the fault factor status data and fault results to obtain a neural network model for distribution network fault prediction.

[0059] The inventive concept of the present invention is further explained as follows.

[0060] At present, most distribution network fault and load online monitoring systems can only accurately judge and find the area where the short-circuit fault that caused the line tripping occurs, but often cannot accurately judge and find the frequent grounding faults on the line, and lack effective monitoring and early warning management for other operating parameters of the distribution network line operation status. The main reason is that due to the complex structure of the distribution network, the large number of equipment on the power supply line, and the complex and changeable operating conditions and environmental conditions, the distribution network line fault is not only affected by a single factor within the system, but more by the combined influence of multiple factors inside and outside the system. Common fault data monitoring methods generally stay within the system, such as monitoring of equipment line current, voltage, power, harmonics, electricity, phase angle and other operating parameters, while the causes of distribution network failures mainly include self-factors, natural factors and human factors.

[0061] ① Personal factors

[0062] The distribution network’s own factors include operational factors and equipment factors.

[0063] 1) Operational factors

[0064] The operating factors that affect distribution network failures are mainly current, voltage, and grid harmonics. The current and voltage in the distribution network are prone to cause problems such as heavy overload, low voltage, and three-phase imbalance. Heavy overload and low voltage not only cause the distribution network equipment to deviate from the rated operating conditions, but also cause the operating temperature of the equipment to rise; three-phase imbalance will increase the energy consumption of the distribution network, and overload of the heavy load phase will also cause the temperature of the equipment to rise. Grid harmonics refer to waves that are different from the power frequency of the AC power grid. Harmonics can cause problems such as equipment overheating, vibration, and noise. Harmonics can also cause resonance, and further amplified harmonics will cause greater harm. These operating factors of the distribution network may affect the performance and life of the distribution network equipment.

[0065] 2) Equipment factors

[0066] The factors of distribution network equipment are mainly line type, line length, operation time, equipment manufacturer and medical history. Distribution network lines are divided into overhead lines, cable lines and mixed lines, and the types and lengths of different lines are different. As the operation time of distribution network equipment increases, the degree of aging will also increase, which will affect its performance. Distribution network equipment may sometimes have family defects, and the manufacturer is a key factor affecting the quality of equipment. The maintenance history of distribution equipment is also an important factor, because equipment that has been repaired many times is prone to failure and its life will be reduced. Differences in distribution network equipment factors will lead to different failure probabilities.

[0067] ②Natural factors

[0068] The natural factors affecting distribution network failures include meteorological factors, geographical factors, and animal factors.

[0069] 1) Meteorological factors

[0070] The power distribution network is exposed outdoors for a long time, and the power distribution equipment will be affected by meteorological factors such as temperature, lightning strikes, wind, rainfall, snow accumulation and ice. If the outside temperature is too high or too low, it will affect the withstand voltage and insulation characteristics of the power distribution equipment, increasing the risk of failure; lightning strikes will directly cause equipment overvoltage, destroying the equipment in an instant and causing irreversible damage; excessive wind can easily cause foreign objects to hang on the wires, and can also cause the feeder to dance violently, leading to disconnection; rainfall increases the ambient temperature, which can easily cause the line to get damp and the equipment to rust; snow accumulation and ice accumulation will increase the weight of the distribution lines and equipment, causing unbalanced line tension, which can easily break the line, and the insulation strength of the line will also be reduced after ice coverage.

[0071] 2) Geographical factors

[0072] Geographical factors include power supply area, substation, tree barriers and mountains. Power supply areas are divided into A+, A, B, C, D and E. Different power supply areas have different load densities and failure rates. Distribution lines belong to different substations, so their pollution levels will be different. When the distribution network passes through parks or forests with dense trees, it is very easy for branches to touch the lines and trees to fall and break the lines. Hillsides with soft soil are also prone to landslides, causing poles to tilt or fall.

[0073] 3) Animal factors

[0074] Animal influencing factors are mainly small animals such as birds, mice and snakes. The impact of birds on distribution network faults includes bird nesting, bird flight and bird droppings flashover, which can easily lead to line grounding or short circuit faults; mice climbing onto distribution network lines or crawling into distribution equipment can easily cause short circuits and may also bite off insulation and lines; snakes crawling into ring network cabinets and cable boxes can also easily cause short circuits.

[0075] ③Human factors

[0076] Human influencing factors mainly include improper operation and maintenance, external force damage and user reasons.

[0077] 1) Improper operation and maintenance

[0078] Improper operation and maintenance construction mainly results from untimely operation and maintenance and poor construction quality.

[0079] 2) External force damage

[0080] External force damage includes vehicles hanging on wires, cables being dug up during construction, and circuit breakers being cut by felled trees.

[0081] 3) User reasons

[0082] Electricity consumption by users includes industrial and residential electricity consumption. The number of user lines and equipment is increasing rapidly. Users’ private wiring, improper use of private equipment, inadequate management and untimely maintenance may all cause distribution network line failures or protective tripping, causing power outages. In recent years, the construction of new energy vehicle charging piles has brought many uncertainties to the distribution network.

[0083] The fault categories of the distribution network can be divided according to the fault phenomena. Common fault types of the distribution network include short circuit fault, single-phase grounding fault and line break fault.

[0084] 1) Short circuit fault

[0085] A short circuit fault refers to a fault caused by a short circuit in a circuit or part of a circuit. Short circuit faults can be divided into single-phase short circuit, multi-phase short circuit and multi-phase grounding, etc., which are common fault types in distribution networks. The hazards of short circuit faults in distribution networks are instantaneous overcurrent and voltage drop, which will cause line protection tripping, and may also cause the temperature of distribution network equipment to rise rapidly, causing impact on the equipment, or even damaging the equipment, and more likely to induce a chain of faults. The factors that cause short circuit faults in distribution networks generally include lightning strikes, overloads, equipment aging, and trees touching the lines.

[0086] 2) Single-phase ground fault

[0087] The grounding line is a line directly connected to the earth. It is a protection device for the distribution line. It can be divided into three types: neutral, lightning protection, and tower grounding. Single-phase grounding refers to the single-phase grounding of the small current grounding system of the distribution network. A small current flows through the grounding point, which generally does not cause a short circuit. The system can continue to operate for a period of time without affecting the continuous power supply to users. However, if the distribution network operates for a long time when a single-phase grounding fault occurs, overvoltage may be generated and even a phase-to-phase short circuit may be caused. Single-phase grounding faults often occur in humid or rainy weather, and may also be caused by factors such as tree barriers and small animals. It can be caused by a variety of factors.

[0088] 3) Disconnection failure

[0089] The disconnection faults in the distribution network are mostly caused by the wire’s own factors and external force damage. The wire’s own factors include wire aging and rust, and the external force damage includes disconnection caused by natural or human factors, which is sporadic.

[0090] The distribution network has a complex line structure and a changeable operating environment. Its failures are often not caused by a single reason, but the result of the combined effect of multiple factors. Among the many factors that cause a certain failure, they are also divided into major failure factors and minor failure factors. For example, equipment factor failures may be affected by operating factors and meteorological factors in addition to the equipment itself, but the equipment itself is the main failure factor. The premise of fault prediction is to analyze which combination sequence of fault factors and their parameter values ​​can lead to the occurrence of a specific fault. Due to the large amount of collected monitoring data, it is impossible to effectively classify them by manual analysis methods. Therefore, it is necessary to use data mining technology to find the mapping relationship between different fault factor sequences and faults.

[0091] The present invention first establishes two different models, the first model is a rule set model, and the second model is a neural network prediction model.

[0092] The rule set model is based on the association of various fault factors and quantifies various factors obtained from the detection of each node. For example, the wind speed is calculated according to the actual wind speed, and the humidity and temperature can be input with corresponding numerical values, or through numerical mapping, such as mapping the temperature from -30 to 50°C to the interval of 0-1 respectively; for non-quantifiable factors, such as intrusion factors, the corresponding numerical item is marked as 1 when an abnormal proximity is detected within a time period, or the corresponding numerical item is marked as 0 when an abnormal proximity is detected within a time period. A time period here can be set to 1h or 24h. Due to the existence of intrusion factors, it does not necessarily lead to circuit failures. Setting a shorter time will result in the loss of potential fault information. After determining the measurement of the fault factor, an association model between the fault factor and the fault is constructed according to the historical fault data. The association model includes multiple rule sets to form a searchable or matchable association rule set. A potential most similar rule can be matched according to a series of input parameters. This similar rule matches at least one historical fault information, and the corresponding fault factor and the factor obtained by the current actual measurement generally differ. Based on the consideration of approximation, the missing data of the current environmental factors are used to adjust the approximation of the relevant data and the numerical value when the fault occurs to obtain an updated fault factor.

[0093] The neural network prediction model can use a common RNN model. During training, the fault factors quantified above can be input into the neural network to output the expected fault category. Based on the difference between the expected fault category and the actual fault category, the loss value during training is obtained, and the neural network model is adjusted.

[0094] In one embodiment of the present invention, the selected RNN model has the following structure: the number of neurons in the first layer is consistent with the number m of input fault factor states, the number of neurons in the first hidden layer is 2m, and it is fully connected to the input layer, using the tanh activation function, the number of neurons in the second hidden layer is m, and it is fully connected to the first hidden layer, using the ReLu activation function, the number of output layers is consistent with the number of fault types, which is consistent with the number of second hidden layers, and uses the Sigmod activation function, and the learning speed is selected as 0.001. A similar structure can be used to train historical data, where the historical data is real and has not been adjusted by the rule set model. By setting data from different sources, the factors of correlation can be mined, and maintenance personnel can troubleshoot faults based on the prediction results.

[0095] In one embodiment of the present invention, the data input to the RNN is the second fault factor status of multiple nodes. By tracking the status of multiple nodes, it is avoided that only tracking a single node misses potential factors, thereby interfering with the risk fault factor troubleshooting to be performed globally.

[0096] It should be understood that the fault factor state of the present invention is a series of specific measurable values ​​reflecting the environment or equipment, and the association rule set is the support of each factor in the fault for the fault.

[0097] In one embodiment provided by the present invention, the April algorithm is used to mine the correlation between fault factors. For example, based on the previously collected distribution network fault data as sample data, each sample data contains the fault factor status data (including environmental parameters / voltage / current / frequency, etc.) and the diagnosed fault results (such as short circuit, grounding or disconnection fault) on the fault node when the fault occurs. For internal factors of the distribution network system, relevant status data can be obtained through system logs, equipment logs, program logs, and manual record logs of maintenance personnel; and for external factors of the distribution network system, they can be obtained from external data platforms, such as meteorological information can be obtained from weather monitoring data disclosed by the Meteorological Bureau.

[0098] In one embodiment provided by the present invention, when it is necessary to predict a fault on a fault node, the fault factor status data that has been monitored on the fault node is counted to form a fault factor set, and then the fault factor set is matched with the strong association rule set to find the most similar association rule. At this time, similarity algorithms such as cosine and Pearson can be used for similarity matching. The most similar association rule found here includes the fault factors that have been detected in the fault factor set, and also includes fault factors that may arise in the future. The fault factors that do not appear in the association rules are added to the fault factor set to obtain the input data set of the neural network, and finally the trained neural network is used to calculate the output prediction result.

[0099] In one embodiment provided by the present invention, the fault factor status data of the sample data mentioned above is used as training input material, and the fault classification result corresponding to the fault factor status data is used as output material to train the neural network to obtain the final decision model for distribution network fault classification.

[0100] According to one embodiment of the present invention, the fault factor states include vibration, gas concentration, smoke, voltage, current, frequency, temperature, humidity, wind speed and intrusion factors, and the fault prediction results include lightning short circuit fault, overload short circuit fault, equipment aging short circuit fault, tree-touching line short circuit fault, climate factor one-way grounding fault, animal-induced one-way grounding fault, equipment line break fault, and weather-induced line break fault.

[0101] By setting different failure factors, the correlation between different types of environments and stocks can be explored.

[0102] When testing sets, training sets, and deploying, normal data can be processed to prevent the neural network model from learning incorrect features. For example, for a certain node, when the input factor is temperature, its normal operating temperature range is 15-29°C, then the data in this range can be set to 0, temperatures below this range can be mapped to 0-0.5, and temperatures above this range can be mapped to 0.5-1.0. The temperature in this temperature range is set to 0.5 to avoid unnecessary interference.

[0103] According to an embodiment of the present invention, the association rule set is obtained using April or FP-Tree algorithm.

[0104] The contribution of each fault factor and fault type can be obtained through April or FP-Tree algorithm, on this basis, hidden factors can be further discovered and potential factors can be adjusted.

[0105] According to an embodiment of the present invention, the process of acquiring the association rule set includes:

[0106] Calculate the support of all single fault factors, select the fault factors that meet the threshold setting to establish frequent item sets, and in each subsequent traversal, use the frequent item set obtained in the previous traversal as the seed item set, substitute the support threshold, and generate a new frequent item set; among them, the newly generated frequent items have one more item than the seed item set;

[0107] Repeat the above steps until no new frequent item sets can be found;

[0108] All frequent item sets generated in this process are association rule sets.

[0109] This method can quickly determine the correlation between various fault factors, avoiding the defect of being unable to comprehensively locate the fault during single factor analysis.

[0110] According to an embodiment of the present invention, obtaining the rule associated with the first fault factor state based on the association rule set includes:

[0111] The fault factor status data monitored on the statistical node is formed into a fault factor set, and then the fault factor set is matched with the association rule set to find the most similar association rule; the matching algorithm used during matching is the cosine algorithm or the Pearson algorithm.

[0112] In this way, historical faults related to the current state can be determined, and then similar associated rules can be determined, and corresponding factors can be completed based on the associated rules.

[0113] According to an embodiment of the present invention, the completion according to the matching degree between the first fault factor state and the associated rule includes:

[0114] determining a proximity based on the first fault factor state and the associated rule;

[0115] Determine the to-be-completed value of the to-be-completed fault factor based on the historical fault data and the first fault factor state associated by the association rule;

[0116] The completion value is determined based on the value to be completed and the approximation.

[0117] In this way, an adjusted input vector with potential factors can be determined. At this time, the input vector has reflected the risk factor. If the prediction result corresponding to this risk factor has a greater risk, that is, the fault type is output, then there may be a greater risk; and when there is no fault in the input, the possibility of risk is low.

[0118] Taking temperature as the completed value as an example, in the input vector, if the value corresponding to the temperature in the historical fault data is 0.7 and the determined approximation is 0.8, then the completed temperature value is 0.5+(0.7-0.5)*0.8=0.66, where the adjustment value is the product of the difference term and the approximation.

[0119] According to one embodiment of the present invention, a fault prediction system for a power distribution network is characterized by comprising:

[0120] An associated rule acquisition unit is used to collect the first fault factor state on the nodes in the power distribution network, and obtain the associated rule matching the first fault factor state based on the associated rule set, wherein the nodes in the power distribution network include substations, ring network cabinets, pole switches and branch boxes;

[0121] A second fault factor state acquisition unit is used to adjust the correlation coefficient of the first fault factor state of the current node according to the correlation rule to obtain the second fault factor state;

[0122] A fault prediction unit, used to obtain a fault prediction result of the current node according to the second fault factor state;

[0123] The association rule algorithm model is a similarity algorithm model;

[0124] The process of obtaining the association rule set includes:

[0125] Input the fault factor status data into the association rule algorithm model to generate frequent item sets about the status data;

[0126] Obtain association rule sets based on frequent item sets;

[0127] The adjustment of the correlation coefficient includes completing the missing fault factors in the first fault factor based on the correlation rule according to the matching degree between the state of the first fault factor and the correlation rule.

[0128] In one embodiment of the present invention, a fault is predicted by a neural network, and the acquisition process of the neural network model includes:

[0129] Collect statistics on distribution network fault sample data, including fault factor status data and fault results on fault nodes;

[0130] The neural network model is trained using the fault factor status data and fault results to obtain a neural network model for distribution network fault prediction.

[0131] According to one embodiment of the present invention, the fault factor states include vibration, gas concentration, smoke, voltage, current, frequency, temperature, humidity, wind speed and intrusion factors, and the fault prediction results include lightning short circuit fault, overload short circuit fault, equipment aging short circuit fault, tree-touching line short circuit fault, climate factor one-way grounding fault, animal-induced one-way grounding fault, equipment line break fault, and weather-induced line break fault.

[0132] According to an embodiment of the present invention, the association rule set is obtained using April or FP-Tree algorithm.

[0133] According to an embodiment of the present invention, obtaining the rule associated with the first fault factor state based on the association rule set includes:

[0134] The fault factor status data monitored on the statistical node is formed into a fault factor set, and then the fault factor set is matched with the association rule set to find the most similar association rule; the matching algorithm is the cosine algorithm or the Pearson algorithm.

[0135] Those of ordinary skill in the art will appreciate that the modules and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0136] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and equipment can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0137] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0138] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0139] In addition, each functional module in the embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0140] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the energy-saving signal sending / receiving method of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0141] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the present application.

[0142] It should be understood that the size of the sequence number of each step in the content of the invention and the embodiments of the present invention does not absolutely mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention. For the purpose of example and description, the foregoing description of the implementation of the present disclosure has been given. The foregoing description is not exhaustive and is not intended to limit the present disclosure to the exact form disclosed. Various deformations and modifications may exist according to the above teachings, or various deformations and modifications may be obtained from the practice of the present disclosure. These embodiments are selected and described to illustrate the principles of the present disclosure and its practical application, so that those skilled in the art can use the present disclosure in various embodiments and various modifications suitable for the specific purpose conceived.

Claims

1. A fault prediction method for a medium and low voltage power distribution network, characterized in that: include: Collecting the first fault factor state on the nodes in the power distribution network, and obtaining the association rule matching the first fault factor state based on the association rule set, wherein the nodes in the power distribution network include a substation, a ring network cabinet, a pole switch and a branch box; According to the correlation rule, the correlation coefficient of the first fault factor state of the current node is adjusted to obtain the second fault factor state; Obtaining a fault prediction result of the current node according to the second fault factor state; The process of obtaining the association rule set includes: Input the fault factor status data into the association rule algorithm model to generate frequent item sets about the status data; Obtain association rule sets based on frequent item sets; The adjustment of the correlation coefficient includes completing the missing fault factors in the first fault factor based on the correlation rule according to the matching degree between the state of the first fault factor and the correlation rule.

2. A method for predicting faults in a medium and low voltage power distribution network as claimed in claim 1, characterized in that: The fault factor states include vibration, gas concentration, smoke, voltage, current, frequency, temperature, humidity, wind speed and intrusion factors; the fault prediction results include lightning short circuit fault, overload short circuit fault, equipment aging short circuit fault, tree-line short circuit fault, climate factor one-way grounding fault, animal-induced one-way grounding fault, equipment line break fault, and weather-induced line break fault.

3. A method for predicting faults in a medium and low voltage power distribution network as claimed in claim 1, characterized in that: The association rule algorithm model is an Apriori or FP-Tree algorithm model.

4. A method for predicting faults in a medium and low voltage power distribution network as claimed in claim 3, characterized in that: The process of obtaining the association rule set includes: Calculate the support of all single fault factors, select the fault factors that meet the threshold setting to establish frequent item sets, and in each subsequent traversal, use the frequent item set obtained in the previous traversal as the seed item set, substitute the support threshold, and generate a new frequent item set; among them, the newly generated frequent items have one more item than the seed item set; Repeat the above steps until no new frequent item sets can be found; All frequent item sets generated in this process are association rule sets.

5. A method for predicting faults in a medium and low voltage power distribution network as claimed in claim 1, characterized in that: The obtaining of the association rule based on the association rule set and matching the first fault factor state includes: The fault factor status data monitored on the statistical node is formed into a fault factor set, and then the fault factor set is matched with the association rule set to find the most similar association rule; the matching algorithm used during matching is the cosine algorithm or the Pearson algorithm.

6. A method for predicting faults in a medium and low voltage power distribution network as claimed in claim 1, characterized in that: The completing according to the matching degree between the first fault factor state and the associated rule includes: determining a proximity based on the first fault factor state and the associated rule; Determine the to-be-completed value of the to-be-completed fault factor based on the historical fault data and the first fault factor state associated by the association rule; The completion value is determined based on the value to be completed and the approximation.

7. A fault prediction system for medium and low voltage power distribution networks, characterized in that: The invention is characterized by comprising: An associated rule acquisition unit is used to collect the first fault factor state on the nodes in the power distribution network, and obtain the associated rule matching the first fault factor state based on the associated rule set, wherein the nodes in the power distribution network include substations, ring network cabinets, pole switches and branch boxes; A second fault factor state acquisition unit is used to adjust the correlation coefficient of the first fault factor state of the current node according to the correlation rule to obtain the second fault factor state; A fault prediction unit, used to obtain a fault prediction result of the current node according to the second fault factor state; The process of obtaining the association rule set includes: Input the fault factor status data into the association rule algorithm model to generate frequent item sets about the status data; Obtain association rule sets based on frequent item sets; The adjustment of the correlation coefficient includes completing the missing fault factors in the first fault factor based on the correlation rule according to the matching degree between the state of the first fault factor and the correlation rule.

8. A fault prediction system for medium and low voltage power distribution networks as claimed in claim 7, characterized in that: The fault factor states include vibration, gas concentration, smoke, voltage, current, frequency, temperature, humidity, wind speed and intrusion factors; the fault prediction results include lightning short circuit fault, overload short circuit fault, equipment aging short circuit fault, tree-line short circuit fault, climate factor one-way grounding fault, animal-induced one-way grounding fault, equipment line break fault, and weather-induced line break fault.

9. A fault prediction system for medium and low voltage power distribution networks as claimed in claim 7, characterized in that: The association rule algorithm model is an Apriori or FP-Tree algorithm model.

10. A fault prediction system for medium and low voltage power distribution networks as claimed in claim 7, characterized in that: The obtaining of the association rule based on the association rule set and matching the first fault factor state includes: The fault factor status data monitored on the statistical node is formed into a fault factor set, and then the fault factor set is matched with the association rule set to find the most similar association rule; the matching algorithm used during matching is the cosine algorithm or the Pearson algorithm.

Citation Information

Cited By

  • Power distribution network medium-voltage line scheduling method and device

    CN119382130A

  • A method and device for dispatching medium-voltage lines in a distribution network

    CN119382130B

  • Power equipment fault state monitoring method and device, electronic equipment and storage medium

    CN120454311A

  • Power equipment fault state monitoring method and device, electronic equipment and storage medium

    CN120454311B