CN117217546B - Power transmission line lightning trip prediction model, method, system and storage medium - Google Patents
Power transmission line lightning trip prediction model, method, system and storage medium Download PDFInfo
- Publication number
- CN117217546B CN117217546B CN202311473535.6A CN202311473535A CN117217546B CN 117217546 B CN117217546 B CN 117217546B CN 202311473535 A CN202311473535 A CN 202311473535A CN 117217546 B CN117217546 B CN 117217546B
- Authority
- CN
- China
- Prior art keywords
- tripping
- prediction
- transmission line
- model
- predicted
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000005540 biological transmission Effects 0.000 title claims abstract description 68
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000013528 artificial neural network Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 21
- 238000010276 construction Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 238000012502 risk assessment Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 238000010801 machine learning Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 5
- 230000002265 prevention Effects 0.000 abstract description 2
- 238000012545 processing Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000002372 labelling Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000013531 bayesian neural network Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000012633 leachable Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the technical field of lightning strike risk prevention of power transmission lines, in particular to a power transmission line lightning strike tripping prediction model, a power transmission line lightning strike tripping prediction method, a power transmission line lightning strike tripping prediction system and a power transmission line lightning strike tripping prediction storage medium. According to the invention, the lightning data of each period is obtained through time slicing, and the terrain is rasterized and filled with the altitude value, namely the time segment Lei Shuju, so that obviously, the processing of more refined characteristics of the invention is beneficial to capturing more information by the neural network, and the accuracy of the information is further improved. According to the invention, the relevant characteristics of the lightning trip probability are processed in a finer granularity way, so that the prediction accuracy is improved. In the invention, the uncertainty and the average tripping probability are further combined by combining the uncertainty parameter value mode of the neural network, so that the accuracy of a model is improved, and the uncertainty caused by some disturbance factors in actual conditions is considered.
Description
Technical Field
The invention relates to the technical field of lightning strike risk prevention of power transmission lines, in particular to a power transmission line lightning strike tripping prediction model, a power transmission line lightning strike tripping prediction method, a power transmission line lightning strike tripping prediction system and a power transmission line lightning strike tripping prediction storage medium.
Background
Along with the increasing huge construction scale of the high-speed rail power supply network, the span of the power transmission line and the voltage level are correspondingly improved, the risk of lightning strike on the line is increased, and the economic loss and the potential safety hazard of the high-speed rail power supply network caused by lightning disasters are more serious. The traditional lightning protection is passive lightning protection, namely the same lightning protection equipment is installed on each section of transmission line, and the defects of low pertinence, high investment cost and the like exist. Therefore, it is important to develop active lightning protection. The active lightning protection is to predict the probability of lightning trip-out in the next period according to some risk factors related to the lightning trip-out, divide different risk classes for the areas with different lightning trip-out probabilities, and achieve different protection works aiming at different risk classes.
The existing lightning trip risk assessment method mainly comprises a traditional method and a deep learning method, wherein the traditional method is used for completing prediction by manually designing a mathematical model of lightning trip risk factors and lightning trip probability and then dividing risk classes, and has two defects: (1) Requiring the mathematical model designer to have mathematical and electrical related expertise; (2) A large amount of historical lightning trip data is not fully utilized. The deep learning method can complete end-to-end training and then predict by only putting a large amount of historical lightning data into the model without the need of users to have professional knowledge in the relevant fields. The existing deep learning method for lightning trip risk assessment is a fully-connected neural network, and has two problems: (1) Lightning trip risk factors with temporal and spatial characteristics (such as lightning coordinate data of a certain area in a period of time) cannot be well processed; (2) The basis of the classification risk is only the predicted lightning trip probability, the uncertainty of the prediction result is not considered, and the full-connection network parameter is fixed, so that the uncertainty of the prediction result cannot be characterized because the prediction value is also a fixed value, and the uncertainty of the prediction result cannot be included in the classification risk.
Disclosure of Invention
In order to overcome the defects of strong specialization, less leachable characteristics and low precision of the lightning risk prediction method in the prior art, the invention provides a lightning trip prediction model of a power transmission line, and the trip prediction accuracy of the power transmission line is greatly improved.
According to the method for constructing the power transmission line lightning trip prediction model, firstly, a basic model is constructed based on a neural network, parameters of the basic model are updated by combining a learning sample, and the converged basic model is obtained and used as a trip prediction model;
the learning sample is denoted as { P (t 0), y (t 0) }; p (t 0) is a time sample of the transmission line on a time node t0, y (t 0) is a known trip probability label of the transmission line on a time period taking t0 as a starting time, y (t 0) epsilon {0,1};
the construction of the time sample P (t 0) comprises the following steps;
s11, acquiring a square prediction area taking a tower pole to be predicted as a center, dividing the prediction area into multiple stages of buffer zones from inside to outside, and setting the weight of each buffer zone; on the prediction area, the weight of each buffer zone is sequentially reduced from the tower pole outwards;
s12, rasterizing a prediction area into an N multiplied by N grid matrix, and counting lightning data in the last G continuous time periods before a time node t0 in the prediction area, wherein the time period t-G represents the (g+1) th time period before the time node t0, G is more than or equal to 0 and less than or equal to G-1, and G is more than or equal to 2; let x (i, j) represent the ith row and jth column grid areas in the pre-determined area, and r (i, j, t-g) represent the number of lightning occurrences in grid area x (i, j) over period t-g; i is more than or equal to 1 and less than or equal to N, j is more than or equal to 1 and less than or equal to N; let v (i, j) represent the height of the grid region x (i, j);
s13, constructing a time sample P (t 0) of the transmission line in a period t 0;
P(t0)={q(i,j,t0)×ε(i,j)|1≤i≤N,1≤j≤N}
q(i,j,t0)={r(i,j,t-G+1),r(i,j,t-G+2),…,r(i,j,t-g),…,r(i,j,t);v(i,j)}
epsilon (i, j) is the weight of the buffer zone where the grid region x (i, j) is located; q (i, j, t 0) is an intermediate parameter, and r (i, j, t), r (i, j, t-g+2), r (i, j, t-g+1) represent the number of lightning occurrences of the grid region x (i, j) on the 1 st, G-1 st and G th period, respectively, of the time node t0 onward.
Preferably, the training of the base model comprises the steps of:
s1, combining historical data to construct a learning sample { P (t 0), y (t 0) };
s2, constructing a basic model based on a neural network and initializing model parameters, wherein the input of the basic model is a time sample of a designated time node, and the output of the basic model is a tripping probability predicted value y ', y' E [0,1] of the power transmission line on a period taking the designated time node as the starting time;
s3, dividing the learning sample into a training data set and a verification data set;
s4, selecting a plurality of training samples from the training data set, enabling the basic model to perform machine learning on the training samples, and selecting parameters from the set parameter distribution to update parameters of the basic model in the learning process; then selecting M test samples from the verification data set, and calculating the loss of the basic model by combining the test samples;
s5, judging whether the basic model converges or not; if not, optimizing the set parameter distribution through gradient descent to reduce the loss, and returning to the step S4; if yes, the converged base model is made to be the tripping prediction model.
Preferably, in S2, the initial parameters of the base model are randomly selected from the gamma distribution.
Preferably, the parameter distribution set in S4 selects a generalized extremum distribution.
Preferably, let the control parameters of the gamma distribution be denoted as α and β, and the control parameters of the generalized extremum distribution be denoted as μ and σ; s5, optimizing the set parameter distribution through optimizing control parameters mu and sigma; the calculation formula of the loss in S4 is as follows:
Loss=∑ H h=1 ln[Gev(w(h)|µ,σ)]-∑ H h=1 ln[Gev(w(h)|α,β)]+[∑ M m=1 (y(m)-y'(m)) 2 ]/M
wherein H is the number of nodes in the base model, w (H) is a parameter of the H node in the base model, gev (w (H) |μ, σ) represents the probability of w (H) in the generalized extremum distribution under the control of the parameters μ and σ, gev (w (H) |α, β) represents the probability of w (H) in the gamma distribution under the control of the parameters α and β; m is the number of test samples used for each round of test, y (M) is a known trip probability label in the mth test sample, and y' (M) is a trip probability predicted value of the mth test sample output by the basic model.
The invention provides a method for predicting lightning trip of a power transmission line, which greatly improves the pre-storing precision of the trip and comprises the following steps:
st1, obtaining the generalized extremum distribution finally optimized in the basic model training process as target distribution;
st2, randomly sampling from target distribution to update each node parameter in the basic model, and taking the updated basic model as a target model; inputting a time sample P (t 0) of a time node to be predicted of the power transmission line to be predicted into a target model, outputting a tripping probability prediction value y ' by the target model, and adding the tripping probability prediction value y ' into a set prediction set, wherein y ' is the tripping probability of the power transmission line in a time period to be predicted, and the time period to be predicted is a time period taking the time node to be predicted as the starting time;
st3, judging whether the number of predicted values y' in the predicted set reaches N; if not, returning to the step St2; if yes, calculating an average value of the predicted values y' in the predicted set as an average trip rate y (avg); and calculating the mean square error of the predicted value y' in the predicted set as an uncertainty coefficient y (p);
st4, calculating a tripping risk coefficient y (r) of the transmission line to be predicted in a period to be predicted by combining the average tripping rate y (avg) and the uncertainty coefficient y (p); y (r) =e -y(p)/y(avg) 。
Preferably, step St4 further comprises: acquiring a tripping risk level evaluation result according to the mapping relation between the tripping risk coefficient and the tripping risk level; the number of the tripping risk levels is set manually, and the value range of the tripping risk coefficient is divided into value intervals corresponding to the tripping risk levels one by one.
The invention also provides a lightning trip-out prediction system and a storage medium of the power transmission line, and provides a carrier for the lightning trip-out prediction method of the power transmission line.
The invention provides a lightning trip-out prediction system of a power transmission line, which is characterized by comprising the following components:
the sample construction module is used for constructing time samples and power transmission line tripping conditions on all known time nodes by combining historical data, and constructing learning samples by combining the time samples and the power transmission line tripping conditions;
the model training module is used for training the constructed basic model by combining the learning sample so as to obtain a tripping prediction model;
the risk prediction module is used for acquiring a time sample of the power transmission line to be predicted on the time node to be predicted, inputting a tripping prediction model, and outputting a tripping probability prediction value by the tripping prediction model.
Preferably, the system further comprises a risk coefficient calculation module and a risk assessment module;
the risk prediction module selects a plurality of groups of model parameters from the final set parameter distribution, and obtains a tripping probability prediction value which is output by a tripping prediction model adopting each group of model parameters aiming at a time sample of the power transmission line to be predicted on a time node to be predicted;
the risk coefficient calculation module obtains an average value y (avg) and a mean square error y (p) of the tripping probability prediction values to be substituted into a formula to calculate a tripping risk coefficient y (r); y (r) =e -y(p)/y(avg) ；
And the risk assessment module acquires a tripping risk level assessment result according to the mapping relation between the tripping risk coefficient and the tripping risk level.
The storage medium is stored with a computer program, a tripping prediction model and optimized generalized extremum distribution, and the computer program is used for realizing the power transmission line lightning trip prediction method when being executed.
The invention has the advantages that:
(1) The power transmission line lightning trip prediction model (trip prediction model for short) provided by the invention is a data-driven prediction method, is an end-to-end method, and can realize the prediction effect only by collecting data to complete the training of the model. The invention needs to model equipment such as lightning tower poles, has simple operation and lower professional difficulty.
(2) According to the invention, the relevant characteristics of the lightning trip probability are processed in a finer granularity way, so that the prediction accuracy is improved. The relevant characteristics of the lightning trip probability selected in the invention are lightning strike data in a tower pole area and landform data of a certain area of the tower pole. The method is different from the traditional method in that average lightning falling data in one year is taken, and the landform data are simply summarized as hills, mountains and the like; according to the invention, the lightning data of each period is obtained through time slicing, and the terrain is rasterized and filled with the altitude value, namely the time segment Lei Shuju, so that obviously, the processing of more refined characteristics of the invention is beneficial to capturing more information by the neural network, and the accuracy of the information is further improved.
(3) The invention predicts the tripping rate in the future period, is beneficial to actively protecting the tower pole when a thunderstorm occurs, for example, a certain tower pole is predicted to be tripped by lightning, the electric brake can be actively cut off to avoid the breakdown of the lightning, and the time period predicted to be tripped is re-overlapped. Therefore, the time-sharing prediction method not only improves the time precision of prediction, but also is beneficial to active protection.
(4) The parameters of the neural network are valued from the appointed probability distribution, which is favorable for capturing uncertainty of the prediction result, and is more in line with the actual situation that the same weather condition and landform in the real world do not necessarily have the same tripping result, thereby further improving the accuracy of the model prediction point.
(5) In the invention, the uncertainty and the average tripping probability are further combined by combining the uncertainty parameter value mode of the neural network, so that the accuracy of a model is improved, and the uncertainty caused by some disturbance factors in actual conditions is considered.
Drawings
FIG. 1 is a flow chart of a time sample construction method;
FIG. 2 is a flow chart of a method for constructing a lightning trip prediction model of a power transmission line;
FIG. 3 is a flow chart of a method for predicting lightning trip of a power transmission line;
FIG. 4 is a schematic diagram of prediction region rasterization;
FIG. 5 is a graph showing the comparison of the performance of three models in the examples;
fig. 6 is a graph showing the performance of the trip prediction model and the trip prediction method according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1 and 2, the method for constructing a lightning trip prediction model of a power transmission line according to the present embodiment includes the following steps S1 to S5.
S1, combining historical data to construct a learning sample; the learning sample is denoted as { P (t 0), y (t 0) }; p (t 0) is a time sample of the transmission line at time node t0, y (t 0) is a known trip probability signature of the transmission line at a time period starting at t0, y (t 0) ∈ {0,1}. The length of the time period is a set value, and can be specifically taken from 5 minutes to 1 hour.
The construction of the time samples P (t 0) includes the following steps S11-S13.
S11, acquiring a square prediction area taking a tower pole to be predicted as a center, dividing the prediction area into multiple stages of buffer zones from inside to outside, and setting the weight of each buffer zone; on the prediction area, the weight of each buffer zone is sequentially reduced from the tower pole outwards.
S12, rasterizing a prediction area into an N multiplied by N grid matrix, and counting lightning data in the last G continuous time periods before a time node t0 in the prediction area, wherein the time period t-G represents the (g+1) th time period before the time node t0, G is more than or equal to 0 and less than or equal to G-1, and G is more than or equal to 2; let x (i, j) represent the ith row and jth column grid areas in the pre-determined area, and r (i, j, t-g) represent the number of lightning occurrences in grid area x (i, j) over period t-g; i is more than or equal to 1 and less than or equal to N, j is more than or equal to 1 and less than or equal to N; let v (i, j) denote the height of the grid region x (i, j).
S13, constructing a time sample P (t 0) of the transmission line on a time node t 0;
P(t0)={q(i,j,t0)×ε(i,j)|1≤i≤N,1≤j≤N}
q(i,j,t0)={r(i,j,t-G+1),r(i,j,t-G+2),…,r(i,j,t-g),…,r(i,j,t);v(i,j)}
epsilon (i, j) is the weight of the buffer zone where the grid region x (i, j) is located; q (i, j, t 0) is an intermediate parameter, and r (i, j, t), r (i, j, t-g+2), r (i, j, t-g+1) represent the number of lightning occurrences of the grid region x (i, j) on the 1 st, G-1 st and G th period, respectively, of the time node t0 onward.
Specifically, let the time period length be T, and the designated time node be T0, the time interval of the time period r (i, j, T) be [ T0-T, T0], the time interval of r (i, j, T-g+2) be [ T0- (G-1) T, T0- (G-2) T ], and the time interval of r (i, j, T-g+1) be [ T0-GT, T0- (G-1) T ]; the time interval of the period with the time node T0 as the start time is [ T0, t0+t ].
S2, constructing a basic model based on a neural network and initializing model parameters, wherein the input of the basic model is a time sample of a designated time node, and the output of the basic model is a tripping probability predicted value y ', y' E [0,1] of the power transmission line on a period taking the designated time node as the starting time.
When the step is implemented, parameters of the basic model are randomly selected from gamma distribution, and control parameters of the gamma distribution are alpha and beta.
S3, dividing the learning sample into a training data set and a verification data set.
S4, selecting a plurality of training samples from the training data set, enabling the basic model to perform machine learning on the training samples, and selecting parameters from the set generalized extremum distribution to update parameters of the basic model in the learning process; then M test samples are selected from the validation data set, and the loss of the base model is calculated in combination with the test samples.
S5, judging whether the basic model converges or not; if not, optimizing control parameters mu and sigma of generalized extremum distribution through gradient descent to reduce loss, and returning to the step S4; if yes, the converged base model is made to be the tripping prediction model.
The loss function is:
Loss=∑ H h=1 ln[Gev(w(h)|µ,σ)]-∑ H h=1 ln[Gev(w(h)|α,β)]+[∑ M m=1 (y(m)-y'(m)) 2 ]/M
wherein H is the number of nodes in the base model, w (H) is a parameter of the H node in the base model, gev (w (H) |μ, σ) represents the probability of w (H) in the generalized extremum distribution under the control of the parameters μ and σ, gev (w (H) |α, β) represents the probability of w (H) in the gamma distribution under the control of the parameters α and β; m is the number of test samples used for each round of test, y (M) is a known trip probability label in the mth test sample, and y' (M) is a trip probability predicted value of the mth test sample predicted by the basic model.
In the loss function, Σ H h=1 ln[Gev(w(h)|µ,σ)]-∑ H h=1 ln[Gev(w(h)|α,β)]The method is beneficial to ensuring that the difference between posterior distribution, namely generalized extremum distribution, and prior distribution, namely gamma distribution, is as small as possible, and the last item ensures that the difference between a predicted value and a true value is as small as possible, so that the convergence speed and the accuracy in the model training process are balanced.
Referring to fig. 3, a lightning trip prediction method for a power transmission line according to the present embodiment includes the following steps St1 to St4.
St1, obtaining final generalized extremum distribution as target distribution;
st2, randomly sampling from target distribution to update each node parameter in the basic model, and taking the updated basic model as a target model; inputting a time sample P (t 0) of a time node to be predicted of the power transmission line to be predicted into a target model, outputting a tripping probability prediction value y ' by the target model, and adding the tripping probability prediction value y ' into a set prediction set, wherein y ' is the tripping probability of the power transmission line in a time period to be predicted, and the time period to be predicted is a time period taking the time node to be predicted as the starting time;
st3, judging whether the number of predicted values y' in the predicted set reaches N; if not, returning to the step St2; if yes, calculating an average value of the predicted values y' in the predicted set as an average trip rate y (avg); and calculating the mean square error of the predicted value y' in the predicted set as an uncertainty coefficient y (p);
st4, calculating a tripping risk coefficient y (r) of the transmission line to be predicted in a period to be predicted by combining the average tripping rate y (avg) and the uncertainty coefficient y (p), and acquiring a tripping risk level assessment result according to the mapping relation of the tripping risk coefficient and the tripping risk level.
y(p)=(1/N)×∑ N j=1 (y(avg)-y'(j))
y(avg)=(1/N)×∑ N j=1 y'(j)
y(r)=e -y(p)/y(avg)
y' (j) is the j-th trip probability predictor in the prediction set.
The trip prediction model described above is verified in connection with the specific embodiment below.
In this embodiment, a basic model is constructed based on a convolutional cyclic bayesian neural network. In this embodiment, the prediction area is divided into three levels of buffer zones from inside to outside, specifically, as shown in fig. 4, the first level of buffer zone is a square area with the prediction area center as the center and the side length of L, the second level of buffer zone is a remaining area after the first level of buffer zone is removed from the square area with the prediction area center as the center and the side length of 2L, the third level of buffer zone is a remaining area after the second level of buffer zone is removed from the square area with the prediction area center as the center and the side length of 3L, and 3L is the side length of the prediction area.
Let n=8 in this embodiment to rasterize the prediction region; for a grid spanning multiple buffer zones, its weight takes the greatest weight in the buffer zone spanned, i.e., the grid is given to the grid where its inner region is located. In this embodiment, let g=3, and the length of each period is 15 minutes. In the embodiment, a tripping sample and an unbuckled sample are built based on all lightning monitoring data of a certain section of high-speed rail power supply network in 2015-2020 and line details of 110kV and above power transmission lines, and are collectively called as labeling samples. The tripping sample is selected from a lightning leading area of a tower pole, wherein lightning tripping actually occurs, and the characteristic quantity data is taken as lightning falling data close to the lightning leading area within 45 minutes before tripping and the landform parameters of the lightning leading area. The untripping sample is taken as a lightning leading area of a tower pole which has lightning falling but has no tripping on the same day, and the characteristic quantity data is taken as the lightning falling data close to the lightning leading area and the landform parameters of the lightning leading area in 45min before different towers at random time on the same day.
In this embodiment, the labeling samples are divided into training samples and test samples, and the sample distribution is shown in table 1.
Table 1: labeling sample distribution
In this embodiment, the above method for constructing a lightning trip prediction model of a power transmission line is performed in combination with training samples, so as to obtain the trip prediction model provided by the present invention. At this time, the training samples are first preprocessed into learning samples { P (t 0), y (t 0) }, and then the training samples are combined to obtain the trip prediction model through training in steps S1-S5.
In this embodiment, the convolutional neural network and the LSTM network are further directly trained in combination with the training samples, so as to obtain a convolutional neural network model and an LSTM network model as a comparison model.
In this embodiment, the trip prediction model and the two comparison models are verified on the test sample, and the comparison results of the three models on the MAE index are shown in fig. 5, and it can be known in conjunction with fig. 5 that the accuracy and stability of the trip prediction model are far higher than those of the two comparison models.
As shown in fig. 3, in this embodiment, the power transmission line lightning trip prediction method provided by the invention and combined with the trip risk coefficient y (r) is further verified for the trip prediction model.
In this embodiment, the mapping relationship between the trip risk coefficient and the trip risk level is:
y (r) is more than or equal to 0 and less than or equal to 0.25, and the tripping risk level is 1 level;
y (r) is more than or equal to 0.25 and less than or equal to 0.5, and the tripping risk level is 2;
y (r) is more than or equal to 0.5 and less than or equal to 0.75, and the tripping risk level is 3;
y (r) is less than or equal to 0.75 and less than or equal to 1, and the tripping risk level is 4.
In this embodiment, the steps St1-St4 are performed on the test sample in combination with the generalized extremum distribution optimized in the training process of the trip prediction model, and the prediction method obtains the MAE index of the trip probability and the MAE index pair of the trip probability directly output by the trip prediction model, as shown in fig. 6, so that the accuracy and stability of the trip probability value calculated in combination with the trip risk coefficient y (r) are further improved.
It will be understood by those skilled in the art that the present invention is not limited to the details of the foregoing exemplary embodiments, but includes other specific forms of the same or similar structures that may be embodied without departing from the spirit or essential characteristics thereof. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.
Furthermore, it should be understood that although the present disclosure describes embodiments, not every embodiment is provided with a separate embodiment, and that this description is provided for clarity only, and that the disclosure is not limited to the embodiments described in detail below, and that the embodiments described in the examples may be combined as appropriate to form other embodiments that will be apparent to those skilled in the art.
The technology, shape, and construction parts of the present invention, which are not described in detail, are known in the art.
Claims (7)
1. The power transmission line lightning trip prediction method is characterized in that a basic model is firstly constructed based on a neural network, parameter updating is carried out on the basic model in combination with a learning sample, and the converged basic model is obtained and used as a trip prediction model;
the learning sample is denoted as { P (t 0), y (t 0) }; p (t 0) is a time sample of the transmission line on a time node t0, y (t 0) is a known trip probability label of the transmission line on a time period taking t0 as a starting time, y (t 0) epsilon {0,1}; the length of the time period is a set value;
the construction of the time sample P (t 0) comprises the following steps;
s11, acquiring a square prediction area taking a tower pole to be predicted as a center, dividing the prediction area into multiple stages of buffer zones from inside to outside, and setting the weight of each buffer zone; on the prediction area, the weight of each buffer zone is sequentially reduced from the tower pole outwards;
s12, rasterizing a prediction area into an N multiplied by N grid matrix, and counting lightning data in the last G continuous time periods before a time node t0 in the prediction area, wherein the time period t-G represents the (g+1) th time period before the time node t0, G is more than or equal to 0 and less than or equal to G-1, and G is more than or equal to 2; let x (i, j) represent the ith row and jth column grid areas in the pre-determined area, and r (i, j, t-g) represent the number of lightning occurrences in grid area x (i, j) over period t-g; i is more than or equal to 1 and less than or equal to N, j is more than or equal to 1 and less than or equal to N; let v (i, j) represent the height of the grid region x (i, j);
s13, constructing a time sample P (t 0) of the transmission line on a time node t 0;
P(t0)={q(i,j,t0)×ε(i,j)|1≤i≤N,1≤j≤N}
q(i,j,t0)={r(i,j,t-G+1),r(i,j,t-G+2),…,r(i,j,t-g),…,r(i,j,t);v(i,j)}
epsilon (i, j) is the weight of the buffer zone where the grid region x (i, j) is located; q (i, j, t 0) is an intermediate parameter, r (i, j, t), r (i, j, t-g+2), r (i, j, t-g+1) respectively represent the number of lightning occurrences of the grid region x (i, j) on the 1 st, G-1 st and G th periods of time node t0 onward;
the training of the basic model comprises the following steps:
s1, combining historical data to construct a learning sample { P (t 0), y (t 0) };
s2, constructing a basic model based on a neural network and initializing model parameters, wherein the input of the basic model is a time sample of a designated time node, and the output of the basic model is a tripping probability predicted value y ', y' E [0,1] of the power transmission line on a period taking the designated time node as the starting time;
s3, dividing the learning sample into a training data set and a verification data set;
s4, selecting a plurality of training samples from the training data set, enabling the basic model to perform machine learning on the training samples, and selecting parameters from the set parameter distribution to update parameters of the basic model in the learning process; then selecting M test samples from the verification data set, and calculating the loss of the basic model by combining the test samples;
s5, judging whether the basic model converges or not; if not, optimizing the set parameter distribution through gradient descent to reduce the loss, and returning to the step S4; if yes, the converged basic model is used as a tripping prediction model;
s4, selecting generalized extremum distribution of the parameter distribution set in the S;
the lightning trip-out prediction method of the power transmission line comprises the following steps:
st1, obtaining the generalized extremum distribution finally optimized in the basic model training process as target distribution;
st2, randomly sampling from target distribution to update each node parameter in the basic model, and taking the updated basic model as a target model; inputting a time sample P (t 0) of a time node to be predicted of the power transmission line to be predicted into a target model, outputting a tripping probability prediction value y ' by the target model, and adding the tripping probability prediction value y ' into a set prediction set, wherein y ' is the tripping probability of the power transmission line in a time period to be predicted, and the time period to be predicted is a time period taking the time node to be predicted as the starting time;
st3, judging whether the number of predicted values y' in the predicted set reaches N; if not, returning to the step St2; if yes, calculating an average value of the predicted values y' in the predicted set as an average trip rate y (avg); and calculating the mean square error of the predicted value y' in the predicted set as an uncertainty coefficient y (p);
st4, calculating a tripping risk coefficient y (r) of the transmission line to be predicted in a period to be predicted by combining the average tripping rate y (avg) and the uncertainty coefficient y (p); y (r) =e -y(p)/y(avg) 。
2. The transmission line lightning trip prediction method according to claim 1, wherein in S2, initial parameters of the base model are randomly selected from gamma distribution.
3. The method for predicting lightning trip of power transmission line according to claim 1, wherein control parameters of gamma distribution are denoted as α and β, and control parameters of generalized extremum distribution are denoted as μ and σ; s5, optimizing the set parameter distribution through optimizing control parameters mu and sigma; the calculation formula of the loss in S4 is as follows:
Loss=∑ H h=1 ln[Gev(w(h)|µ,σ)]-∑ H h=1 ln[Gev(w(h)|α,β)]+[∑ M m=1 (y(m)-y'(m)) 2 ]/M
wherein H is the number of nodes in the base model, w (H) is a parameter of the H node in the base model, gev (w (H) |μ, σ) represents the probability of w (H) in the generalized extremum distribution under the control of the parameters μ and σ, gev (w (H) |α, β) represents the probability of w (H) in the gamma distribution under the control of the parameters α and β; m is the number of test samples used for each round of test, y (M) is a known trip probability label in the mth test sample, and y' (M) is a trip probability predicted value of the mth test sample output by the basic model.
4. The transmission line lightning trip-out prediction method according to claim 1, wherein step St4 further comprises: acquiring a tripping risk level evaluation result according to the mapping relation between the tripping risk coefficient and the tripping risk level; the number of the tripping risk levels is set manually, and the value range of the tripping risk coefficient is divided into value intervals corresponding to the tripping risk levels one by one.
5. A power transmission line lightning trip prediction system employing the power transmission line lightning trip prediction method according to any one of claims 1 to 4, comprising:
the sample construction module is used for constructing time samples and power transmission line tripping conditions on all known time nodes by combining historical data, and constructing learning samples by combining the time samples and the power transmission line tripping conditions;
the model training module is used for training the constructed basic model by combining the learning sample so as to obtain a tripping prediction model;
the risk prediction module is used for acquiring a time sample of the power transmission line to be predicted on the time node to be predicted, inputting a tripping prediction model, and outputting a tripping probability prediction value by the tripping prediction model.
6. The power transmission line lightning trip prediction system of claim 5, further comprising a risk factor calculation module and a risk assessment module;
the risk prediction module selects a plurality of groups of model parameters from the final set parameter distribution, and obtains a tripping probability prediction value which is output by a tripping prediction model adopting each group of model parameters aiming at a time sample of the power transmission line to be predicted on a time node to be predicted;
the risk coefficient calculation module obtains an average value y (avg) and a mean square error y (p) of the tripping probability prediction values to be substituted into a formula to calculate a tripping risk coefficient y (r); y (r) =e -y(p)/y(avg) ；
And the risk assessment module acquires a tripping risk level assessment result according to the mapping relation between the tripping risk coefficient and the tripping risk level.
7. A storage medium, characterized in that a computer program, a trip prediction model and an optimized generalized extremum distribution are stored, which computer program, when executed, is adapted to carry out the transmission line lightning trip prediction method according to any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311473535.6A CN117217546B (en) | 2023-11-08 | 2023-11-08 | Power transmission line lightning trip prediction model, method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311473535.6A CN117217546B (en) | 2023-11-08 | 2023-11-08 | Power transmission line lightning trip prediction model, method, system and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117217546A CN117217546A (en) | 2023-12-12 |
| CN117217546B true CN117217546B (en) | 2024-01-12 |
Family
ID=89042986
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311473535.6A Active CN117217546B (en) | 2023-11-08 | 2023-11-08 | Power transmission line lightning trip prediction model, method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117217546B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004205233A (en) * | 2002-12-24 | 2004-07-22 | Tohoku Electric Power Co Inc | Short-term prediction system by heterogeneous phenomenon |
| CN103399233A (en) * | 2013-07-31 | 2013-11-20 | 国家电网公司 | Lightning trip-out risk prediction method for power transmission line |
| CN108108520A (en) * | 2017-11-29 | 2018-06-01 | 海南电网有限责任公司电力科学研究院 | A kind of transmission line of electricity damage to crops caused by thunder risk forecast model based on Artificial neural network ensemble |
| CN112149898A (en) * | 2020-09-21 | 2020-12-29 | 广东电网有限责任公司清远供电局 | Fault rate prediction model training method, fault rate prediction method and related device |
| CN114966233A (en) * | 2022-05-16 | 2022-08-30 | 国网电力科学研究院武汉南瑞有限责任公司 | Lightning forecasting system and method based on deep neural network |
| CN115034586A (en) * | 2022-05-25 | 2022-09-09 | 国网安徽省电力有限公司电力科学研究院 | Method and device for predicting power failure risk of overhead line in strong convection weather |
| CN115219810A (en) * | 2022-05-18 | 2022-10-21 | 四川大学 | Line trip prediction method based on lightning positioning system |
| CN116070794A (en) * | 2023-03-29 | 2023-05-05 | 合肥工业大学 | Current collecting line impact trip probability prediction and alarm method and system |
-
2023
- 2023-11-08 CN CN202311473535.6A patent/CN117217546B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004205233A (en) * | 2002-12-24 | 2004-07-22 | Tohoku Electric Power Co Inc | Short-term prediction system by heterogeneous phenomenon |
| CN103399233A (en) * | 2013-07-31 | 2013-11-20 | 国家电网公司 | Lightning trip-out risk prediction method for power transmission line |
| CN108108520A (en) * | 2017-11-29 | 2018-06-01 | 海南电网有限责任公司电力科学研究院 | A kind of transmission line of electricity damage to crops caused by thunder risk forecast model based on Artificial neural network ensemble |
| CN112149898A (en) * | 2020-09-21 | 2020-12-29 | 广东电网有限责任公司清远供电局 | Fault rate prediction model training method, fault rate prediction method and related device |
| CN114966233A (en) * | 2022-05-16 | 2022-08-30 | 国网电力科学研究院武汉南瑞有限责任公司 | Lightning forecasting system and method based on deep neural network |
| CN115219810A (en) * | 2022-05-18 | 2022-10-21 | 四川大学 | Line trip prediction method based on lightning positioning system |
| CN115034586A (en) * | 2022-05-25 | 2022-09-09 | 国网安徽省电力有限公司电力科学研究院 | Method and device for predicting power failure risk of overhead line in strong convection weather |
| CN116070794A (en) * | 2023-03-29 | 2023-05-05 | 合肥工业大学 | Current collecting line impact trip probability prediction and alarm method and system |
Non-Patent Citations (1)
| Title |
|---|
| "Lightning trip-out risk assessment and differential lightning protection of 35 kV transmission lines in mountain wind farm";Ying Yang;《Energy Reports》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117217546A (en) | 2023-12-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Titley et al. | A global evaluation of multi‐model ensemble tropical cyclone track probability forecasts | |
| CN106651007A (en) | Method and device for GRU-based medium and long-term prediction of irradiance of photovoltaic power station | |
| CN114966233A (en) | Lightning forecasting system and method based on deep neural network | |
| CN112149887A (en) | PM2.5 concentration prediction method based on data space-time characteristics | |
| CN112285807A (en) | Meteorological information prediction method and device | |
| CN106779222B (en) | Airport ground waiting time prediction method and device | |
| CN105471647A (en) | Power communication network fault positioning method | |
| CN112308281A (en) | Temperature information prediction method and device | |
| CN113505927A (en) | Method, device, equipment and medium for selecting battery capacity of solar bird repelling equipment | |
| CN111985719A (en) | Power load prediction method based on improved long-term and short-term memory network | |
| Madadi et al. | Probabilistic real-time dynamic line rating forecasting based on dynamic stochastic general equilibrium with stochastic volatility | |
| CN111785093A (en) | Air traffic flow short-term prediction method based on fractal interpolation | |
| CN117217546B (en) | Power transmission line lightning trip prediction model, method, system and storage medium | |
| KR101725629B1 (en) | System and method for predicting vehicular traffic based on genetic programming using fitness function considering error magnitude | |
| CN113469457A (en) | Power transmission line fault probability prediction method fused with attention mechanism | |
| CN110222878B (en) | Short-term load prediction method based on artificial fish swarm neural network | |
| CN112836843A (en) | Method and device for predicting base station out-of-service alarm | |
| Solheim et al. | Modelling transmission line failures due to lightning using reanalysis data and a Bayesian updating scheme | |
| CN108509537B (en) | System and method for forecasting galloping probability of power transmission line | |
| Rykov et al. | Risk tree as an assistant tool for the decision-maker | |
| CN115618286A (en) | Transformer partial discharge type identification method, system, equipment, terminal and application | |
| CN113298296A (en) | Method for predicting day-ahead load probability of power transmission substation from bottom to top | |
| CN116610484B (en) | Model training method, fault prediction method, system, equipment and medium | |
| CN112147718B (en) | Lightning stroke prediction method and device for overhead transmission line and storage medium | |
| CN117077546B (en) | Power system load modeling method and system based on data driving |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |