Power transmission line comprehensive risk early warning method considering conductor deicing jump

Abstract

The invention discloses a power transmission line comprehensive risk early warning method considering conductor deicing jump, comprising the following steps: acquiring meteorological data and icing thickness data when a power transmission line has a historical icing fault to form a historical fault data set, and constructing an icing rate prediction model based on data driving; acquiring online monitoring and meteorological numerical forecasting information of a target area, preprocessing the information, inputting the information into the icing rate prediction model, performing criterion judgment on an output icing rate value and corresponding meteorological information, and predicting a deicing moment and a deicing duration; fuzzy judgment is carried out to output probability density distribution of the deicing rate; correcting the actual allowable tensile strength of the line; calculating a flashover probability and a disconnection probability, further calculating a comprehensive fault probability of the line, comparing the comprehensive fault probability with a preset risk grade division standard, and outputting a comprehensive risk early warning grade of the deicing jump event; the technical problems of universality, probabilistic evaluation, line strength correction and evaluation accuracy in the prior art are solved.

Description

Technical Field

[0001] This invention belongs to the field of power transmission line risk early warning technology, and in particular relates to a comprehensive risk early warning method for transmission lines that takes into account conductor de-icing and jumping. Background Technology

[0002] Severe icing of transmission lines can jeopardize the safe and stable operation of the power grid. Icing can induce structural changes and electrical failures in transmission lines, leading to faults such as galloping, ice flashover, ice shedding jumps, or line breaks. Although the frequency of line breaks has decreased with the continuous improvement of the anti-icing capabilities of transmission lines, the risks of flashovers and exceeding instantaneous stress limits caused by ice shedding jumps remain key issues for power grid prevention. According to existing statistics, a large number of icing faults are caused by electrical failures due to ice shedding jumps. The accuracy of risk assessment for such events is closely related to the accurate prediction of ice thickness and the probabilistic description of ice shedding loads.

[0003] Traditional methods for monitoring and predicting icing thickness rely heavily on empirical formulas or physical models. However, existing physical models are limited to single meteorological parameters, lacking universality and adaptability. Furthermore, current risk assessment models fail to accurately predict the de-icing rate during de-icing jump events, typically employing a 100% de-icing assumption, resulting in inaccurate risk assessments. Additionally, when assessing line breakage risk, current technologies generally use the line's design strength, neglecting the degradation of actual allowable tensile strength due to fatigue accumulation during long-term operation. This leads to discrepancies between the assessed probability of line breakage and the actual operating condition of the line. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a comprehensive risk early warning method for transmission lines that takes into account conductor de-icing jumps, so as to accurately assess and warn of the comprehensive fault probability of de-icing jump events, and solve the technical problems of existing methods in terms of universality, probabilistic assessment, line strength correction and assessment accuracy.

[0005] Technical solution of the present invention:

[0006] A comprehensive risk early warning method for transmission lines that takes into account conductor de-icing and jumping includes:

[0007] Step 1: Obtain meteorological data and ice thickness data of historical icing faults of transmission lines to form a historical fault dataset, and construct a data-driven icing rate prediction model.

[0008] Step 2: Obtain online monitoring and meteorological numerical forecast information for the target area. Input the preprocessed information into the icing rate prediction model, and make judgments on the output icing rate value and corresponding meteorological information to predict the icing time and duration.

[0009] Step 3, fuzzy judgment is performed on the meteorological information corresponding to the de-icing duration period and the ice accretion rate value to output the probability density distribution of the de-icing rate;

[0010] Step 4, the target area line digital archives and meteorological statistical data are acquired to correct the actual allowable tensile strength of the line;

[0011] Step 5, the line de-icing rate is acquired based on Monte Carlo sampling, the flashover probability and the line breakage probability are calculated, and then the comprehensive failure probability of the line is calculated;

[0012] Step 6, the comprehensive failure probability is compared with the preset risk level division standard, and the comprehensive risk warning level of the de-icing jump event is output.

[0013] The method for acquiring the historical failure data set comprises the following steps:

[0014] The historical ice thickness records are acquired through the power transmission line monitoring terminal, including the ice thickness value and the reporting time; the meteorological data are acquired through the meteorological sensor of the power transmission line monitoring terminal, including the temperature, the humidity, the wind speed and the reporting time;

[0015] The ice thickness information and the meteorological information are subjected to data cleaning and interpolation processing, the interpolation is performed by using the front and rear weighted mean value method, and the time scale is corresponded to generate the ice thickness and meteorological time sequence information containing the same time interval; the difference value of the ice thickness information of adjacent times is subjected to time differentiation processing, the meteorological information of adjacent times is averaged, and then the maximum-minimum normalization method is used for normalization processing to form the historical failure data set.

[0016] The ice accretion rate prediction model structure based on data driving comprises a historical encoder, a reparameterization layer and a prediction decoder;

[0017] The historical encoder comprises 2 layers of LSTM, a full connection layer, a full connection mean value layer and a full connection variance layer; the number of neurons of the first layer of LSTM in the historical encoder is 128, which is used for learning the dependence relationship of the time sequence, the number of neurons of the second layer of LSTM is 64, the number of neurons of the full connection layer is 32, and ReLU is used as the activation function; the full connection mean value layer is used for calculating the mean vector of the latent space distribution, and the number of neurons is 16; the full connection variance layer is used for calculating the logarithmic variance vector of the latent space distribution, and the number of neurons is 16;

[0018] The reparameterization layer comprises a sampling layer and a full connection layer; the reparameterization layer first samples the derivable latent context vector z from the historical encoder, and uses the full connection layer to map the initial condition in the prediction decoder;

[0019] The prediction decoder comprises a 1-layer LSTM and a full connection layer; the number of neurons of the LSTM in the prediction decoder is 32, and the output is obtained through the full connection layer, linear activation is adopted, and the number of neurons is 1;

[0020] The known historical fault data set samples are divided into a training set and a test set;

[0021] The loss function combines the accuracy of sequence prediction and the regularization of latent space, wherein the accuracy of sequence prediction adopts mean square error, the regularization of latent space adopts KL divergence, and the total loss function is:

[0022] ;

[0023] In the formula, is a prediction loss, is a KL divergence, wherein the hyperparameter The initial weight can be selected as 0.5, and can be dynamically adjusted according to the performance of the test set;

[0024] During training, the learning rate preheating and weight decay regularization strategy are introduced, the adaptive moment estimation Adam optimizer optimization algorithm is used to update the parameters of the network, and the joint training is ended until the loss function tends to be stable or reaches the preset maximum training number.

[0025] The icing rate prediction model is obtained by evaluating and fine-tuning based on the test set.

[0026] The method for predicting the ice shedding time and the ice shedding duration period comprises:

[0027] The meteorological data and the icing thickness data are preprocessed, including time scale conversion, data interpolation processing, average value and differential calculation and normalization processing;

[0028] The preprocessed data are input into the icing rate prediction model, and an icing rate time sequence in a future time period is output;

[0029] It is judged whether the output icing rate meets the state change criterion, and when the state change criterion is met, the thermodynamic ice shedding criterion is judged to predict the ice shedding time, which specifically comprises:

[0030] When the icing change rate appears negative growth and the value of the negative growth is less than a preset critical value, it is judged whether the corresponding predicted temperature value, solar radiation intensity and corresponding duration are greater than the preset critical value, wherein the state change criterion is:

[0031] ;

[0032] In the formula, is an icing rate value;

[0033] The thermodynamic ice shedding criterion is:

[0034] ;

[0035] wherein, is a temperature value, is a solar radiation intensity, is a duration satisfying the first two criteria in the thermodynamic ice shedding criterion;

[0036] When the state change criterion and the thermodynamic ice shedding criterion are satisfied simultaneously, the starting time is defined as the ice shedding time, the ice shedding duration is selected as 3h after the ice shedding time, the ice shedding rate before the ice shedding time is integrated, and the ice thickness at the ice shedding time is obtained by summing the ice thickness monitored by the online monitoring system.

[0037] The method for outputting the probability density distribution of the ice shedding rate comprises:

[0038] The average environmental temperature, the average solar radiation intensity and the average ice negative growth value in the ice shedding duration are taken as input variables, and the expected value and the standard deviation of the ice shedding rate are taken as output variables, and the membership degrees are divided; the triangular membership function is used to divide the membership degrees;

[0039] The parameters corresponding to the average environmental temperature, the average solar radiation intensity and the average ice negative growth value are combined and matched to construct the corresponding fuzzy control rule base;

[0040] According to the Mamdani model, the “max-min” reasoning method is used, the inference is carried out according to the rule base, and the membership degree value of each input index is substituted into the corresponding rule, so that the membership degrees of the fuzzy sets of the expected value and the standard deviation of the output about the ice shedding rate are inferred;

[0041] The centroid method is used to calculate the gravity center of the final output fuzzy set, the fuzzy sets of the expected value and the standard deviation of the ice shedding rate are de-fuzzied, the determined expected value and standard deviation are obtained, and the probability density distribution of the ice shedding rate is obtained.

[0042] The method for correcting the actual allowable tensile strength of the line comprises:

[0043] The S-N curve, the line operation time and the historical ice shedding jump fault report of the target line are obtained through the digital archives of the target line; the historical wind speed value of the target line is obtained through the meteorological bureau;

[0044] The residual strength expression of the line is obtained according to the S-N curve of the line;

[0045] The wind speed range is determined according to the maximum and minimum wind speed in the historical data, the wind speed range is divided into M discrete intervals, and M groups of discrete wind speed-probability pairs are calculated based on the historical data.

[0046] The AR linear filtering method and the minimum AIC criterion are used to obtain the random vibration wind speed spectrum under each representative wind speed, and the wind speed time history is obtained; the stress time history of the transmission line in the first 400s after the start of the truncated vibration is constructed by using the ANSYS finite element simulation;

[0047] The rain flow counting method is used to obtain the corresponding stress spectrum under each stress time history; based on the operation time of the line, the stress cycles and their corresponding frequencies during the operation time of the line are inferred, and the contribution value of each wind speed-probability pair to the degradation of the line resistance is calculated;

[0048] Based on the fault report of the target line history ice shedding jump, the dynamic stress time history is simulated according to the relevant information, and the contribution value of each ice shedding jump event to the degradation of the line resistance is obtained based on the same idea;

[0049] The total line resistance degradation contribution value is summed up, and the actual allowable tensile stress of the corresponding transmission line is calculated based on the residual strength expression of the element.

[0050] The method for calculating the comprehensive failure probability of the line comprises:

[0051] Based on the probability density distribution of the ice shedding rate, the single ice shedding rate is obtained by Monte Carlo sampling, and is substituted into the regression curve of the maximum ice jump height and the maximum instantaneous stress fitted by the simulation experimental data to obtain the maximum ice jump height and the maximum instantaneous stress under the ice shedding rate;

[0052] The minimum gap distance between the conductors after ice shedding is calculated The critical flashover voltage is calculated according to the minimum gap distance, compared with the maximum operating voltage, whether the sampling occurs flashover failure is judged, the above sampling and judgment process is repeated, and the flashover probability is calculated based on the judgment result;

[0053] Whether the sampling occurs line breakage failure is judged by comparing the maximum instantaneous stress with the actual allowable tensile stress, the above sampling and judgment process is repeated, and the line breakage probability is calculated based on the judgment result;

[0054] The comprehensive failure probability of the transmission line when the ice shedding jump occurs is calculated by the flashover probability and the line breakage probability.

[0055] The risk level division standard is:

[0056] When , blue warning is carried out;

[0057] When , yellow warning is carried out;

[0058] When , orange warning is carried out;

[0059] When , a red early warning is carried out;

[0060] According to the obtained line comprehensive fault probability, in combination with the risk grade division standard, the transmission line comprehensive risk early warning grade considering the conductor ice shedding jump is output.

[0061] The beneficial effects of the present application are:

[0062] For the prediction of ice thickness, the traditional method depends on empirical formula or physical model, is limited by single meteorological parameter, and has poor universality and adaptability. The present application constructs an ice thickness prediction model based on variational autoencoder and long short-term memory network, can fully capture the nonlinear relationship and long-term dependence of time series, more accurately predict the ice thickness, and thus improve the accuracy of risk early warning.

[0063] The existing evaluation model lacks correct prediction of the ice shedding rate in the ice shedding jump event, usually adopts the way of 100% ice shedding for evaluation, and thus leads to inaccurate risk evaluation result. The present application outputs the probability density distribution of the ice shedding rate through fuzzy judgment, more truly reflects the uncertainty of the ice shedding process, and makes the risk evaluation result more reliable.

[0064] In the evaluation of the line breaking risk, the traditional method mostly adopts the design strength of the line, and does not fully consider the degradation of the actual allowable tensile strength of the line caused by fatigue accumulation in long-term operation. The present application introduces a line actual allowable tensile strength correction technology based on fatigue accumulation theory, obtains the S-N curve, operation time and historical fault report of the line, and combines AR linear filtering and rain flow counting method to calculate the resistance degradation contribution value of the line, so as to obtain the actual allowable tensile stress, and make the evaluation result of the line breaking probability more consistent with the actual operation state of the line.

[0065] The present application combines the data-driven ice thickness prediction, the fuzzy reasoning-based ice shedding rate probabilistic processing and the fatigue accumulation theory-based line actual allowable tensile strength correction, can accurately evaluate and early warn the comprehensive fault probability of the ice shedding jump event, and solves the technical problems of the above-mentioned prior art method in universality, probabilistic evaluation, line strength correction and evaluation accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0066] Figure 1 The present application is a flowchart;

[0067] Figure 2 The present application is a schematic diagram for obtaining the probability density distribution of the ice shedding rate based on fuzzy theory;

[0068] Figure 3 The present application is a step diagram for calculating the resistance degradation contribution value;

[0069] Figure 4Fitting curves of the maximum ice jump height of the present application under typical line parameters, different ice thickness and ice shedding rate;

[0070] Figure 5 Fitting curves of the maximum instantaneous stress of the present application under typical line parameters, different ice thickness and ice shedding rate. DETAILED DESCRIPTION

[0071] A power transmission line comprehensive risk early warning method considering conductor ice shedding jump, comprising:

[0072] S1, obtaining meteorological data and ice thickness data at the time of historical icing failure of the power transmission line to form a historical failure data set, and constructing an icing rate prediction model based on data driving;

[0073] S2, obtaining online monitoring and meteorological numerical prediction information of the target area, inputting the pretreated information into the icing rate prediction model, and performing criterion determination on the output icing rate value and corresponding meteorological information, predicting the ice shedding time and ice shedding duration;

[0074] S3, performing fuzzy judgment on the meteorological information and icing rate value corresponding to the ice shedding duration to output the probability density distribution of the ice shedding rate;

[0075] S4, obtaining line digital archives and meteorological statistical data of the target area, and correcting the actual allowable tensile strength of the line;

[0076] S5, obtaining the line ice shedding rate based on Monte Carlo sampling, calculating the flashover probability and the line breaking probability, and then calculating the line comprehensive failure probability;

[0077] S6, comparing the comprehensive failure probability with the preset risk level division standard, and outputting the ice shedding jump event comprehensive risk early warning level.

[0078] Specifically, when step S1 is performed,

[0079] The historical ice thickness record is obtained through the power transmission line monitoring terminal, including but not limited to the ice thickness value and the report time; the meteorological data is obtained through the meteorological sensor of the power transmission line monitoring terminal, including but not limited to the temperature, humidity, wind speed and report time;

[0080] The ice thickness information and meteorological information are preprocessed, and the historical failure data set is constructed, specifically including:

[0081] The icing thickness information and the weather are subjected to data cleaning and interpolation processing, the front and rear weighted mean method is adopted for interpolation, and the corresponding time scale is generated, icing thickness and weather time sequence information containing the same time interval are generated, the difference of icing information of adjacent time is subjected to time differentiation, the average of adjacent time weather information is taken, then the maximum-minimum normalization method is adopted for normalization processing, and a historical fault data set is formed.

[0082] The maximum-minimum normalization calculation formula is:

[0083]

[0084] In the formula, is the normalized value of the i-th data, and is the maximum value and the minimum value in the data, i corresponds to icing rate, temperature, humidity or wind speed data.

[0085] An icing rate prediction model based on a variational autoencoder and an LSTM is established, wherein the structure of the prediction model comprises a historical encoder, a reparameterization layer and a prediction decoder;

[0086] The historical encoder comprises 2 layers of LSTM, a full connection layer, a full connection mean layer and a full connection variance layer; the reparameterization layer comprises a sampling layer and a full connection layer; the prediction decoder comprises 1 layer of LSTM and a full connection layer;

[0087] The number of neurons of the first layer of LSTM in the historical encoder is 128, which is used to learn the dependence relationship of the time sequence, the number of neurons of the second layer of LSTM is 64, the number of neurons of the full connection layer is 32, and ReLU is used as an activation function, the number of neurons of the full connection mean layer is 16, which is used to calculate the mean vector of the latent space distribution, and the number of neurons of the full connection variance layer is 16, which is used to calculate the logarithmic variance vector of the latent space distribution;

[0088] The reparameterization layer first samples a derivable latent context vector z from the historical encoder, and utilizes the full connection layer to map the initial condition in the prediction decoder;

[0089] The number of neurons of the LSTM in the prediction decoder is 32, which is output through the full connection layer, and adopts linear activation, and the number of neurons is 1;

[0090] The known historical fault data set samples are divided into a training set and a test set, wherein the training set accounts for 70%, and the test set accounts for 30%;

[0091] The loss function combines the accuracy of sequence prediction and the regularization of the latent space, wherein the accuracy of sequence prediction adopts mean square error, the regularization of the latent space adopts KL divergence, and the total loss function is:

[0092]

[0093] wherein, is the prediction loss, is the KL divergence, wherein the hyperparameter The initial weight of the can be selected as 0.5, and can be dynamically adjusted according to the performance of the test set;

[0094] During training, the learning rate preheating and weight decay regularization strategies are introduced, the adaptive moment estimation (Adam) optimizer is used to update the parameters of the network until the loss function tends to be stable or the preset maximum training number of rounds is reached to end the joint training.

[0095] The icing rate prediction model can be obtained by evaluating the fine-tuning based on the test set.

[0096] When performing step S2, specifically comprising:

[0097] Based on the longitude and latitude information of the terminal of the meteorological center, the numerical prediction information of the target area is obtained, including but not limited to the reporting time, the prediction time, the temperature, the humidity, the wind speed, and the solar radiation intensity; the existing cumulative process value of the icing thickness of the target area this time and the corresponding meteorological data are obtained through the power line monitoring terminal, including but not limited to the icing thickness, the temperature, the humidity, the wind speed, and the reporting time;

[0098] The meteorological data and the icing thickness data are preprocessed, including time scale conversion, data interpolation processing, average value and differential calculation, and normalization processing;

[0099] The processed data is input into the icing rate prediction model, and the icing rate time series in a future period of time is output;

[0100] It is judged whether the output icing rate meets the state change criterion, and when the state change criterion is met, the thermodynamic deicing criterion is judged to predict the deicing time, specifically comprising:

[0101] When the icing change rate appears negative growth and the value of the negative growth is less than a preset critical value, it is judged whether the corresponding predicted temperature value, solar radiation intensity, and corresponding duration are greater than a preset critical value, wherein the state change criterion is:

[0102]

[0103] wherein, is the icing rate value.

[0104] The thermodynamic deicing criterion is:

[0105]

[0106] wherein, T is a temperature value, I is a solar radiation intensity, T is a duration satisfying the first two criteria in the thermodynamic de-icing criterion.

[0107] When the state change criterion and the thermodynamic de-icing criterion are satisfied simultaneously, the starting time is defined as the de-icing time, preferably, the de-icing duration can be selected as 3h after the de-icing time, the icing rate before the de-icing time is integrated and summed with the icing thickness monitored by the online monitoring system to obtain the icing thickness value at the predicted de-icing time.

[0108] In the step S3, specifically comprises:

[0109] The average ambient temperature, the average solar radiation intensity and the average negative icing growth value in the de-icing duration are taken as input variables, preferably, the triangular membership function is used for membership division;

[0110] The piecewise formula of the triangular membership function is as follows:

[0111]

[0112] In the formula, a is the left end point, at which the membership starts to be nonzero, b is the peak point, at which the membership is 1, and c is the right end point, at which the membership is 0, and here x can be the specific value of the average ambient temperature, the average solar radiation intensity, the average negative icing growth value, the expectation of the de-icing rate and the standard deviation.

[0113] Preferably, the average ambient temperature is covered by 3 fuzzy subsets: low temperature (L), ice melting temperature (M) and high temperature (H); the average solar radiation intensity is covered by 3 fuzzy subsets: weak radiation (L), medium radiation (M) and strong radiation (H); and the average negative icing growth value is covered by 3 fuzzy subsets: low negative growth (L), medium negative growth (M) and high negative growth (H).

[0114] For the output parameter fuzzy set, the fuzzy sets of the expectation and the standard deviation of the de-icing rate are defined, preferably, 3 fuzzy subsets are used to cover the parameter range of the expectation: low (L), medium (M) and high (H); and 3 fuzzy subsets are used to cover the parameter range of the standard deviation: low (L), medium (M) and high (H), and the division of the fuzzy membership function can be continuously tested and improved in application.

[0115] The parameters corresponding to the average ambient temperature, the average solar radiation intensity and the average negative icing growth value are combined and matched to construct the corresponding fuzzy control rule base.

[0116] According to the Mamdani model, a "max-min" reasoning method is adopted, a rule base is used for reasoning, membership values of each input index are substituted into corresponding rules, and membership of the output fuzzy set of the expectation and standard deviation of the ice shedding rate is inferred;

[0117] The gravity of the final output fuzzy set is calculated by using the centroid method, the fuzzy sets of the expectation and standard deviation of the ice shedding rate are de-fuzzified, and the determined expectation and standard deviation are obtained, and the calculation formula is as follows:

[0118]

[0119] In the formula, is the de-fuzzified value, z is each possible value of the index to be de-fuzzified on the domain, is the membership of the output fuzzy set obtained by the "max-min" reasoning method.

[0120] According to the de-fuzzified expectation and standard deviation, the probability density distribution of the ice shedding rate is obtained.

[0121] When step S4 is performed, specifically comprising:

[0122] The S-N curve, line operation time and historical ice shedding jump fault report of the target line are obtained through the target line digital archive, and the historical wind speed value of the target line is obtained through the meteorological bureau;

[0123] The residual strength expression of the line is obtained according to the S-N curve of the line, and the expression of the S-N curve is:

[0124]

[0125] In the formula, is the cyclic stress amplitude, N is the number of cycles, and m and C are both constants in the S-N curve.

[0126] After being subjected to k-level stress impact, the residual strength expression of the component is:

[0127]

[0128] In the formula, is the residual strength value, which is the actual allowable tensile stress of the transmission line in this study , is the initial stress, is the i-th stress value, is the number of times corresponding to the stress value, and b is a constant obtained from the S-N curve, satisfying:

[0129]

[0130] The wind speed range is determined according to the maximum and minimum wind speeds in the historical data, that is, the wind speed range is The wind speed range is divided into M discrete intervals, and the probability of the wind speed in each interval is obtained according to the historical data, and the average of the values at both ends of the discrete interval is taken as the representative wind speed of the interval, that is:

[0131]

[0132] In the formula, is the left boundary of the i-th wind speed interval, is the right boundary of the i-th wind speed interval.

[0133] M sets of discrete wind speed-probability pairs can be obtained, that is,

[0134] The contribution value of each wind speed-probability pair to the degradation of the line resistance is calculated;

[0135] Take for example, take as the average wind speed, use the AR linear filtering method and the minimum AIC criterion to obtain the random vibration wind speed spectrum, and obtain the wind speed time history;

[0136] Based on the wind speed time history, the stress time history of the transmission line within 400s after the start of the truncated vibration is constructed by using ANSYS finite element simulation;

[0137] The rain flow counting method is used to obtain the corresponding stress spectrum under each stress time history, and the maximum stress value in each stress cycle is taken as the representative of the stress cycle, and the stress cycle and the corresponding frequency are obtained;

[0138] The frequency obtained above is the frequency within 400s, and based on the line operation time, the corresponding frequency within the line operation time is inferred, which is represented as .

[0139] The contribution value of the wind speed-probability pair to the degradation of the line resistance is calculated, and the calculation formula of the contribution value of the line resistance degradation is:

[0140]

[0141] In the formula, is the probability corresponding to the representative wind speed, and are the y-th set of stress cycles and the corresponding frequency under the representative wind speed.

[0142] Based on the fault report of the target line historical ice shedding jump, the dynamic stress time history is simulated according to the relevant information, and based on the same idea, the contribution value of each ice shedding jump event to the degradation of the line resistance can be obtained, wherein the value of is 1.​

[0143] Summing up the total line resistance degradation contribution value, the actual allowable tensile stress of the power transmission line is calculated based on the residual strength expression of the element , where the residual strength expression is The total line resistance degradation contribution value is

[0144] When performing step S5, specifically includes:

[0145] Based on the probability density distribution of the ice shedding rate, the single ice shedding rate is obtained by Monte Carlo sampling, which is substituted into the regression curve of the maximum ice jump height and the maximum instantaneous stress fitted offline to obtain the maximum ice jump height and the maximum instantaneous stress under the ice shedding rate. Preferably, the regression curve of the maximum ice jump height and the maximum instantaneous stress can be obtained by ANSYS finite element simulation software, constructing an ice shedding jump simulation model, calculating the maximum ice jump height and the maximum instantaneous stress of the power transmission line under different ice thickness and ice shedding rate, and using Python to fit the independent variable as the ice thickness and the ice shedding rate, and the dependent variable as the maximum ice jump height and the maximum instantaneous stress.

[0146] According to the obtained maximum ice jump height H, the minimum gap distance between the conductors after ice shedding is calculated , and the calculation formula is as follows:

[0147]

[0148] In the formula, is the horizontal distance between the upper phase conductor and the lower phase conductor; is the vertical distance between the upper phase conductor and the lower phase conductor.

[0149] The critical flashover voltage of the line under the minimum gap is calculated, and the calculation formula is as follows:

[0150]

[0151] In the formula, d is the interphase gap distance.

[0152] The critical flashover voltage calculated according to the minimum gap distance is compared with the maximum operating voltage. When it is greater than the maximum operating voltage, no flashover failure occurs, otherwise flashover failure occurs. By multiple sampling, the ratio of the number of flashover failures to the total number of samples is calculated to obtain the flashover probability ;

[0153] The maximum instantaneous stress is compared with the actual allowable tensile stress. When it is greater than the actual allowable tensile stress, wire breakage failure occurs, otherwise no wire breakage failure occurs. By multiple sampling, the ratio of the number of wire breakage failures to the total number of samples is calculated to obtain the wire breakage probability ;

[0154] The comprehensive failure probability of the power transmission line when ice shedding jump is about to occur is calculated by flashover probability and wire breakage probability The calculation formula is:

[0155]

[0156] When step S6 is executed,

[0157] The risk level division standard is:

[0158] When , blue pre-warning is carried out;

[0159] When , yellow pre-warning is carried out;

[0160] When , orange pre-warning is carried out;

[0161] When , red pre-warning is carried out.

[0162] According to the obtained line comprehensive failure probability, combined with the risk level division standard, the comprehensive risk pre-warning level of the power transmission line considering the conductor ice shedding jump is output.

[0163] The present application takes a flashover failure caused by ice shedding jump of a power transmission line in a certain region on January 15, 2024, 12:08 as an example, uses the meteorological and line data before the occurrence of the event, outputs the pre-warning level of this failure, verifies the correctness of the method, and the specific method is as follows:

[0164] S1, meteorological data and ice thickness data at the time of historical icing failure of the power transmission line are obtained to form a historical failure data set, and a data-driven ice rate prediction model is constructed;

[0165] S2, the ice thickness information and meteorological information in the whole process of the failure in the on-line monitoring system are obtained, the ice thickness information and meteorological information before 23:00 on January 14, 2024, and the meteorological information after 23:00 are preprocessed, and input into the ice rate prediction model to predict the ice rate in the next 24 hours, the mean square error is 0.08 mm / h, and the prediction effect is good; based on the meteorological information, when the starting time of simultaneously meeting the state change criterion and the thermodynamic ice shedding criterion is 11:00 and the ice shedding duration is 11:00 to 14:00, the time period covers the actual ice shedding jump time, the ice thickness at the predicted ice shedding time is obtained by integrating the ice rate before the ice shedding time, which is 14.2 mm;

[0166] S3, calculated during the de-icing period, yielded an average temperature of 5.83℃, an average solar radiation intensity of 776.67 W / m², and an average negative icing growth rate of -1.13 mm / h. Based on fuzzy set partitioning and membership calculations, the average temperature falls within the melting temperature range (0.166 membership), the average solar radiation intensity falls within the strong radiation range (0.44 membership), and the average negative icing growth rate falls within the medium negative growth range (0.175 membership). According to the input membership, the expected value and standard deviation of the output from the Mamdani inference and fuzzy control rule base fall within the medium range. Through centroid defuzzification, the expected de-icing rate is 40% with a standard deviation of 20%. Therefore, the probability density distribution of the de-icing rate is... ;

[0167] S4, obtain the SN curve of the line through the target line digital file, where m is 4 and C is... The remaining strength expression of the component can be obtained. Wind speed values ​​recorded over many years for the line are obtained from the meteorological bureau, and the wind speed range is divided into 5 discrete intervals. The probability of wind speed in each interval is calculated, resulting in a set of wind speed-probability pairs. The AR linear filtering method and the minimum AIC criterion are used to calculate the random vibration wind speed spectrum under 5 representative wind speeds, obtaining the wind speed time history. ANSYS finite element simulation is used to construct the stress time history for the first 400 seconds after the start of the transmission line's truncation vibration. The rainflow counting method is used to obtain the stress spectrum corresponding to each stress time history, and the maximum stress value in each stress cycle represents that stress cycle, obtaining each stress cycle and its corresponding frequency. The line has been in operation for 15 years; the stress cycles and their corresponding frequencies experienced at this time scale are inferred, and their resistance degradation contribution value is calculated. The resistance degradation contribution value of 5 ice-breaking jump impact events in the historical record is calculated using the same approach. The initial design allows a tensile strength of 220 MPa; substituting this into the component remaining strength expression, the existing actual allowable tensile strength is calculated to be 170 MPa.

[0168] S5, from A total of 10,000 samples were taken. Taking a single ice removal rate of 41% as an example, the ice thickness of 14.2 mm and the ice removal rate of 48% were substituted into the regression curves of the maximum ice jump height and the maximum instantaneous stress. The results showed that the maximum ice jump height was 8.64 m and the maximum instantaneous stress was 75 MPa. At this point, the minimum gap... The value is 4.1m, corresponding to the critical flashover voltage under this gap. The voltage is 1086.4 kV, and the maximum operating voltage of this line is 777.82 kV. It is determined that a flashover fault occurred during this sampling, but no open-circuit fault occurred. The probability of flashover is obtained through multiple sampling. The probability of a disconnection is 85%. If the overall failure probability is approximately 0%, then... It is 85%;

[0169] S6, combined with the risk level classification standard, output orange warning, consistent with the actual risk situation.

Claims

1. A comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping, characterized in that: The method includes: Step 1: Obtain meteorological data and ice thickness data of historical icing faults of transmission lines to form a historical fault dataset, and construct a data-driven icing rate prediction model. Step 2: Obtain online monitoring and meteorological numerical forecast information for the target area. Input the preprocessed information into the icing rate prediction model, and make judgments on the output icing rate value and corresponding meteorological information to predict the icing time and duration. Step 3: Perform fuzzy judgment on the meteorological information and icing rate value corresponding to the duration of de-icing and output the probability density distribution of the de-icing rate; Step 4: Obtain digital archives and meteorological statistics of the target area's railway lines, and correct the actual allowable tensile strength of the railway lines; Step 5: Based on Monte Carlo sampling, obtain the line de-icing rate, calculate the flashover probability and the line break probability, and then calculate the overall line fault probability. Step 6: Compare the overall failure probability with the preset risk level classification standard, and output the overall risk warning level of the ice-breaking jump event.

2. The comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping according to claim 1, characterized in that: Methods for obtaining historical fault datasets include: Historical icing thickness records, including icing thickness values ​​and reporting times, are obtained through the transmission line monitoring terminal; meteorological data, including temperature, humidity, wind speed, and reporting times, are obtained through the meteorological sensors of the transmission line monitoring terminal. Data cleaning and interpolation were performed on the icing thickness and meteorological information. The weighted mean method was used for interpolation, and the time scale was matched to generate icing thickness and meteorological time series information with the same time interval. The difference between icing information at adjacent time was differentiated in time, and the meteorological information at adjacent time was averaged. Then, the maximum-minimum normalization method was used for normalization to form a historical fault dataset.

3. The comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping according to claim 1, characterized in that: The data-driven icing rate prediction model structure includes a history encoder, a reparameterization layer, and a prediction decoder. The history encoder consists of two LSTM layers, a fully connected layer, a fully connected mean layer, and a fully connected variance layer. The first LSTM layer in the history encoder has 128 neurons and is used to learn the dependencies of the time series. The second LSTM layer has 64 neurons. The fully connected layer has 32 neurons and uses ReLU as the activation function. The fully connected mean layer is used to calculate the mean vector of the latent spatial distribution and has 16 neurons. The fully connected variance layer is used to calculate the log-variance vector of the latent spatial distribution and has 16 neurons. The reparameterization layer includes a sampling layer and a fully connected layer; The reparameterization layer first samples a differentiable latent context vector z from the historical encoder and then uses a fully connected layer to map the initial conditions in the predictive decoder. The predictive decoder consists of one LSTM layer and a fully connected layer; the LSTM in the predictive decoder has 32 neurons, and the output is through the fully connected layer, using linear activation with 1 neuron. The known historical fault dataset samples are divided into training set and test set; The loss function combines the accuracy of sequence prediction with the regularization of the latent space. The accuracy of sequence prediction is expressed using mean squared error, and the regularization of the latent space is expressed using KL divergence. The total loss function is: ； In the formula, To predict losses, Let KL divergence be the hyperparameter, where KL divergence is the hyperparameter. The initial weight can be set to 0.5 and can be dynamically adjusted based on the performance of the test set. During training, a learning rate warm-up and weight decay regularization strategy is introduced, and the Adam optimizer algorithm with adaptive moment estimation is used to update the network parameters until the loss function tends to stabilize or the preset maximum number of training rounds is reached to end the joint training. The icing rate prediction model was obtained by evaluating and fine-tuning based on the test set.

4. The comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping according to claim 1, characterized in that: Methods for predicting the timing and duration of ice shedding include: Preprocessing of meteorological and icing thickness data includes time scale transformation, data interpolation, average and differential calculations, and normalization. The preprocessed data is input into the icing rate prediction model, and the output is the time series of icing rate in the future period. Determine whether the output icing rate satisfies the state change criterion. If the state change criterion is satisfied, perform a thermodynamic de-icing criterion judgment to predict the de-icing time, specifically including: When the rate of icing change shows a negative increase and the negative increase is less than a preset critical value, it is determined whether the corresponding predicted temperature value, solar radiation intensity, and corresponding duration are greater than the preset critical values. The state change criterion is as follows: ； In the formula, This represents the icing rate value. The thermodynamic de-icing criterion is: ; In the formula, This is the temperature value. The intensity of solar radiation. To satisfy the duration of the first two criteria in the thermodynamic de-icing criterion; When both the state change criterion and the thermodynamic de-icing criterion are satisfied, the starting time is defined as the de-icing time, and the de-icing duration is selected as 3 hours after the de-icing time. The icing rate before the de-icing time is integrated and summed with the icing thickness monitored by the online monitoring system at this time to obtain the icing thickness value at the predicted de-icing time.

5. The comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping according to claim 1, characterized in that: Methods for outputting the probability density distribution of the de-icing rate include: The average ambient temperature, average solar radiation intensity, and average negative growth of icing during the de-icing period were used as input variables, and the expected value and standard deviation of the de-icing rate were used as output variables to divide the membership degree; the triangular membership function was used to divide the membership degree. By combining and matching the parameters corresponding to the average ambient temperature, average solar radiation intensity, and average negative growth value of icing, a corresponding fuzzy control rule base is constructed. Based on the Mamdani model, the "max-min" reasoning method is adopted. Reasoning is performed based on the rule base. The membership value of each input index is substituted into the corresponding rule to reason about the membership of the fuzzy set of the expected value and standard deviation of the de-icing rate. The centroid method is used to calculate the centroid of the final output fuzzy set, thereby defuzzifying the fuzzy set of the expected value and standard deviation of the ice removal rate, obtaining the definite expected value and standard deviation, and thus obtaining the probability density distribution of the ice removal rate.

6. The comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping according to claim 1, characterized in that: Methods for correcting the actual allowable tensile strength of the line include: Obtain the line's SN curve, line operating time, and historical de-icing jump fault reports through the target line's digital archive; obtain the target line's historical wind speed values ​​through the meteorological bureau; The expression for the residual strength of the line is obtained from the line's SN curve; The wind speed range is determined based on the maximum and minimum wind speeds in historical data. The wind speed range is divided into M discrete intervals, and M discrete wind speed-probability pairs are calculated based on historical data. The random vibration wind speed spectrum under each representative wind speed was obtained by using the AR linear filtering method and the minimum AIC criterion, and the wind speed time history was obtained. The stress time history of the first 400s after the start of the truncated vibration of the transmission line was constructed by using ANSYS finite element simulation. The stress spectrum corresponding to each stress time history is obtained by rainflow counting method; based on the line operation time, the stress cycles and their corresponding frequencies during the line operation time are inferred, thereby calculating the contribution of each wind speed-probability to the line resistance degradation. Based on the fault reports of historical de-icing jumps of the target line, the dynamic stress time history is simulated according to relevant information. Based on the same idea, the contribution value of each de-icing jump event to the line resistance degradation is obtained. The total contribution value of line resistance degradation is summed, and the actual allowable tensile stress of the corresponding transmission line is calculated based on the residual strength expression of the component.

7. The comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping according to claim 1, characterized in that: Methods for calculating the overall fault probability of a line include: Based on the probability density distribution of the de-icing rate, the de-icing rate of a single instance is obtained through Monte Carlo sampling. This rate is then substituted into the regression curves of the maximum ice jump height and the maximum instantaneous stress obtained by fitting simulation experimental data to obtain the maximum ice jump height and the maximum instantaneous stress under that de-icing rate. Calculate the minimum phase-to-phase gap distance after conductor de-icing. The critical flashover voltage is calculated based on the minimum gap distance and compared with the maximum operating voltage to determine whether a flashover fault has occurred in this sampling. The above sampling and judgment process is repeated, and the flashover probability is calculated based on the judgment result. By comparing the maximum instantaneous stress with the actual allowable tensile stress, it is determined whether a wire breakage fault has occurred in this sampling. The above sampling and judgment process is repeated, and the probability of wire breakage is calculated based on the judgment result. The combined fault probability of a transmission line undergoing de-icing jump is calculated using flashover probability and line break probability.

8. The comprehensive risk early warning method for transmission lines considering conductor de-icing and jumping according to claim 1, characterized in that: The risk level classification criteria are as follows: when A blue alert has been issued. when A yellow alert has been issued. when An orange alert has been issued. when A red alert was issued. Based on the obtained comprehensive line fault probability and combined with the risk level classification standard, the comprehensive risk warning level of the transmission line considering conductor de-icing and jumping is output.

Images