An artificial intelligence-based power line state evaluation method and system

Abstract

The application discloses an artificial intelligence-based power line state evaluation method and system, relates to the technical field of power system operation and maintenance, and comprises the following steps: simulating an input signal based on a line level to obtain initial ideal data; obtaining historical fault data of a line to be evaluated, and combining the initial ideal data to construct a health value evaluation model; evaluating a current monitoring signal according to the health value evaluation model to obtain health evaluation data; determining the resistance of the line to be evaluated to each fault based on the health evaluation data to obtain resistance data; statistically processing the monitoring signal and the health evaluation data of the line to be evaluated based on time data to obtain a risk prediction curve, determining the change rate of each index item in the risk prediction curve, and predicting the occurrence possibility of a corresponding fault risk based on the change rate of each index item to obtain a risk probability. The application has the effect of improving the accuracy of line state evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system operation and maintenance, in particular to a power line state evaluation method and system based on artificial intelligence. BACKGROUND

[0002] Traditional power line inspection mainly relies on manual visual inspection, infrared temperature measurement, unmanned aerial vehicle inspection and the like, which is low in efficiency, high in cost, limited in coverage, and greatly affected by weather and terrain. The existing online monitoring devices (such as video monitoring, microclimate, ice monitoring, etc.) generate massive data, but lack effective intelligent analysis means, and the data value is not fully excavated. The early warning of potential risks of the line (such as tree barriers, loose fittings, deteriorated insulators, external damage risks, and meteorological disaster impacts) is not timely and accurate. The line state evaluation relies on a single data source or simple threshold judgment, and the evaluation results are not comprehensive and objective, which is difficult to support fine operation and maintenance decisions, and there is room for improvement. SUMMARY

[0003] In order to improve the accuracy of line state evaluation, the present application provides a power line state evaluation method and system based on artificial intelligence.

[0004] In a first aspect, the present application provides a power line state evaluation method based on artificial intelligence, which adopts the following technical solution:

[0005] A power line state evaluation method based on artificial intelligence, comprising:

[0006] obtaining the line grade of the line to be evaluated and performing grade matching between the line grade of the line to be evaluated and the built-in line specification table to determine the line specification under the corresponding line grade;

[0007] obtaining the input signal of the line endpoint and simulating the input signal of the line endpoint based on the line specification of the line to be evaluated to determine the performance of the input signal under the line specification and obtain initial ideal data;

[0008] obtaining the historical fault data of the line to be evaluated and constructing a health value evaluation model of the line in combination with the initial ideal data;

[0009] obtaining the monitoring signal of the current state of the line to be evaluated and evaluating the monitoring signal according to the health value evaluation model to obtain health evaluation data;

[0010] based on the health evaluation data, determining the resistance of the current working state of the line to be evaluated to each fault to obtain resistance data, and predicting the fault risk of the line according to the resistance data to obtain the risk grade of the fault risk, and outputting and displaying the risk grade of each fault risk;

[0011] The monitoring signal of the to-be-evaluated line and the health evaluation data thereof are counted based on the time data to obtain a risk prediction curve, the change rate of each index item in the risk prediction curve is determined, the occurrence possibility of the corresponding fault risk is predicted based on the change rate of each index item, the risk probability is obtained, and the risk probability of each fault risk is output and displayed.

[0012] Preferably, S31, based on the time point when the historical fault occurs, the historical fault data before the fault occurs is marked as first fault data, and the historical fault data when the fault occurs is marked as second fault data;

[0013] S32, the first fault data and the second fault data are matched with index data, the index item that changes compared with the index data before the fault occurs when the fault occurs in the same historical fault event is determined, and the index item is marked as a key index item;

[0014] S33, the key index items of multiple historical fault events of the same type of fault are counted to determine a fault evaluation index;

[0015] S34, the index data of the fault evaluation index items of multiple fault events of the same type of fault are counted to determine fault judgment data;

[0016] S35, the initial ideal data is taken as the line data when the health value is the first health threshold, the fault judgment data is taken as the line data when the health value is the second health threshold, the relationship between the line data and the health value is constructed, and the health value evaluation model of each fault evaluation index is obtained.

[0017] Preferably, the key index items of multiple historical fault events of the same fault are counted to determine whether the distribution proportion of each key index item meets the Pareto distribution;

[0018] If it is determined that the distribution of the key index items meets the Pareto distribution, the distribution proportion of each key index item is taken as the credibility of the result that the corresponding key index item can reflect the line health value, the credibility data is obtained, and the key index item located in the head of the Pareto distribution is marked as a fault evaluation index;

[0019] If it is determined that the distribution of the key index items does not meet the Pareto distribution, it is determined that each index item does not have credibility, and a fault unpredictable signal is output.

[0020] Preferably, the initial ideal data is taken as the line data when the health value is the first health threshold, the fault judgment data is taken as the line data when the health value is the second health threshold, the relationship between the line data and the health value is constructed, and the initial health value evaluation model is obtained;

[0021] The quantity judgment is performed on the fault evaluation indexes of the same type of fault, and the index property of the fault evaluation indexes of the same type of fault is determined;

[0022] If the number of the fault evaluation indexes is equal to 1, the initial health value evaluation model is taken as the health value evaluation model corresponding to the fault;

[0023] If the number of the fault evaluation indexes is greater than 1, the fault evaluation data of the plurality of fault evaluation indexes are compared with each other to determine whether the data relationship between the plurality of fault evaluation indexes is the same;

[0024] If the data relationship between the plurality of fault evaluation indexes is the same, it is determined that the index property of the fault evaluation indexes is singularity, and the initial health value evaluation model is taken as the health value evaluation model of the fault evaluation indexes corresponding to the fault;

[0025] If the data relationship between the plurality of fault evaluation indexes is different, each fault evaluation index in the same fault event is evaluated based on the initial health value evaluation model to determine the health value of each fault evaluation index in the fault event, and an event health value set is obtained;

[0026] Second health threshold judgment is performed on the event health value set to determine whether there is at least one fault evaluation index whose health value reaches the second health threshold in each fault event;

[0027] If there is at least one fault evaluation index whose health value reaches the second health threshold in each fault event, it is determined that the index property of the fault evaluation indexes of the same type of fault is diversity, and the initial health value evaluation model is taken as the health value evaluation model of each fault evaluation index corresponding to the fault;

[0028] If there is a case where the health value of all fault evaluation indexes does not reach the second health threshold in the fault event, the index data of the key index items in the plurality of fault events are constructed, the comprehensive data of the fault category is determined according to the constructed data relationship, the relationship between the line data and the health value is constructed based on the comprehensive data and the initial ideal data, the second health value evaluation model is obtained, and the second health value evaluation model is taken as the health value evaluation model in the same type of fault event.

[0029] Preferably, the fault evaluation indexes of the to-be-evaluated line under each fault category are obtained, and the external factors that can affect the fault evaluation indexes are determined according to the fault evaluation indexes;

[0030] Based on the health evaluation data of the to-be-evaluated line in the current state, the condition required for the external factor to reduce the health value of the to-be-evaluated line from the current health value to the second health threshold is determined, and the condition is marked as first fault condition data;

[0031] The implementability of the first fault condition data is judged to determine the probability of reaching the first fault condition data under non-human conditions, and fault resistance data is obtained;

[0032] The fault risk level table of the line is matched based on the fault resistance data, and the risk level of the line to be evaluated relative to each fault category is obtained.

[0033] Preferably, the corresponding fault evaluation indicators in the monitoring signal of the line to be evaluated are subjected to index data statistics based on the fault category, and a risk prediction curve of the corresponding fault evaluation indicator is drawn;

[0034] The rate of change of the risk prediction curve is judged to obtain prediction change data;

[0035] The current indicator data of the fault evaluation indicator and the fault evaluation data of the corresponding indicator are subjected to difference calculation to obtain a predicted change value;

[0036] Based on the prediction change data and the predicted change value, the predicted change duration is determined, and the predicted change durations of each fault evaluation indicator within the same fault are compared with each other according to the fault category. The shortest predicted change duration is matched with the built-in risk probability table to determine the risk probability of the corresponding fault category and output the display.

[0037] Preferably, the fault evaluation indicators of each fault category are obtained, and the index data of the monitoring signal of the line to be evaluated on the corresponding fault evaluation indicator is determined based on the fault evaluation indicators;

[0038] The index data of each fault evaluation indicator is input into a health value evaluation model to obtain a health evaluation value of the fault evaluation indicator;

[0039] According to the fault category, the health evaluation values of the fault evaluation indicators of the same fault are statistically calculated to obtain health evaluation data of the corresponding fault.

[0040] If the number of fault evaluation indicators of the fault is 1, the health evaluation value of the fault evaluation indicator is taken as the health evaluation data of the fault;

[0041] If the number of fault evaluation indicators of the fault is greater than 1, the index properties of the fault evaluation indicators of the fault are obtained. If the index property is single, the mean of the health evaluation values of the multiple fault evaluation indicators of the fault is taken as the health evaluation data of the fault;

[0042] If the index property is diversity, the smallest health evaluation value of the health evaluation values of the multiple fault evaluation indicators of the fault is taken as the health evaluation data of the fault;

[0043] If the index property is regularity, the comprehensive data of the fault evaluation index under the fault is input into the second health value evaluation model for evaluation to obtain health evaluation data of the to-be-evaluated line compared with the fault.

[0044] In a second aspect, the application provides an artificial intelligence-based power line state evaluation system, which adopts the following technical solution:

[0045] An artificial intelligence-based power line state evaluation system comprises an initial state analysis module, an evaluation model construction module, and a line state evaluation and display module.

[0046] The initial state analysis module is configured to obtain the line grade of the to-be-evaluated line and perform grade matching between the line grade of the to-be-evaluated line and a built-in line specification table to determine the line specification under the corresponding line grade, obtain the input signal of the line endpoint, and simulate the input signal of the line endpoint based on the line specification of the to-be-evaluated line to determine the performance of the input signal under the line specification and obtain initial ideal data.

[0047] The evaluation model construction module is configured to obtain the historical fault data of the to-be-evaluated line and construct a health value evaluation model of the line in combination with the initial ideal data.

[0048] The line state evaluation and display module is configured to obtain the monitoring signal of the current state of the to-be-evaluated line, evaluate the monitoring signal based on the health value evaluation model to obtain health evaluation data, determine the resistance of the current working state of the to-be-evaluated line to each fault based on the health evaluation data to obtain resistance data, predict the fault risk of the line based on the resistance data to obtain the risk grade of the fault risk, and output and display the risk grade of each fault risk, and statistically analyze the monitoring signal of the to-be-evaluated line and its health evaluation data based on time data to obtain a risk prediction curve, determine the change rate of each index item in the risk prediction curve, predict the occurrence probability of the corresponding fault risk based on the change rate of each index item to obtain a risk probability, and output and display the risk probability of each fault risk.

[0049] In summary, the application has at least one of the following beneficial technical effects:

[0050] By utilizing big data statistics, a line specification table is constructed, the input requirement of the to-be-evaluated line is simplified, the initial state of the to-be-evaluated line is determined, and the applicability of the system to the to-be-evaluated line in different states is ensured. By statistically analyzing the historical fault data of the to-be-evaluated line and taking the initial ideal data as the ideal state of the to-be-evaluated line, a health value evaluation model for evaluating the health state of the line relative to each fault is constructed, and then the health value of the to-be-evaluated line relative to each fault category can be evaluated according to the current monitoring signal of the to-be-evaluated line, so that the evaluated health value is more targeted, and the accuracy of the evaluation result is improved. At the same time, by comparing the current state of the to-be-evaluated line with the fault condition when the fault occurs, the difficulty of the to-be-evaluated line to cause the corresponding fault is determined, the judgment accuracy of the resistance data is improved, and at the same time, the data change of the to-be-evaluated line is determined by statistically analyzing the monitoring signal of the to-be-evaluated line. Then, the probability of the to-be-evaluated line to cause a fault under the data change is determined to achieve the prediction effect of the fault risk, so that the prediction result is more accurate. The state evaluation of the line as a whole is converted to the state evaluation of the line relative to each fault, so that the evaluation result is more targeted, and the accuracy of the state evaluation result is improved.

[0051] By utilizing the initial ideal data and fault judgment data of each fault evaluation index, an initial health value evaluation model is constructed, and based on the initial health value evaluation model, targeted revision is carried out for different fault categories. Through quantity judgment, data relationship judgment and event health value judgment of the fault evaluation index, the initial health value evaluation model is adjusted according to different judgment results, so that the data evaluated by the adjusted health value evaluation model is more accurate, the accuracy of the evaluation result is improved, and the effectiveness of the health value evaluation model evaluation result is ensured.

[0052] By determining the external influencing factors corresponding to each fault evaluation index under each fault category, and combining the current state of the to-be-evaluated line to determine the condition required to reduce the health value of the fault evaluation index to the second health threshold under the corresponding external influencing factor, and taking the condition as the first fault condition data, finally, the probability that can be achieved under non-human conditions is determined through the implementability judgment of the first fault condition, and the resistance of the current line to sudden changes is determined, so that the evaluation is more accurate, and the accuracy of the risk assessment of each fault risk is improved. BRIEF DESCRIPTION OF DRAWINGS

[0053] Fig. 1 The step flow chart of the power line state evaluation method based on artificial intelligence of the present embodiment;

[0054] Fig. 2A module block diagram of the power line state evaluation system based on artificial intelligence for the embodiment.

[0055] Reference signs: 1, initial state analysis module; 2, evaluation model construction module; 3, line state evaluation and display module. DETAILED DESCRIPTION

[0056] The following will be described in detail below with reference to the accompanying Figs. 1-2 The application is further described in detail.

[0057] The embodiment of the application discloses a power line state evaluation method and system based on artificial intelligence.

[0058] Embodiment: as shown in the figure, the power line state evaluation method based on artificial intelligence comprises: Fig. 1

[0059] S1, the line grade of the line to be evaluated is obtained, and the line grade of the line to be evaluated is matched with the built-in line specification table to determine the line specification under the corresponding line grade; wherein the line grade is the line specification required by different regional spans or power transmission, and the line grade is obtained according to the line specification. The input signal of the line end point is simulated with the line specification to obtain the corresponding initial ideal data. Wherein the line specification table can be divided into cross-province line, cross-city line, cross-area line according to the length of transmission distance, and the divided conditions are corresponded to the line specification to obtain the line specification table. For example, the line specification of high-voltage distribution line is P1, the line specification of medium-voltage distribution line is P2, etc.

[0060] S2, the input signal of the line end point is obtained, and the input signal of the line end point is simulated based on the line specification of the line to be evaluated to determine the performance of the input signal under the line specification, and the initial ideal data is obtained.

[0061] S3, the historical fault data of the line to be evaluated is obtained, and the health value evaluation model of the line is constructed in combination with the initial ideal data; wherein the health value evaluation model is only used for evaluating the health value of each index item in each fault category in the line to be evaluated.

[0062] S4, the monitoring signal of the current state of the line to be evaluated is obtained, and the monitoring signal is evaluated according to the health value evaluation model to obtain the health evaluation data.

[0063] ​S5, based on the health assessment data, determine the resistance of the current working state of the to-be-evaluated line to each fault, obtain resistance data, and predict the fault risk of the line according to the resistance data, obtain the risk level of the fault risk, and output and display the risk level of each fault risk;

[0064] S6, based on the time data, statistically analyze the monitoring signal and the health assessment data of the to-be-evaluated line, obtain a risk prediction curve, determine the change rate of each index item in the risk prediction curve, and predict the occurrence probability of the corresponding fault risk based on the change rate of each index item, obtain the risk probability, and output and display the risk probability of each fault risk.

[0065] In this embodiment, by using big data statistics, a line specification table is constructed, which simplifies the input requirements of the to-be-evaluated line, determines the initial state of the to-be-evaluated line, and ensures the applicability of the system to the to-be-evaluated line in different states. By statistically analyzing the historical fault data of the to-be-evaluated line and taking the initial ideal data as the ideal state of the to-be-evaluated line, a health value evaluation model for evaluating the health state of the line relative to each fault is constructed, and then the health value of the to-be-evaluated line relative to each fault category can be evaluated according to the current monitoring signal of the to-be-evaluated line, so that the evaluated health value is more targeted, and the accuracy of the evaluation result is improved. At the same time, by comparing the current state of the to-be-evaluated line with the fault condition when the fault occurs, the difficulty of the to-be-evaluated line to occur the corresponding fault is determined, the judgment accuracy of the resistance data is improved, and at the same time, the data change of the to-be-evaluated line is determined by statistically analyzing the monitoring signal of the to-be-evaluated line, and then the probability of the to-be-evaluated line to occur the fault under the data change is determined, so as to achieve the prediction effect of the fault risk, and the prediction result is more accurate. The state evaluation of the line as a whole is converted to the state evaluation of the line relative to each fault, so that the evaluation result is more targeted, and the accuracy of the state evaluation result is improved.

[0066] For example, when the state of the line is evaluated and predicted, the ideal state of the to-be-evaluated line is determined first, so that the ideal state is taken as the initial reference value, and the evaluation result is more accurate and effective.

[0067] After determining the ideal state of the line to be evaluated, the historical fault data of the line to be evaluated is analyzed to determine the data relationship between the data at the time of fault and the data in the ideal state, and then the data relationship is taken as the basis for evaluating the health value of the line, for example, the line data in the ideal state is 100, the line data at the time of fault is 50, the line health value in the ideal state is 100, and the line health value at the time of fault is 0, that is, the closer the line data of the line to the line data at the time of fault, the closer the corresponding line health value to 0, and then a simple health value evaluation equation is constructed: y=2(x-50), where y is the health value and x is the current line data.

[0068] Since the line data corresponding to different faults is different, for example, the line data required for fault A is that index item a is less than 50 and index item b is less than 50, the line data required for fault B is that index item c is less than 50 and index item d is less than 50, and when fault A occurs, fault B does not necessarily occur, but the health value of the line changes, so the health value evaluation model constructed is only used for single judgment of each index item in one fault, not the overall judgment result.

[0069] Therefore, after constructing the health value evaluation model of the line to be evaluated, the health value of each index item with respect to each fault is obtained by judging the current monitoring signal of the line through the health value evaluation model, so as to evaluate the health value of the line in each fault, and the difficulty of the condition required to be achieved by the line to be evaluated from the current state to the fault state is determined according to the size of the health value, so as to determine the resistance data of the line to the fault. For example, when the health value of the line is 100, if A fault occurs, the line needs to withstand a tension of 100, which is impossible without human intervention, so it can be determined that the resistance of the line to fault A is very high, and the risk level of the corresponding fault risk is safe (assuming that the risk level includes safe, general, dangerous, and high risk).

[0070] At the same time, the monitoring signal of the line to be evaluated is counted according to the time data to determine the change rate of the index data of each index item of the line to be evaluated, and then the time data required for each fault to occur is predicted based on the historical fault data at the time of fault and the current change rate, and the corresponding risk probability is obtained.

[0071] Finally, the health value, risk level and risk probability of the line to be evaluated with respect to each fault category are output and displayed on the central display for relevant staff to view.

[0072] In step S3, the historical fault data of the line to be evaluated is obtained, and the health value evaluation model of the line is constructed in combination with the initial ideal data, including the following steps:

[0073] S31, based on the time point when the historical fault occurs, marking the historical fault data before the fault occurs as first fault data, and marking the historical fault data when the fault occurs as second fault data;

[0074] S32, performing index data matching on the first fault data and the second fault data, determining the index item that changes when the fault occurs compared to the index data before the fault occurs in the same historical fault event, and marking the index item as a key index item;

[0075] S33, counting the key index items of multiple historical fault events of the same type of fault to determine a fault evaluation index;

[0076] S34, counting the index data of the fault evaluation index items of multiple fault events of the same type of fault to determine fault judgment data;

[0077] S35, taking the initial ideal data as the line data when the health value is the first health threshold, taking the fault judgment data as the line data when the health value is the second health threshold, constructing the relationship between the line data and the health value, and obtaining a health value evaluation model of each fault evaluation index. The first health threshold refers to the value set to represent when the health degree of the line is 100% healthy, which can be 100. The second health threshold refers to the value set to represent when the line fails, which can be 0.

[0078] In this embodiment, by extracting the index item that changes when the fault occurs, the key index item and its corresponding index data that need to be triggered when the fault occurs are reduced, the amount of data that needs to be analyzed for fault risk is reduced, the data processing efficiency is improved, and the data quality is improved. By counting the key index items in the same type of fault, the selected fault evaluation index can more accurately reflect the fault. Similarly, by counting the index data of the fault evaluation index of the same type of fault, the fault judgment data obtained can more accurately reflect whether the fault occurs, and the evaluation result of the health value evaluation model of the line constructed according to the fault judgment data of the corresponding fault evaluation index and the initial ideal data is more accurate, thereby improving the accuracy of the line fault risk warning.

[0079] For example, during the operation of the to-be-evaluated line L, the monitored index items a, b, and c change, and in the fault event 1, the fault event 2, …, and the fault event n, only the index item a changes when the fault occurs, and thus the index item a is determined as the fault evaluation index of the fault category, that is, the last straw that broke the camel's back. Therefore, when monitoring the fault category, only the index item a needs to be monitored, and the index items b and c do not need to be monitored, and thus when judging the fault risk of the fault category, only the index data of the index item a needs to be evaluated to determine whether the data of the index item a is in the fault judgment data close to the response, thereby simplifying the evaluation process and improving the evaluation efficiency.

[0080] In step S33, the key index items of the plurality of historical fault events of the same fault category are counted to determine the fault evaluation index, including the following steps.

[0081] In S331, the key index items of the plurality of historical fault events of the same fault are counted to determine whether the distribution of each key index item meets the Pareto distribution; wherein the Pareto distribution is a data distribution method in the mathematical field.

[0082] In S332, if it is determined that the distribution of the key index items meets the Pareto distribution, the distribution of each key index item is used as the credibility of the result that the key index item can reflect the line health value, to obtain the credibility data, and the key index item located in the head of the Pareto distribution is marked as the fault evaluation index.

[0083] In S333, if it is determined that the distribution of the key index items does not meet the Pareto distribution, it is determined that each index item does not have credibility, and a fault unpredictable signal is output.

[0084] In this embodiment, the key index items in the plurality of fault events of the same fault category are subjected to mathematical statistical analysis, so as to determine the distribution of the key index items, and then the credibility of each key index item is determined according to the distribution, and the key index items are screened based on the credibility, so as to ensure that the screened key index items can accurately reflect the occurrence of the fault, and thus when the fault analysis and early warning are performed according to the screened key index items, the obtained result is more accurate and effective, and the accuracy of the fault risk prediction is improved.

[0085] For example, assuming that the fault evaluation index of A fault is a, and when A fault occurs, the change of index item a may lead to the change of index items b and c, or the change of index items c and d. Therefore, when determining the fault evaluation index of A fault, the changed index items involved in each fault event need to be counted to determine which index item appears most frequently, thereby determining the accuracy of the changed index reflecting the occurrence of the fault. Assuming that the change of index item a occurs every time A fault occurs, the numerical change of index item a reflects the possibility of occurrence of A fault, thereby ensuring the accuracy of the fault risk predicted by analyzing the data of index item a.

[0086] In step S35, the initial ideal data is taken as the line data when the health value is the first health threshold, the fault judgment data is taken as the line data when the health value is the second health threshold, the relationship between the line data and the health value is constructed, and the health value evaluation model of each fault evaluation index is obtained, including the following steps:

[0087] S351, the initial ideal data is taken as the line data when the health value is the first health threshold, the fault judgment data is taken as the line data when the health value is the second health threshold, the relationship between the line data and the health value is constructed, and the initial health value evaluation model is obtained;

[0088] S352, the number of fault evaluation indexes of the same type of fault is determined, and the index property of the fault evaluation index of the same type of fault is determined;

[0089] S353, if the number of fault evaluation indexes is equal to 1, the initial health value evaluation model is taken as the health value evaluation model of the corresponding fault;

[0090] S354, if the number of fault evaluation indexes is greater than 1, the fault judgment data of multiple fault evaluation indexes is compared with each other to determine whether the data relationship between the multiple fault evaluation indexes is the same;

[0091] S355, if the data relationship between the multiple fault evaluation indexes is the same, it is determined that the index property of the fault evaluation index is single, and the initial health value evaluation model is taken as the health value evaluation model of the corresponding fault evaluation index;

[0092] S356, if the data relationship between the multiple fault evaluation indexes is different, each fault evaluation index in the same fault event is evaluated based on the initial health value evaluation model to determine the health value of each fault evaluation index in the fault event, and an event health value set is obtained;

[0093] S357, performing a second health threshold judgment on the event health value set to determine whether the health value of at least one fault evaluation index in each fault event reaches the second health threshold;

[0094] S358, if the health value of at least one fault evaluation index in each fault event reaches the second health threshold, determining that the index property of the fault evaluation index of the fault class is diversity, and taking the initial health value evaluation model as the health value evaluation model of each fault evaluation index in the corresponding fault;

[0095] S359, if the health value of all fault evaluation indexes in the fault event is not the second health threshold, constructing a data relationship of the index data of the key index item in the plurality of fault events, determining comprehensive data of the fault class according to the constructed data relationship, constructing a relationship between the line data and the health value based on the comprehensive data and the initial ideal data, obtaining a second health value evaluation model, and taking the second health value evaluation model as the health value evaluation model in the fault event.

[0096] In the embodiment, when constructing the health value evaluation model under different faults, the initial health value evaluation model is constructed by using the initial ideal data and the fault judgment data of each fault evaluation index, and the initial health value evaluation model is revised for different fault classes. The initial health value evaluation model is adjusted according to the number judgment, data relationship judgment and event health value judgment of the fault evaluation index of the fault, so that the data evaluated by the adjusted health value evaluation model is more accurate, the accuracy of the evaluation result is improved, and the effectiveness of the evaluation result of the health value evaluation model is ensured.

[0097] For example, since the health value evaluation model is used to evaluate the health value of the current line, a health value will also be output when the current line is in a normal state, for example, the ideal state of the line when not in use. At this time, the health value is 100 (the health value 100 refers to the ideal state for any kind of fault). When the line is used, the state of the line changes due to aging and other reasons, and the health value may be 90 relative to A fault and 80 relative to B fault, and the corresponding health value needs to be output.

[0098] When A fault is evaluated and B fault is evaluated in the line state, the fault evaluation indexes that need to be evaluated are different, so A fault needs to be evaluated separately and B fault needs to be evaluated separately.

[0099] If the number of failure evaluation indexes of A failure is 1 (for example, index item a), only index item a needs to be evaluated to determine the data relationship between the current index data of index item a and the failure judgment data, and then determine the corresponding health value according to the data relationship. For example, the index data of index item a in the ideal state is 100, the failure judgment data is 50, and the relationship between the health value and the index data and the evaluation data is y = 100 / 50 x (x-50), where y is the current health value of index item a, and x is the current index data of index item a. When the current index data is 70, the evaluated health value is 40.

[0100] If the number of failure evaluation indexes of B failure is 2 (for example, index items b and c), not only index item b but also index item c needs to be evaluated when the line health value is evaluated, and before evaluating index item b and index item c, the data relationship between index item b and index item c on the failure judgment data can be determined to determine whether separate evaluation is needed.

[0101] Assuming that in all historical failure events, the index data of index item b at failure is 50, and the index data of index item c at failure is 30, it indicates that the data relationship between index item c and index item b is always consistent, and therefore it can be determined that when the index data of index item b reaches 50, the index data of index item c will necessarily reach 30. Therefore, when evaluating the health value of B failure in this case, only index item b or index item c needs to be evaluated.

[0102] If the failure judgment data of index item b in the historical failure events is 50, and the failure judgment data of index item c can be 10, 30 or 20, then index item c and index item b do not have a failure data relationship, and therefore index item b and index item c need to be considered comprehensively when judging B failure.

[0103] Assuming that indicator item b alone can cause the line to fail B, and indicator item c alone can also cause the line to fail, and assuming that the failure evaluation data of indicator item b is 50 and the failure evaluation data of indicator item c is 30, when the line fails B, at least one indicator item reaches the failure evaluation data, that is, the indicator data of indicator item b is 50 or the indicator data of indicator item c is 30 or the indicator data of indicator item b is 50 and the indicator data of indicator item c is 30. Similarly, when indicator item b and indicator item c reach the corresponding failure evaluation data, B failure will occur. Therefore, in the process of comprehensively considering indicator item b and indicator item c, by separately evaluating the health value of indicator item b and indicator item c, it is determined whether all historical failure events satisfy the above rule. If all historical failure events satisfy the above rule, only one health value evaluation of each indicator item is needed, and the minimum health value is taken as the reference item of the failure evaluation.

[0104] If all historical failure events do not basically satisfy the above rule, for example, 3 events satisfy the rule of indicator item b, 3 events satisfy the rule of indicator item c, and 4 events do not satisfy the above rule, it is determined that all historical failure events do not basically satisfy the above rule. When the above rule is not satisfied, since the final data direction is consistent (B failure), it can be known by reverse deduction of the final result direction that there must be a certain data relationship between the key indicator items involved in B failure. By using artificial intelligence to summarize a large amount of data, the data relationship is determined to obtain a unique comprehensive data, and the comprehensive data is taken as the basis for failure discrimination to construct a second health value evaluation model that meets the situation.

[0105] In step S5, based on the health evaluation data, the resistance of the current working state of the line to be evaluated to each failure is determined to obtain resistance data, and the failure risk of the line is predicted according to the resistance data to obtain the risk level of the failure risk, including the following steps:

[0106] S51, obtaining the failure evaluation indicators of the line to be evaluated under each failure category, and determining the external factors that can affect the failure evaluation indicators according to the failure evaluation indicators;

[0107] S52, based on the health evaluation data of the line to be evaluated in the current state, determining the condition required for the external factors to reduce the health value of the line to be evaluated from the current health value to the second health threshold, and marking the condition as first failure condition data;

[0108] S53, performing implementability judgment on the first failure condition data to determine the probability of reaching the first failure condition data under non-human conditions to obtain failure resistance data;

[0109] S54, match the fault risk level table of the line based on the fault resistance data, and obtain the risk level of the line to be evaluated relative to each fault category. The risk level can be safe, general, dangerous, and high-risk.

[0110] In this embodiment, since the resistance judgment on the line refers to the resistance to suddenness data, i.e., the possibility of the external environment suddenly reaching the condition, by determining the external influencing factors corresponding to each fault evaluation index under each fault category, and combining the current state of the line to be evaluated to determine the condition required to reduce the health value of the fault evaluation index to 0 under the corresponding external influencing factor, and taking the condition as the first fault condition data, and finally determining the probability that can be achieved under non-human conditions through the implementability judgment of the first fault condition, the resistance of the current line to suddenness is determined, which makes the evaluation more accurate and improves the accuracy of the risk assessment of each fault risk.

[0111] For example, assuming that the fault evaluation index of the line to be evaluated under A fault is tension, and the factors in the external environment that can cause tension to the line to be evaluated are wind, snow, etc. Assuming that the health value of the current fault index is 70, and the condition required to reduce the health value of the fault index with a health value of 70 to a health value of 0 is wind of 12 levels, snow of 1 m, and in the actual process, the possibility of the wind in this area reaching 12 levels is 0%, and the possibility of the snow reaching 1 m is 0%, it is determined that the resistance of the current state is 100%, and the risk level corresponding to the resistance of 100% is safe. If the possibility of the wind in this area reaching 12 levels is 50%, and the possibility of the snow reaching 1 m is 70%, it is determined that the resistance of the current state is 30%, and the risk level corresponding to the resistance of 30% is dangerous.

[0112] In step S6, the monitoring signals and health evaluation data of the line to be evaluated are statistically analyzed based on the time data to obtain a risk prediction curve, determine the change rate of each index item in the risk prediction curve, and predict the occurrence probability of the corresponding fault risk based on the change rate of each index item to obtain a risk probability, including the following steps:

[0113] S61, based on the fault category, index data of the corresponding fault evaluation index in the monitoring signal of the line to be evaluated is statistically analyzed, and a risk prediction curve of the corresponding fault evaluation index is drawn;

[0114] S62, change rate judgment is performed on the risk prediction curve to obtain prediction change data;

[0115] S63, difference calculation is performed on the current index data of the fault evaluation index and the fault judgment data of the corresponding index to obtain a predicted change value;

[0116] S64, based on the predicted change data and the predicted change value, determine the predicted change duration, and compare the predicted change durations of each fault evaluation index in the same fault according to the fault category, match the shortest predicted change duration with the built-in risk probability table, determine the risk probability of the corresponding fault category and output display. Wherein, the risk probability table is the result obtained by big data analysis on the historical fault events of this type of fault based on artificial intelligence.

[0117] In this embodiment, by performing index data statistics on the fault evaluation indexes that need to be monitored, the relationship between the state of the fault evaluation index and the time is determined, and then the remaining predicted change value is judged according to the change relationship of the data, so that the predicted change duration of each fault evaluation index is determined, and the shortest predicted change duration is matched with the risk probability table as the judgment standard of this type of fault, to determine the risk probability of the corresponding fault of the line under the current state, and improve the credibility of the power line fault risk warning.

[0118] For example, assuming that the data change curve of the fault evaluation index a in the line to be evaluated is a linear curve, y=100-x, where y is the current state value and x is the time data. It can be known that the state value of the line decreases by 1 unit every 1 unit of time. If the fault judgment data of the fault evaluation index a is 50 and the current state value is 70, it is determined that the predicted change duration is 20, and thus the probability of fault risk occurrence is determined according to the predicted change duration 20. For example, by big data matching, a corresponding relationship table of predicted duration and risk occurrence probability is obtained, so that the corresponding risk probability is matched according to the predicted change duration, improving the accuracy and comprehensiveness of risk prediction.

[0119] In step S4, the monitoring signal of the current state of the line to be evaluated is obtained, and the monitoring signal is evaluated according to the health value evaluation model to obtain health evaluation data, including the following steps:

[0120] S41, obtaining the fault evaluation indexes of each fault category, and determining the index data of the monitoring signal of the line to be evaluated on the corresponding fault evaluation index based on the fault evaluation index;

[0121] S42, inputting the index data of each fault evaluation index into the health value evaluation model to obtain the health evaluation value of the fault evaluation index;

[0122] S43, according to the fault category, the health evaluation values of the fault evaluation indexes of the same fault are statistically calculated to obtain the health evaluation data of the corresponding fault.

[0123] S44, if the number of fault evaluation indexes of the fault is 1, the health evaluation value of the fault evaluation index is taken as the health evaluation data of the fault;

[0124] S45, if the number of fault evaluation indexes of the fault is greater than 1, obtaining an index property of the fault evaluation indexes of the fault, if the index property is singularity, taking the mean value of the health evaluation values in the multiple fault evaluation indexes of the fault as the health evaluation data of the fault;

[0125] S46, if the index property is diversity, taking the minimum health evaluation value of the health evaluation values in the multiple fault evaluation indexes of the fault as the health evaluation data of the fault;

[0126] S47, if the index property is regularity, inputting the comprehensive data of the fault evaluation indexes under the fault into the second health value evaluation model for evaluation to obtain the health evaluation data of the to-be-evaluated line relative to the fault.

[0127] For example, after the corresponding health value evaluation model is constructed, since the health value evaluated by the health value evaluation model is the health value of the corresponding fault evaluation index, it is necessary to determine the health value of the to-be-evaluated line relative to each fault according to the health value of each fault evaluation index. The overall line state analysis is divided into state analysis of each fault.

[0128] By determining different fault categories, the output results of the health value evaluation models corresponding to the fault categories are used to judge the line relative to each fault. When the number of fault evaluation indexes of the fault is 1, it indicates that the fault evaluation index can reflect the health value of the line state relative to the fault, and therefore the health value of the fault evaluation index is directly taken as the health value of the line relative to the fault. When the number of fault evaluation indexes of the fault is greater than 1 and the index property is singularity, since the data of the fault evaluation indexes varies and the fault evaluation data varies, the health values evaluated by each fault evaluation index are different, and since the fault evaluation indexes are associated, the mean value of the fault evaluation indexes is calculated to make the comprehensive evaluation result more reliable.

[0129] When the index property is diversity, it indicates that as long as one of the health values of the multiple fault evaluation indexes is zero, the fault occurs, and therefore the minimum health evaluation value of the health evaluation values in the multiple fault evaluation indexes is taken as the health evaluation data of the fault.

[0130] When the index property is regularity, since the health value evaluation model evaluates the comprehensive data under the fault, the comprehensive evaluation result can be taken as the health value of the line relative to the fault, thereby ensuring the accuracy of the data.

[0131] Based on the description of the above-mentioned power line state evaluation method based on artificial intelligence, the embodiment of the application further discloses a power line state evaluation system based on artificial intelligence:

[0132] As Fig. 2 shown, an artificial intelligence-based power line state evaluation system, by applying the artificial intelligence-based power line state evaluation method as described above, includes an initial state analysis module 1, an evaluation model construction module 2, and a line state evaluation and display module 3.

[0133] The initial state analysis module 1 is used to obtain the line grade of the line to be evaluated and match the line grade of the line to be evaluated with the built-in line specification table to determine the line specification under the corresponding line grade; obtain the input signal of the line endpoint, and simulate the input signal of the line endpoint based on the line specification of the line to be evaluated to determine the performance of the input signal under the line specification, and obtain the initial ideal data.

[0134] The evaluation model construction module 2 is used to obtain the historical fault data of the line to be evaluated, and construct a health value evaluation model of the line in combination with the initial ideal data.

[0135] The line state evaluation and display module 3 is used to obtain the monitoring signal of the current state of the line to be evaluated, and evaluate the monitoring signal according to the health value evaluation model to obtain health evaluation data; based on the health evaluation data, determine the resistance of the current working state of the line to be evaluated to each fault to obtain resistance data, and predict the fault risk of the line according to the resistance data to obtain the risk level of the fault risk, and output and display the risk level of each fault risk; based on the time data, the monitoring signal of the line to be evaluated and its health evaluation data are counted to obtain a risk prediction curve, the change rate of each index item in the risk prediction curve is determined, and the occurrence probability of the corresponding fault risk is predicted based on the change rate of each index item to obtain a risk probability, and the risk probability of each fault risk is output and displayed.

[0136] Compared with the existing artificial intelligence-based power line state evaluation method and system, the accuracy of line state evaluation is improved.

[0137] The above are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, therefore: any equivalent changes made on the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A power line condition assessment method based on artificial intelligence, characterized in that, include: Step S1: Obtain the line level of the line to be evaluated and match the line level of the line to be evaluated with the built-in line specification table to determine the line specification under the corresponding line level. Step S2: Obtain the input signal at the line endpoint and simulate the input signal at the line endpoint based on the line specification of the line to be evaluated to determine the performance of the input signal under the line specification and obtain the initial ideal data. Step S3: Obtain historical fault data of the line to be evaluated, and construct a health value evaluation model of the line by combining it with the initial ideal data; Step S4: Obtain the monitoring signal of the current status of the line to be evaluated, and evaluate the monitoring signal according to the health value evaluation model to obtain health evaluation data; Step S5: Based on the health assessment data, determine the resistance of the current working state of the line to be assessed to each fault, obtain resistance data, predict the fault risk of the line based on the resistance data, obtain the risk level of the fault risk, and output and display the risk level of each fault risk. Step S6: Based on the time data, the monitoring signals and health assessment data of the line to be evaluated are statistically analyzed to obtain the risk prediction curve, the change rate of each indicator item in the risk prediction curve is determined, and the probability of occurrence of the corresponding fault risk is predicted based on the change rate of each indicator item to obtain the risk probability, and the risk probability of each fault risk is output and displayed. Based on the fault category, the corresponding fault assessment indicators in the monitoring signals of the line to be assessed are statistically analyzed, and the risk prediction curves of the corresponding fault assessment indicators are plotted. The rate of change of the risk prediction curve is determined to obtain the predicted change data; The difference between the current data of the fault assessment indicators and the fault judgment data of the corresponding indicators is calculated to obtain the expected change value. Based on the predicted change data and expected change values, the expected change duration is determined. Then, according to the fault category, the expected change durations of various fault assessment indicators within the same fault are compared with each other. The expected change duration with the shortest expected change duration is matched with the built-in risk probability table to determine the risk probability of the corresponding fault category and output it for display.

2. The power line condition assessment method based on artificial intelligence according to claim 1, characterized in that: Step S3 includes: S31, based on the time point when the historical fault occurred, mark the historical fault data before the fault occurred as the first fault data, and mark the historical fault data when the fault occurred as the second fault data; S32, Match the first fault data and the second fault data with the indicator data, identify the indicator items in the same historical fault event that have changed from the indicator data before the fault occurred, and mark the indicator items as key indicator items. S33, statistically analyze the key indicators of multiple historical fault events of the same type of fault to determine the fault assessment indicators; S34, statistically analyze the indicator data of multiple fault events of the same type of fault to determine the fault judgment data; S35, using the initial ideal data as the line data when the health value is the first health threshold, and the fault judgment data as the line data when the health value is the second health threshold, the relationship between the line data and the health value is constructed to obtain the health value assessment model for each fault assessment index.

3. The power line condition assessment method based on artificial intelligence according to claim 2, characterized in that: Step S33 includes: The key indicators of multiple historical fault events of the same fault are statistically analyzed to determine whether the distribution of each key indicator meets the Pareto distribution. If the distribution of key indicators is determined to satisfy the Pareto distribution, then the distribution ratio of each key indicator is used as the credibility of the result that the corresponding key indicator can reflect the line health value, and credibility data is obtained. The key indicator at the head of the Pareto distribution is marked as the fault assessment indicator. If the distribution of key indicators does not conform to the Pareto distribution, then each indicator is deemed unreliable, and an unpredictable fault signal is output.

4. The power line condition assessment method based on artificial intelligence according to claim 2, characterized in that: Step S35 includes: The initial ideal data is used as the line data when the health value is the first health threshold, and the fault judgment data is used as the line data when the health value is the second health threshold. The relationship between the line data and the health value is constructed to obtain the initial health value evaluation model. Quantitatively assess the fault assessment indicators for the same type of fault to determine the nature of the indicators. If the number of fault assessment indicators is equal to 1, then the initial health value assessment model will be used as the health value assessment model for the corresponding fault. If the number of fault assessment indicators is greater than 1, the fault judgment data of multiple fault assessment indicators will be compared with each other to determine whether the data relationship between multiple fault assessment indicators is the same. If the data relationships between multiple fault assessment indicators are the same, the indicator nature of the fault assessment indicator is determined to be singular, and the initial health value assessment model is used as the health value assessment model of the corresponding fault assessment indicator. If the data relationships between multiple fault assessment indicators are different, then each fault assessment indicator in the same fault event is evaluated based on the initial health value assessment model to determine the health value of each fault assessment indicator in the fault event and obtain the event health value set. A second health threshold is determined for the event health value set to determine whether there is at least one fault assessment indicator whose health value reaches the second health threshold in each fault event. If at least one fault assessment indicator in each fault event has a health value that reaches the second health threshold, then the fault assessment indicator of this type of fault is determined to be of diverse nature, and the initial health value assessment model is used as the health value assessment model for each fault assessment indicator in the corresponding fault. If, in a fault event, the health values ​​of all fault assessment indicators are not at the second health threshold, then data relationships are constructed for the key indicator data in multiple fault events, and comprehensive data for this fault category is determined based on the constructed data relationships. Based on the comprehensive data and the initial ideal data, the relationship between line data and health values ​​is constructed to obtain the second health value assessment model, and the second health value assessment model is used as the health value assessment model for this type of fault event.

5. The power line condition assessment method based on artificial intelligence according to claim 2, characterized in that: Step S5 includes: Obtain the fault assessment indicators for the line to be evaluated under each fault category, and determine the external factors that can affect the fault assessment indicators based on the fault assessment indicators. Based on the health assessment data of the line under evaluation in its current state, determine the conditions that external factors need to be met to reduce the health value of the line under evaluation from the current health value to the second health threshold, and mark this condition as the first fault condition data. The feasibility of the first fault condition data is assessed to determine the probability of reaching the first fault condition data under non-human conditions, thereby obtaining fault resistance data. The fault risk level table of the line is matched based on the fault resistance data to obtain the risk level of the line to be evaluated relative to each fault category.

6. The power line condition assessment method based on artificial intelligence according to claim 4, characterized in that: Step S4 includes: Obtain the fault assessment indicators for each fault category, and determine the indicator data of the monitoring signal of the line to be assessed on the corresponding fault assessment indicators based on the fault assessment indicators. Input the index data of each fault assessment indicator into the health value assessment model to obtain the health assessment value of the fault assessment indicator. Based on the fault category, the health assessment values ​​of the fault assessment indicators for the same type of fault are statistically calculated to obtain the corresponding fault health assessment data. If the number of fault assessment indicators for this fault is 1, then the health assessment value of this fault assessment indicator shall be used as the health assessment data for this fault. If the number of fault assessment indicators for the fault is greater than 1, then the indicator nature of the fault assessment indicator is obtained. If the indicator nature is singular, then the average of the health assessment values ​​among the multiple fault assessment indicators for the fault is used as the health assessment data for the fault. If the indicator is diverse, then the health assessment value with the smallest health assessment value among the multiple fault assessment indicators of the fault shall be used as the health assessment data of the fault. If the indicator is regular, the comprehensive data of the fault assessment indicators under the fault is input into the second health value assessment model for assessment, and the health assessment data of the line to be assessed relative to the fault is obtained.

7. A power line condition assessment system based on artificial intelligence, characterized in that, The system is used to implement an artificial intelligence-based power line condition assessment method according to any one of claims 1-6, comprising: an initial condition analysis module, an assessment model construction module, and a line condition assessment and display module; The initial state analysis module is used to obtain the line level of the line to be evaluated and match the line level of the line to be evaluated with the built-in line specification table to determine the line specification under the corresponding line level; obtain the input signal of the line endpoint and simulate the input signal of the line endpoint based on the line specification of the line to be evaluated to determine the performance of the input signal under the line specification and obtain the initial ideal data. The evaluation model building module is used to obtain historical fault data of the line to be evaluated and combine it with the initial ideal data to build a health value evaluation model of the line. The line status assessment and display module is used to acquire monitoring signals of the current status of the line under assessment, evaluate the monitoring signals according to the health value assessment model, and obtain health assessment data. Based on the health assessment data, it determines the resistance of the current working state of the line under assessment to various faults, obtains resistance data, and predicts the fault risk of the line based on the resistance data, obtaining the risk level of the fault risk, and outputs and displays the risk level of each fault risk. Based on time data, it statistically analyzes the monitoring signals and health assessment data of the line under assessment to obtain a risk prediction curve, determines the rate of change of each indicator in the risk prediction curve, and predicts the probability of occurrence of the corresponding fault risk based on the rate of change of each indicator, obtaining the risk probability, and displays each... The system outputs and displays the probability of fault risk; based on the fault category, it statistically analyzes the corresponding fault assessment indicators in the monitoring signals of the line to be assessed, and plots the risk prediction curves for the corresponding fault assessment indicators; it judges the rate of change of the risk prediction curves to obtain the predicted change data; it calculates the difference between the current indicator data of the fault assessment indicators and the fault judgment data of the corresponding indicators to obtain the expected change value; based on the predicted change data and the expected change value, it determines the expected change duration, and compares the expected change durations of various fault assessment indicators within the same fault according to the fault category. It then matches the expected change duration with the shortest expected change duration with the built-in risk probability table to determine the risk probability of the corresponding fault category and outputs and displays it.

Images