CN115034586A - Method and device for predicting power failure risk of overhead line in strong convection weather - Google Patents
Method and device for predicting power failure risk of overhead line in strong convection weather Download PDFInfo
- Publication number
- CN115034586A CN115034586A CN202210577256.3A CN202210577256A CN115034586A CN 115034586 A CN115034586 A CN 115034586A CN 202210577256 A CN202210577256 A CN 202210577256A CN 115034586 A CN115034586 A CN 115034586A
- Authority
- CN
- China
- Prior art keywords
- decision tree
- data
- overhead line
- strong convection
- characteristic data
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a device for predicting power failure risk of an overhead line in strong convection weather, belonging to the technical field of big data of a power distribution network, wherein the method comprises the following steps: acquiring a training data set, wherein the training data set comprises characteristic data and a data label corresponding to the characteristic data, and the training data comprises strong convection meteorological characteristic data, geographical characteristic data and power grid state characteristic data; initializing a decision tree; taking the training data and the corresponding data labels thereof as the input of a decision tree, and calculating the target function and the output confidence of each leaf in the decision tree; selecting the leaves capable of generating a new tree according to the output confidence of each leaf to generate a new decision tree; and obtaining a sparse decision tree for calculating the power failure probability of the overhead line until all the characteristic data in the training data set are traversed. The invention can efficiently construct the decision tree and has better prediction precision.
Description
Technical Field
The invention relates to the technical field of big data of a power distribution network, in particular to a method and a device for predicting power failure risk of an overhead line in strong convection weather.
Background
Along with the development of economic society, the coverage of a power grid system transformer substation and a power distribution network is continuously enlarged, the invasion of disastrous weather such as strong wind, thunder and lightning, haze and the like to the safe operation of a power grid is increased, and under the influence of extreme weather, a 10kV power distribution overhead line is more likely to trip, so that the line fault of the regional power grid is stopped, and the normal production and life of residents are influenced.
At present, although there are many researches on risk assessment and prevention of power grid meteorological disasters, the main focus is on the influence of weather such as ice, snow, strong wind and the like on a power distribution network. With the increasing of severe weather such as haze and ice coating, the normal operation of power grid equipment is greatly influenced, and a plurality of electric power scientific research units develop deep analysis on power grid meteorological disasters and seek prevention and control measures. The China institute of Electrical science and technology initiates a wide investigation on the technical status quo that China's power grids deal with extreme natural disasters ', determines the types and characteristics of meteorological disasters faced by China's power grids, researches the meteorological influence factors and the space-time distribution characteristics of power grid disasters in respective regions, and lays a foundation for meteorological disaster early warning and defense strategies.
At present, research and analysis on the overhead line power failure prediction in strong convection weather are less, and the method mainly focuses on the overhead line power failure prediction based on a neural network. For example, chinese patent application publication No. CN108596449A discloses a method for predicting reliability of a power distribution network in consideration of the influence of weather on the probability of failure of the power distribution network, which includes the following steps: (1) analyzing the incidence relation between the power distribution network element fault probability and the weather variable; (2) selecting main weather variables by considering both model applicability and model accuracy; (3) forming a power distribution network element fault rate prediction method based on a learnable sparse decision tree; (4) and predicting the reliability of the power distribution network by combining the power distribution network element fault probability prediction result and a power distribution network reliability calculation method. The method can predict the reliability of the power distribution network aiming at the influence of different weather variables on the fault probability of the power distribution network elements.
The overhead line power failure prediction method based on the neural network is characterized in that the factor displays are integrated together for feature extraction and input into the neural network, and the final predicted confidence coefficient is obtained. However, the overhead line power failure prediction has many influencing factors and great differences, such as weather factors, geographic factors, historical failure factors and the like, and the factors have great differences in numerical values, physical meanings and the like. The numerical difference of the input characteristics causes that the neural network training is difficult to converge, so that the high-performance overhead line power failure prediction precision is difficult to realize; and due to the inexplicability of the neural network, the prediction result is difficult to have reasonable physical meaning interpretation.
Disclosure of Invention
The invention aims to solve the technical problem of how to realize the high-precision prediction of the power failure of the overhead line in the strong convection weather.
The invention solves the technical problems through the following technical means:
on one hand, the invention provides a method for predicting power failure risk of an overhead line in strong convection weather, which comprises the following steps:
acquiring a training data set, wherein the training data set comprises characteristic data and a data label corresponding to the characteristic data, and the training data comprises strong convection meteorological characteristic data, geographical characteristic data and power grid state characteristic data;
initializing a decision tree;
taking the training data and the corresponding data labels thereof as the input of the decision tree, and calculating the target function R (d, x, y) and the output confidence of each leaf in the decision tree;
selecting leaves capable of generating a new tree according to the output confidence of each leaf to generate a new decision tree;
until all the characteristic data in the training data set are traversed, obtaining a sparse decision tree;
and calculating the power failure probability of the overhead line in the target area by utilizing the sparse decision tree.
The method is designed and used for training the optimized sparse decision tree by using the strong convection meteorological characteristic data, the geographic characteristic data and the power grid state characteristic data, so that the sparse decision tree has the capability of predicting the power failure probability of the overhead line in the strong convection weather according to the input information; the decision tree algorithm has good capability of processing difference of input data dimensions, and can well cope with difference in numerical values of factors such as weather factors and geographic factors, and when leaves are added in each step of the decision tree to form a sub-tree, each new tree is generated from a leaf which can be divided by an old tree but not from a leaf which can not be divided by the old tree.
Further, the strong convection meteorological characteristic data comprise a strong convection radar intensity maximum value, a lightning current magnitude and lightning attack times, wherein the strong convection radar intensity maximum value is a maximum value in a set time period, the lightning current magnitude is an average value in the set time period, and the radar attack times are lightning attack times in the set time period;
the geographic feature data comprises a local geographic feature and a global geographic feature;
the power grid state characteristic data comprise load times, overload times and defect times.
Further, the acquiring the training set includes:
acquiring corresponding characteristic data from a period of time before a power failure signal;
obtaining the matching of the feature data and the tag information in the time period based on a tag value range, wherein the tag value range is as follows:
further, the initialized decision tree is:
d＝(d un ，δ un ，d split ，δ split ，K，H)
wherein d is un Leaves not changed in d, δ un In order not to alter the corresponding prediction of the leaf, d split For the leaf to be subdivided in d, δ split For the features corresponding to the leaves to be subdivided, H is the total number of leaves and K is the delimiter.
Further, the objective function R (d, x, y) is:
R(d，x，y)＝l(d，x，y)+λH d
wherein d represents a parameter of the decision tree and x represents an input bitThe characteristic data is represented by y, the label corresponding to the characteristic data is represented by l (d, x, y), the decision tree parameter is directly optimized according to the label, and lambda H d A regularization term is represented.
Further, the selecting a leaf capable of generating a new tree according to the output confidence of each leaf to generate a new decision tree includes:
judging the state of the overhead line according to the output confidence of each leaf;
if the state of the overhead line is a power failure state, determining that the leaves are leaves capable of generating a new tree;
and if the state of the overhead line is a normal state, determining that the leaf is a leaf which cannot generate a new tree.
In addition, the invention also provides a device for predicting the power failure risk of the overhead line in the strong convection weather, which comprises:
the system comprises an acquisition module, a data processing module and a data processing module, wherein the acquisition module is used for acquiring a training data set, the training data set comprises characteristic data and a data label corresponding to the characteristic data, and the training data comprises strong convection meteorological characteristic data, geographic characteristic data and power grid state characteristic data;
the initialization module is used for initializing the decision tree;
the training module is used for taking the training data and the corresponding data labels thereof as the input of the decision tree, and calculating the target function R (d, x, y) and the output confidence of each leaf in the decision tree;
the decision tree updating module is used for selecting the leaves capable of generating the new tree according to the output confidence of each leaf to generate the new decision tree;
the output module is used for obtaining a sparse decision tree until all the characteristic data in the training data set are traversed;
and the prediction module is used for calculating the power failure probability of the overhead line in the target area by utilizing the sparse decision tree.
Further, the strong convection meteorological characteristic data comprise a strong convection radar intensity maximum value, a lightning current magnitude and lightning attack times, wherein the strong convection radar intensity maximum value is a maximum value in a set time period, the lightning current magnitude is an average value in the set time period, and the radar attack times are lightning attack times in the set time period;
the geographic feature data comprises a local geographic feature and a global geographic feature;
the power grid state characteristic data comprise load times, overload times and defect times.
Further, the initialized decision tree is:
d＝(d un ，δ un ，d split ，δ split ，K，H)
wherein d is un Leaves not changed in d, δ un In order not to alter the corresponding prediction of the leaf, d split For the leaf to be subdivided in d, δ split For the features corresponding to the leaves to be subdivided, H is the total number of leaves and K is the delimiter.
Further, the objective function R (d, x, y) is:
R(d，x，y)＝l(d，x，y)+λH d
wherein d represents the parameter of the decision tree, x represents the input feature data, y represents the label corresponding to the feature data, l (d, x, y) represents the optimization of the decision tree parameter directly according to the label, and λ H d A regularization term is represented.
The invention has the advantages that:
(1) the method is designed and used for training the optimized sparse decision tree by using the strong convection meteorological characteristic data, the geographic characteristic data and the power grid state characteristic data, so that the sparse decision tree has the capability of predicting the power failure probability of the overhead line in the strong convection weather according to the input information; the decision tree algorithm has good capability of processing difference of input data dimensions, and can well cope with difference in numerical values of factors such as weather factors and geographic factors, and when leaves are added in each step of the decision tree to form a sub-tree, each new tree is generated from a leaf which can be divided by an old tree but not from a leaf which can not be divided by the old tree.
(2) When the leaves are added in each step of the decision tree to form a sub-tree, the decision tree is optimized and constructed by adopting a set loss function, so that the performance of the constructed decision tree is further improved.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
FIG. 1 is a schematic flow chart illustrating a method for predicting a power outage risk of an overhead line in a heavy convection weather according to a first embodiment of the present invention;
FIG. 2 is a schematic block diagram of the prediction of power outage risk of a heavy convection weather overhead line according to a first embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a sparse decision tree model according to a first embodiment of the present invention;
FIG. 4 is a schematic diagram showing a comparison between a sparse decision tree model and a conventional decision tree model in a first embodiment of the present invention, wherein (a) is a structural diagram of the conventional decision tree model, and (b) is a structural diagram of the sparse decision tree model;
FIG. 5 is a schematic structural diagram of a power outage risk prediction system for an overhead line of a heavy convection weather system according to a second embodiment of the present invention;
fig. 6 is a block diagram illustrating the power outage risk prediction of the overhead line in the strong convection weather according to the second embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
As shown in fig. 1 to 2, a first embodiment of the present invention provides a method for predicting a power outage risk of an overhead line in heavy convection weather, where the method includes the following steps:
s10, a training data set is obtained, wherein the training data set comprises characteristic data and a data label corresponding to the characteristic data, and the training data comprises strong convection meteorological characteristic data, geographic characteristic data and power grid state characteristic data.
And S20, initializing a decision tree.
S30, taking the training data and the corresponding data labels as the input of the decision tree, and calculating the target function R (d, x, y) and the output confidence of each leaf in the decision tree.
And S40, selecting the leaves capable of generating the new tree according to the output confidence of each leaf, and generating the new decision tree.
And S50, obtaining a sparse decision tree until all the feature data in the training data set are traversed.
And S60, calculating the power failure probability of the overhead line in the target area by using the sparse decision tree.
It should be noted that, as shown in fig. 3 to fig. 4, unlike the conventional decision tree model, the decision tree in this embodiment is a sparse and optimizable decision tree, which is an iterative decision tree algorithm, the algorithm is composed of a plurality of decision trees, the conclusions of all the trees are accumulated to make the final answer, the construction process of the sparse and optimizable decision tree is hierarchically constructed according to the input data of each different feature, the association between the features is embodied in the construction process of the new tree and is not embodied on the leaves of the old tree, so that each new tree is generated from the reproducible leaves of the old tree and the non-reproducible leaves of the old tree, i.e. d un Based on the analysis, the sparse decision tree uses an asymmetric structure, which is just the embodiment of sparsity, and by the method for constructing the decision tree through the sparsity, a large number of redundant calculations are eliminated, so that the efficiency of constructing the decision tree can be greatly improved, and the performance of the decision tree cannot be influenced on the theoretical analysis.
In the embodiment, the strong convection meteorological characteristic data, the geographic characteristic data and the power grid state characteristic data are designed and used for training the optimized sparse decision tree, so that the sparse decision tree has the capability of predicting the power failure probability of the overhead line in strong convection weather according to the input information; the decision tree algorithm has good capability of processing difference of input data dimensions, can well cope with difference in numerical values of factors such as weather factors and geographic factors, and when leaves are added to the decision tree in each step to form a sub-tree, each new tree is generated from a leaf which can be re-divided by an old tree but not from a leaf which cannot be re-divided by the old tree.
In an embodiment, in the step S10, the strong convection meteorological feature data includes a strong convection radar intensity maximum value f1, a lightning current magnitude f2 and a lightning return attack number f3, where the strong convection radar intensity maximum value f1 is a maximum value in a set time period, the lightning current magnitude f2 is an average value in the set time period, and the radar return attack number f3 is the lightning return attack number in the set time period;
the geographic feature data comprises local geographic features and global geographic features, wherein the local geographic features comprise an elevation l1, a gradient l2 and a slope direction l 3; global geo-feature longitude m1, latitude m 2;
the power grid state characteristic data comprise load times n1, overload times n2 and defect times n 3.
It should be noted that, in this embodiment, according to the description of the weather information, the strong convection weather features may cover most of the information with the maximum intensity of the strong convection radar, the magnitude of the lightning current, and the number of lightning strike-back times.
When the geographic characteristic information is selected, on one hand, local geographic characteristic expressions are considered, including height information, gradient information, slope information and the like. One aspect takes into account global geographic location latitude and longitude information. Considering that the overhead lines in each area are influenced by the historical factors, all parts are the same, and therefore longitude and latitude information needs to be added into the characteristic information to distinguish the overhead lines in different areas.
When the power grid state characteristic data are selected, the stability of the power grid state can be mostly described by considering the stability degree of the overhead line state, including information such as overload times, defect times and the like. If the overload times are too many, the grid loss of the power grid is larger, and the power failure probability of the power grid is relatively increased. These characteristics are associated with overhead line blackouts, as described above.
It should be noted that, in this embodiment, an optimizable sparse decision tree is used to predict the power failure of the overhead line, and the implicit algorithm model has the function of distinguishing the influence degree of different feature dimensions on the power failure state of the overhead line, so that normalization processing on feature data is not required.
Furthermore, in the characteristic information, the weather information changes in real time along with the change of time, and is important for the influence of the power failure prediction result of the overhead line. In this embodiment, a long-time data feature extraction algorithm is designed for meteorological information in an one-hour interval of a certain overhead line, and comprehensive processing is performed to obtain feature information. Specifically, the method comprises the following steps:
(1) intensity maximum of strong convection radar
f1＝max{f1 1 ，f1 2 ...f1 n }
Wherein, f1 1 ，f1 2 ...f1 n The sampling data of the maximum value of the intensity of the strong convection radar of the overhead line in one hour. Considering that the 'maximum value of the intensity of the strong convection radar' has a certain breakdown effect, the maximum value in a period of time is used for representing the maximum value of the intensity of the strong convection radar in 1 hour.
(2) Magnitude of lightning current
f2＝sum{f2 1 ，f2 2 ...f2 n }/n
Wherein, f2 1 ，f2 2 ...f2 n The lightning current of the overhead line is sampled within one hour. Considering that the "lightning current magnitude" has a certain cumulative effect, the average value in a period of time is used to represent the lightning current magnitude in 1 hour.
(3) Number of lightning strike-back
The number of lightning strike-back times is already accumulated data in a period of time, and the number of lightning strike-back times in 1 hour is adopted as the f3 numerical value.
Further, for an overhead line of a certain specific region, the present embodiment is expressed by feature ═ (f1, f2, f3, l1, l2, l3, m1, m2, n1, n2, and n 3). In order to obtain the characteristic information required by the power failure prediction of the overhead lines, the patent collects the big data information of tens of thousands of overhead lines in each area in recent years, including meteorological information, geographic information and power grid state information, and extracts the characteristic information and corresponding labels required by training and verification from the big data information.
The description of the tag information is described in detail below. Considering that the power failure state of the overhead line is a piece of time sequence information, the present embodiment designs the following tag extraction algorithm to obtain the pairing of the feature information and the tag information in a period of time.
Specifically, the step S10 includes the steps of:
and acquiring corresponding characteristic data from a period of time before the power failure signal.
Obtaining the matching of the feature data and the tag information in the time period based on a tag value range, wherein the tag value range is as follows:
it should be noted that the power grid blackout state is considered to be monitored in real time. Once the power grid is powered off, the center console can acquire a power failure signal. Therefore, the power failure state in the big data information is a real-time signal. In this embodiment, corresponding feature information is obtained from the power failure signal y within one hour before the power failure signal y, and a corresponding feature data feature is generated, so that matching between y and feature can be completed.
Further, 100000 data samples are generated for algorithm training and verification through the data acquisition process; in this embodiment, 9: 1, the scale of a training set is 90000 samples, the scale of a testing set is 10000 samples, and the training set and the testing set are respectively used for training and verifying the sparse optimized decision tree.
In one embodiment, in the step S20, the initialized decision tree is:
d＝(d un ，δ un ，d split ，δ split ，K，H)
wherein d is un Leaves not changed in d, δ un In order not to alter the corresponding prediction of the leaf, d split For the leaf to be subdivided in d, δ split For the features corresponding to the leaves to be subdivided, H is the total number of leaves and K is the delimiter.
It should be noted that this embodiment assumes that a tree with H leaves is d ═ (p) 1 ，p 2 ，p 3 ...，p K ，p K+1 ，...，p H ) It means that the tree is subdivided when the first K leaves can not generate a subtree, and the remaining H-K leaves can be subdivided.
Further, it may be rewritten as d ═ d (d) un ，δ un ，d split ，δ split ，K，H)，d un Leaves not changed in d, δ un In order not to alter the corresponding prediction of the leaf, d split For the leaf to be subdivided in d, δ split For the features corresponding to the leaves to be subdivided, H is the total number of leaves and K is the divider.
Based on this, when the leaves are subdivided, a new tree d ' ═ (d ') is obtained ' un ，δ′ un ，d′ split ，δ′ split K ', H'), each new tree is generated from a leaf of the old tree that is reclassified, because in the process of constructing the decision tree, the decision tree is hierarchically constructed according to the input data of each different feature, and the association between the features is embodied in the process of constructing the new tree, but is not embodied in the leaves of the old tree.
In one embodiment, the objective function R (d, x, y) is:
R(d，x，y)＝l(d，x，y)+λH d
where d represents a parameter of the decision tree, x represents input feature data, y represents a label corresponding to the feature data, l (d,x, y) denotes optimization of decision tree parameters directly from tags, λ H d A regularization term is represented.
It should be noted that for each particular decision tree construction, d '═ d' un ，δ′ un ，d′ split ，δ′ split K ', H'), whose objective function can be defined as R (d, x, y) ═ l (d, x, y) + λ H d Wherein, the parameters of the decision tree model are in d, x represents data, y represents a corresponding label, l (d, x, y) represents that the model parameters are directly optimized according to the label, and lambdoh d And a regularization item is represented, so that overfitting of the model in an optimization iteration process is prevented, and optimization iteration can be performed on the model based on the construction of the loss function. When the leaves are added in each step of the decision tree to form a sub-tree, the decision tree is optimized and constructed by adopting a set loss function, so that the performance of the constructed decision tree is further improved.
In one embodiment, the step S40 includes the following steps:
s41, judging the state of the overhead line according to the output confidence of each leaf;
s42, if the state of the overhead line is a power failure state, determining that the leaves are leaves capable of generating a new tree;
and S43, if the state of the overhead line is a normal state, determining that the leaf is a leaf which can not generate a new tree.
Specifically, in consideration of the classification problem, a normal state is assumed when the confidence is higher than 0.5, and a power-off state is assumed when the confidence is lower than 0.5.
It should be noted that the sparse decision tree is trained by using the training data set, so that a strong convection weather overhead line power failure prediction algorithm based on the optimized sparse decision tree is obtained, and a 90% accuracy rate and a 70% recall rate are obtained on a verification set.
Specifically, the decision tree training optimization process in this embodiment is as follows:
inputting: training overhead line training data set D-90000
And (3) outputting: optimizable sparse decision tree
The training process is as follows:
(1) initializing one decision tree d ═ d (d) un ，δ un ，d split ，δ split ，K，H)；
(2) According to the characteristics of the data, feature ═ is (f1, f2, f3, 11, l2, 13, m1, m2, n1, n2, n3), and the depth of the decision tree is 11;
(3) according to input features f i ＝{x i ，y i Calculating R (d, x, y) and an output confidence coefficient of each leaf in d;
(4) after multiple iterations, selecting the leaves capable of generating a new tree according to the output confidence coefficient of each leaf in d;
(5) generating a new tree d ═ (d ') from the new leaves' un ，δ′ un ，d′ split ，δ′ split ，K′，H′)；
(6) Entering the step (3), and performing circulation until all the characteristics are traversed to finally obtain the sparse decision tree d final_model ＝(d final ，H)。
Furthermore, as shown in fig. 5, a second embodiment of the present invention provides an apparatus for predicting a power outage risk of a heavy convection weather overhead line, the apparatus including:
the acquisition module 10 is configured to acquire a training data set, where the training data set includes feature data and data labels corresponding to the feature data, and the training data includes strong convection meteorological feature data, geographic feature data, and power grid state feature data;
an initialization module 20 for initializing a decision tree;
a training module 30, configured to take the training data and the corresponding data labels thereof as inputs of the decision tree, and calculate an objective function R (d, x, y) and an output confidence of each leaf in the decision tree;
a decision tree updating module 40, configured to select a leaf capable of generating a new tree according to the output confidence of each leaf, and generate a new decision tree;
an output module 50, configured to obtain a sparse decision tree until all feature data in the training data set are traversed;
and the prediction module 60 is configured to calculate the power outage probability of the overhead line in the target area by using the sparse decision tree.
In an embodiment, the strong convection meteorological feature data comprise a strong convection radar intensity maximum value, a lightning current magnitude and lightning attack times, wherein the strong convection radar intensity maximum value is a maximum value in a set time period, the lightning current magnitude is an average value in the set time period, and the radar attack times are lightning attack times in the set time period;
the geographic feature data comprises a local geographic feature and a global geographic feature;
the power grid state characteristic data comprise load times, overload times and defect times.
In one embodiment, the initialized decision tree is:
d＝(d un ，δ un ，d split ，δ split ，K，H)
wherein, d un Leaves not changed in d, δ un In order not to alter the corresponding prediction of the leaf, d split For the leaf of d to be subdivided, δ split For the features corresponding to the leaves to be subdivided, H is the total number of leaves and K is the divider.
In one embodiment, the objective function R (d, x, y) is:
R(d，x，y)＝l(d，x，y)+λH d
wherein d represents the parameters of the decision tree, x represents the input feature data, y represents the label corresponding to the feature data, l (d, x, y) represents the optimization of the decision tree parameters directly according to the label, and λ H d A regularization term is represented.
In the embodiment, the strong convection meteorological characteristic data, the geographic characteristic data and the power grid state characteristic data are designed and used for training the optimized sparse decision tree, so that the sparse decision tree has the capability of predicting the power failure probability of the overhead line in strong convection weather according to the input information; the decision tree algorithm has good capability of processing difference of input data dimensions, can well cope with difference in numerical values of factors such as weather factors and geographic factors, and when leaves are added to the decision tree in each step to form a sub-tree, each new tree is generated from a leaf which can be re-divided by an old tree but not from a leaf which cannot be re-divided by the old tree.
It should be noted that, other embodiments or implementation methods of the power outage risk prediction apparatus for heavy convection weather overhead lines according to the present invention can refer to the above-mentioned embodiments, and no redundancy is made herein.
It should be noted that the logic and/or steps shown in the flowcharts or otherwise described herein, such as an ordered listing of executable instructions that can be considered to implement logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium could even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, via for instance optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner if necessary, and then stored in a computer memory.
It should be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or combination of the following techniques, which are known in the art, may be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application specific integrated circuit having an appropriate combinational logic gate circuit, a Programmable Gate Array (PGA), a Field Programmable Gate Array (FPGA), or the like.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise.
Although embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are exemplary and not to be construed as limiting the present invention, and that changes, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. A method for predicting power failure risk of an overhead line in strong convection weather is characterized by comprising the following steps:
acquiring a training data set, wherein the training data set comprises characteristic data and a data label corresponding to the characteristic data, and the training data comprises strong convection meteorological characteristic data, geographical characteristic data and power grid state characteristic data;
initializing a decision tree;
taking the training data and the corresponding data labels thereof as the input of the decision tree, and calculating the target function R (d, x, y) and the output confidence of each leaf in the decision tree;
selecting leaves capable of generating a new tree according to the output confidence of each leaf to generate a new decision tree;
until all the characteristic data in the training data set are traversed, obtaining a sparse decision tree;
and calculating the power failure probability of the overhead line in the target area by using the sparse decision tree.
2. The strong convection weather overhead line power failure risk prediction method of claim 1, wherein the strong convection weather feature data comprises a maximum strong convection radar intensity value, a lightning current value and lightning return attack times, wherein the maximum strong convection radar intensity value is a maximum value within a set time period, the lightning current value is an average value within the set time period, and the radar return attack times are lightning return attack times within the set time period;
the geographic feature data comprises a local geographic feature and a global geographic feature;
the power grid state characteristic data comprise load times, overload times and defect times.
3. The strong convection weather overhead line blackout risk prediction method of claim 2, wherein the obtaining the training set comprises:
acquiring corresponding characteristic data from a period of time before a power failure signal;
obtaining the matching of the feature data and the tag information in the time period based on a tag value range, wherein the tag value range is as follows:
4. the strong convection weather overhead line power failure risk prediction method of claim 1, wherein the initialized decision tree is:
d＝(d un ，δ un ，d split， δ split， K，H)
wherein d is un Leaves not changed in d, δ un To not change the corresponding prediction of the leaf, d split For the leaf to be subdivided in d, δ split For the features corresponding to the leaves to be subdivided, H is the total number of leaves and K is the delimiter.
5. The strong convection weather overhead line blackout risk prediction method of claim 1, wherein the objective function R (d, x, y) is:
wherein d represents a parameter of the decision tree, x represents input feature data, y represents a label corresponding to the feature data,
denotes optimization of decision Tree parameters directly from tags, λ H d A regularization term is represented.
6. The method for predicting the power failure risk of the heavy convection weather overhead line according to claim 1, wherein the selecting the leaves capable of generating the new tree according to the output confidence of each leaf to generate the new decision tree comprises:
judging the state of the overhead line according to the output confidence of each leaf;
if the state of the overhead line is a power failure state, determining that the leaves are leaves capable of generating a new tree;
and if the state of the overhead line is a normal state, determining that the leaf is a leaf which can not generate a new tree.
7. A strong convection weather overhead line power failure risk prediction device, its characterized in that, the device includes:
the system comprises an acquisition module, a comparison module and a comparison module, wherein the acquisition module is used for acquiring a training data set, the training data set comprises characteristic data and a data label corresponding to the characteristic data, and the training data comprises strong convection meteorological characteristic data, geographic characteristic data and power grid state characteristic data;
the initialization module is used for initializing the decision tree;
the training module is used for taking the training data and the corresponding data labels thereof as the input of the decision tree, and calculating the target function R (d, x, y) and the output confidence of each leaf in the decision tree;
the decision tree updating module is used for selecting the leaves capable of generating the new tree according to the output confidence of each leaf to generate the new decision tree;
the output module is used for obtaining a sparse decision tree until all the characteristic data in the training data set are traversed;
and the prediction module is used for calculating the power failure probability of the overhead line in the target area by utilizing the sparse decision tree.
8. The strong convection weather overhead line power failure risk prediction device of claim 7, wherein the strong convection weather characteristic data comprises a strong convection radar intensity maximum value, a lightning current magnitude and a lightning return attack number, wherein the strong convection radar intensity maximum value is a maximum value within a set time period, the lightning current magnitude is an average value within the set time period, and the radar return attack number is a lightning return attack number within the set time period;
the geographic feature data comprises a local geographic feature and a global geographic feature;
the power grid state characteristic data comprise load times, overload times and defect times.
9. The strong convection weather overhead line blackout risk prediction apparatus of claim 7, wherein the initialized decision tree is:
d＝(d un ，δ un ，d split ，δ split ，K，H)
wherein d is un Leaves not altered in d, delta un In order not to alter the corresponding prediction of the leaf, d split For the leaf to be subdivided in d, δ split For the features corresponding to the leaves to be subdivided, H is the total number of leaves and K is the divider.
10. The strong convection weather overhead line blackout risk prediction apparatus of claim 7, wherein the objective function R (d, x, y) is:
wherein d represents a parameter of the decision tree, x represents input feature data, y represents a label corresponding to the feature data,
denotes optimization of decision Tree parameters directly from tags, λ H d A regularization term is represented.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210577256.3A CN115034586A (en) | 2022-05-25 | 2022-05-25 | Method and device for predicting power failure risk of overhead line in strong convection weather |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210577256.3A CN115034586A (en) | 2022-05-25 | 2022-05-25 | Method and device for predicting power failure risk of overhead line in strong convection weather |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115034586A true CN115034586A (en) | 2022-09-09 |
Family
ID=83120630
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210577256.3A Pending CN115034586A (en) | 2022-05-25 | 2022-05-25 | Method and device for predicting power failure risk of overhead line in strong convection weather |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115034586A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117217546A (en) * | 2023-11-08 | 2023-12-12 | 合肥工业大学 | Power transmission line lightning trip prediction model, method, system and storage medium |
-
2022
- 2022-05-25 CN CN202210577256.3A patent/CN115034586A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117217546A (en) * | 2023-11-08 | 2023-12-12 | 合肥工业大学 | Power transmission line lightning trip prediction model, method, system and storage medium |
| CN117217546B (en) * | 2023-11-08 | 2024-01-12 | 合肥工业大学 | Power transmission line lightning trip prediction model, method, system and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kankanala et al. | AdaBoost $^{+} $: An Ensemble Learning Approach for Estimating Weather-Related Outages in Distribution Systems | |
| CN107506868B (en) | Method and device for predicting short-time power load | |
| CN108133225A (en) | A kind of icing flashover fault early warning method based on support vector machines | |
| CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
| CN110969290A (en) | Runoff probability prediction method and system based on deep learning | |
| CN111178585A (en) | Fault reporting amount prediction method based on multi-algorithm model fusion | |
| Hou et al. | Review of failure risk and outage prediction in power system under wind hazards✰ | |
| CN115220133B (en) | Rainfall prediction method, device and equipment for multiple meteorological elements and storage medium | |
| Hou et al. | Spatial distribution assessment of power outage under typhoon disasters | |
| Li et al. | Prediction of power outage quantity of distribution network users under typhoon disaster based on random forest and important variables | |
| Yeh et al. | Real-time leakage current classification of 15kV and 25kV distribution insulators based on bidirectional long short-term memory networks with deep learning machine | |
| Hou et al. | Damage prediction of 10 kV power towers in distribution network under typhoon disaster based on data-driven model | |
| Piotrowski et al. | Baczy nski | |
| CN115034586A (en) | Method and device for predicting power failure risk of overhead line in strong convection weather | |
| US20230324860A1 (en) | Estimating Energy Consumption for a Building Using Dilated Convolutional Neural Networks | |
| Afrasiabi et al. | Probability density function forecasting of electricity price: Deep gabor convolutional mixture network | |
| CN117674119A (en) | Power grid operation risk assessment method, device, computer equipment and storage medium | |
| Zhang et al. | Extreme wind power forecast error analysis considering its application in day‐ahead reserve capacity planning | |
| CN117371207A (en) | Extra-high voltage converter valve state evaluation method, medium and system | |
| JP2003090887A (en) | Predication system and prediction method of instantaneous voltage drop by thunderbolt | |
| Tan et al. | Projected changes of typhoon intensity in a regional climate model: Development of a machine learning bias correction scheme | |
| Ferraz et al. | Analysis of distribution system considering load uncertainties: a k-means clustering approach | |
| CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
| Li et al. | Short-term probabilistic load forecasting method based on uncertainty estimation and deep learning model considering meteorological factors | |
| Kor et al. | Predicting weather-related power outages in distribution grid |
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 |