A method for predicting tree growth on a power transmission line
By expanding and processing the data on tree obstruction hazards along power transmission lines and conducting multiple rounds of collaborative training, high-quality sample data is generated, solving the problem of low accuracy in tree obstruction prediction in existing technologies and achieving efficient tree height growth prediction.
Patent Information
- Application Number
- CN202211089766.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-07
- Publication Date
- 2026-02-27
- Estimated Expiration
- 2042-09-07
AI Technical Summary
Existing technologies for predicting tree obstructions around transmission lines suffer from high workload, difficult calculations, and large errors due to manual inspections. Insufficient training data for machine learning models also leads to low prediction accuracy, making it difficult to meet the needs of diverse tree obstructions and growth patterns around transmission lines.
By acquiring sample data of potential tree obstructions along power transmission lines and performing primary and secondary expansion processing, and using cyclic prediction and backward inference models for multi-round collaborative training, high-quality sample data is generated, thereby improving model training performance.
This improved the prediction accuracy of the tree height growth model, reduced the need for initial sample data, and enhanced the overall performance of the prediction model.
Smart Images
Figure CN116304657B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of tree growth prediction, and in particular to a power transmission line tree barrier growth prediction method. BACKGROUND
[0002] The vegetation in tropical regions grows fast, and there are many plant species. When the trees near the power transmission line grow too high, the swaying of the branches and leaves of the trees caused by strong winds may touch the conductor and cause electric leakage, or even break the conductor. When the trees under the power transmission line grow too high, they may also touch the conductor and cause electric leakage, open circuit, or even cause a power outage. One of the difficulties in tree barrier treatment is to estimate the distance from the trees to the power line. In order to solve the problems of large workload, difficulty in manual distance measurement, and difficulty in avoiding errors in the traditional manual inspection of the power transmission line, the current industry uses a machine learning model to predict the growth height of the trees, thereby assisting the operation and maintenance unit in preparing a tree barrier cleaning plan and reducing the risk of tree barriers to the safe operation of the line. However, there are still some problems, such as the need for a large amount of tree barrier hidden danger and growth training data to meet the coverage requirements of the power transmission line and the growth diversity of different tree species. It is time-consuming and laborious to collect sufficient training data from the natural environment, and the training data obtained from the network may not meet the requirements, which increases the difficulty of training the prediction model and affects the prediction accuracy. SUMMARY
[0003] In view of this, the purpose of the present application is to provide a power transmission line tree barrier growth prediction method to overcome or at least partially solve the above-mentioned problems existing in the prior art.
[0004] To achieve the above-mentioned application purpose, the present application provides a power transmission line tree barrier growth prediction method, which comprises the following steps:
[0005] S101, acquiring power transmission line tree barrier hidden danger sample data;
[0006] S102, performing a one-time expansion processing on the power transmission line tree barrier hidden danger sample data to obtain one-time expansion sample data;
[0007] S103, inputting the one-time expansion sample data into a recurrent prediction model for a two-time expansion processing to obtain two-time expansion sample data;
[0008] S104, training a tree height growth model through the two-time expansion sample data;
[0009] S105, inputting real-time collected power transmission line tree barrier hidden danger data into the tree height growth model to obtain tree height prediction data.
[0010] Further, the transmission line tree barrier hidden danger sample data includes tree species, tree age, canopy density, plant quantity, current tree height, current diameter at breast height, current horizontal distance and vertical distance from the conductor, and growth rate.
[0011] Further, the transmission line tree barrier hidden danger sample data is processed once, specifically including:
[0012] S201, selecting an expansion strategy from the expansion strategy set;
[0013] S202, performing one expansion processing on the transmission line tree barrier hidden danger sample data according to the selected expansion strategy.
[0014] Further, an expansion strategy is selected from the expansion strategy set, specifically including the following steps:
[0015] S301, obtaining a transmission line section information list, and randomly selecting a section as a target section from the transmission line section information list;
[0016] S302, obtaining historical tree barrier removal data of the target section, and selecting an expansion strategy from the expansion strategy set based on the historical tree barrier removal data.
[0017] Further, an expansion strategy is selected from the expansion strategy set based on the historical tree barrier removal data, specifically including the following steps:
[0018] S401, generating a historical tree barrier removal data change curve graph according to the historical tree barrier removal data;
[0019] S402, calculating the average curvature of the curve in the historical tree barrier removal data change curve graph;
[0020] S403, classifying the average curvature, and selecting an expansion strategy from the expansion strategy set according to the level to which the average curvature belongs.
[0021] Further, the once-expanded sample data is input into the cyclic prediction model for secondary expansion processing, specifically including the following steps:
[0022] S501, creating a prediction model and a backstepping model for composing the cyclic prediction model, setting the target of the prediction model to predict tree barrier growth data according to the input tree barrier hidden danger data, and setting the target of the backstepping model to backstepping tree barrier hidden danger data according to the input tree barrier growth data;
[0023] S502, performing multiple rounds of collaborative training on the prediction model and the backstepping model through the once-expanded sample data, generating more accurate tree barrier hidden danger data and tree barrier growth data in each round, and obtaining secondary expansion sample data.
[0024] Further, the step S502 specifically includes:
[0025] S601, extract corresponding tree barrier hidden danger data and tree barrier growth data in the first extended sample data as a set of sample data with solutions;
[0026] S602, pre-train the prediction model and the back-propagation model through the set of sample data with solutions;
[0027] S603, extract tree barrier hidden danger data in the first extended sample data to form a tree barrier hidden danger sample, and extract tree barrier growth data therein to form a tree barrier growth sample;
[0028] S604, cooperatively train the prediction model and the back-propagation model through the tree barrier hidden danger sample and the tree barrier growth sample.
[0029] Further, the step S604 specifically comprises:
[0030] S701, in the first round of cooperative training, input the tree barrier hidden danger sample into the prediction model to obtain tree barrier growth data corresponding to the tree barrier hidden danger sample, and add the tree barrier hidden danger sample and the corresponding tree barrier growth data to a first training data set;
[0031] S702, in the first round of cooperative training, input the tree barrier growth sample into the back-propagation model to obtain tree barrier hidden danger data corresponding to the tree barrier growth sample, and add the tree barrier growth sample and the corresponding tree barrier hidden danger data to a second training data set;
[0032] S703, in each subsequent round of cooperative training, use the first training data set and the second training data set output in the last round of cooperative training to train the prediction model and the back-propagation model respectively, and update parameters of the prediction model and the back-propagation model.
[0033] Compared with the prior art, the present application has the following advantages:
[0034] The method for predicting tree barrier growth of a power transmission line provided by the present application, after obtaining tree barrier hidden danger sample data of the power transmission line, first performs first extension processing on the tree barrier hidden danger sample data of the power transmission line to obtain first extended sample data, then inputs the first extended sample data into a recurrent prediction model to perform second extension processing and obtain second extended sample data, and trains a tree height growth model through the second extended sample data, so that the tree height growth model can be trained to have higher performance using less initial sample data, and the prediction accuracy of the tree height growth model is improved. BRIEF DESCRIPTION OF DRAWINGS
[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only are the preferred embodiments of the present application, and for those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.
[0036] Figure 1 is a schematic diagram of the whole process of the transmission line tree barrier growth prediction method provided by the embodiments of the present application.
[0037] Figure 2 is a schematic diagram of the primary expansion processing process provided by the embodiments of the present application.
[0038] Figure 3 is a schematic diagram of the expansion strategy selection process provided by the embodiments of the present application.
[0039] Figure 4 is a schematic diagram of the expansion strategy selection process based on historical tree barrier removal data provided by the embodiments of the present application.
[0040] Figure 5 is a schematic diagram of the secondary expansion processing process provided by the embodiments of the present application.
[0041] Figure 6 is a schematic diagram of the primary expansion sample data processing process provided by the embodiments of the present application.
[0042] Figure 7 is a schematic diagram of the cooperative training process provided by the embodiments of the present application. DETAILED DESCRIPTION
[0043] The principles and features of the present application are described below in combination with the drawings, and the listed embodiments are only used to explain the present application, and are not used to limit the scope of the present application.
[0044] With reference to Figure 1 The present embodiment provides a transmission line tree barrier growth prediction method, which comprises the following steps:
[0045] S101, obtaining transmission line tree barrier hidden danger sample data.
[0046] In the present embodiment, the transmission line tree barrier hidden danger sample data comprises tree species, tree age, canopy density, plant quantity, current tree height, current diameter at breast height, current horizontal distance and vertical distance from the conductor, and growth rate.
[0047] S102, performing primary expansion processing on the transmission line tree barrier hidden danger sample data to obtain primary expansion sample data.
[0048] S103, input the first extended sample data into the cycle prediction model for secondary expansion processing to obtain secondary extended sample data.
[0049] S104, train the tree height growth model by using the secondary extended sample data.
[0050] S105, input the real-time collected transmission line tree barrier hidden danger data into the tree height growth model to obtain tree height prediction data.
[0051] The method increases the amount of high-quality sample data for subsequent model training by performing first and second expansion processing on the transmission line tree barrier hidden danger sample data. Training the model with sufficient and high-quality training data helps improve the overall performance of the model.
[0052] Reference Figure 2 The first expansion processing of the transmission line tree barrier hidden danger sample data includes the following steps:
[0053] S201, select an expansion strategy from the expansion strategy set. The expansion strategy set stores multiple expansion strategies, and different expansion strategies are used to expand and enhance the data in different ways to increase the amount of data.
[0054] S202, perform first expansion processing on the transmission line tree barrier hidden danger sample data according to the selected expansion strategy.
[0055] Reference Figure 3 In conventional data enhancement processing, the same data enhancement method is used to process sample data, which will result in the same trend of the expanded sample data, making it difficult to meet the data diversity requirement. Therefore, this embodiment randomly selects an expansion strategy to perform first expansion processing on the transmission line tree barrier hidden danger sample data, which can improve the diversity of the expanded sample data when multiple rounds of sample data are collected. Selecting an expansion strategy from the expansion strategy set includes the following steps:
[0056] S301, obtain a transmission line section information list, and randomly select a section as a target section from the transmission line section information list. In this embodiment, the entire transmission line to be operated and maintained is divided into multiple sections in advance, and the transmission line section information list stores the unique identification information of each section in advance.
[0057] S302, obtain the historical tree barrier removal data of the target section, and select an expansion strategy from the expansion strategy set based on the historical tree barrier removal data. For example, the historical tree barrier removal data can be historical tree barrier removal plant quantity data or branch length removal data.
[0058] Reference Figure 4, selecting an expansion strategy from the expansion strategy set based on the historical tree barrier clearing data, specifically comprising the following steps:
[0059] S401, generating a historical tree barrier clearing data change curve diagram according to the historical tree barrier clearing data.
[0060] S402, calculating the average curvature of the curve in the historical tree barrier clearing data change curve diagram.
[0061] S403, classifying the average curvature, and selecting an expansion strategy from the expansion strategy set according to the level to which the average curvature belongs.
[0062] In this embodiment, a plurality of levels are divided according to the curvature value range, and the level of the historical tree barrier clearing data change curve average curvature is determined, different levels correspond to different expansion strategies in the expansion strategy set, and the corresponding expansion strategy is selected through the two variables of section and historical data change trend to realize randomization. For example, the historical tree barrier clearing data used to select the expansion strategy each time should be updated compared with the data used to select the expansion strategy last time, or different historical period data is used.
[0063] Referring to Figure 5 , the one-time expansion sample data is input into the cyclic prediction model for secondary expansion processing, specifically comprising the following steps:
[0064] S501, creating a prediction model and an inverse model for composing a cyclic prediction model, setting the target of the prediction model to predict tree barrier growth data according to the input tree barrier hidden danger data, and setting the target of the inverse model to inverse the tree barrier hidden danger data according to the input tree barrier growth data. The tree barrier hidden danger data is of the same type as the data in the tree barrier hidden danger sample data. The tree barrier growth data is the current horizontal distance and vertical distance of the tree barrier from the conductor after a certain period of time.
[0065] S502, performing multi-round collaborative training on the prediction model and the inverse model through the one-time expansion sample data, generating more accurate tree barrier hidden danger data and tree barrier growth data in each multi-round, and obtaining secondary expansion sample data.
[0066] Specifically, referring to Figure 6 , the step S502 specifically comprises:
[0067] S601, extracting the corresponding tree barrier hidden danger data and tree barrier growth data in the one-time expansion sample data as a solution sample data set. The corresponding tree barrier hidden danger data and tree barrier growth data means that the corresponding tree barrier growth data can be obtained from the tree barrier hidden danger data, and this association can be pre-specified by the user or obtained in actual tree barrier clearing work.
[0068] S602, pre-training the prediction model and the reverse model by using the sample data set with solutions.
[0069] S603, extracting tree barrier hidden danger data in the once extended sample data to form a tree barrier hidden danger sample, and extracting tree barrier growth data to form a tree barrier growth sample.
[0070] S604, co-training the prediction model and the reverse model by using the tree barrier hidden danger sample and the tree barrier growth sample.
[0071] Referring to Figure 7 , the step S604 specifically comprises:
[0072] S701, in the first round of co-training, inputting the tree barrier hidden danger sample into the prediction model to obtain tree barrier growth data corresponding to the tree barrier hidden danger sample, and adding the tree barrier hidden danger sample and the corresponding tree barrier growth data to the first training data set.
[0073] S702, in the first round of co-training, inputting the tree barrier growth sample into the reverse model to obtain tree barrier hidden danger data corresponding to the tree barrier growth sample, and adding the tree barrier growth sample and the corresponding tree barrier hidden danger data to the second training data set.
[0074] S703, in each subsequent round of co-training, using the first training data set and the second training data set output in the last round of co-training to train the prediction model and the reverse model respectively, and updating the parameters of the prediction model and the reverse model, the performance of the model and the quantity and quality of the training data set are improved in each round of co-training, and the co-training is stopped after the model converges.
[0075] The method provided by the embodiment pre-trains the prediction model and the reverse model by using the sample data set with solutions, and then co-trains the prediction model and the reverse model by using the tree barrier hidden danger sample and the tree barrier growth sample respectively, and constantly updates the training data in multiple rounds of co-training, thereby improving the quantity and quality of the training data and the performance of the model, so that the method can fully train the model using less initial training data.
[0076] The above only describes the preferred embodiments of the present application, and is not intended to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A method for predicting tree growth along power transmission lines, characterized in that, The method comprises the following steps: S101, acquiring power transmission line tree barrier hidden danger sample data; S102, performing one-time extension processing on the power transmission line tree barrier hidden danger sample data to obtain one-time extension sample data; S103, inputting the one-time extension sample data into a cyclic prediction model for two-time extension processing to obtain two-time extension sample data; S104, training a tree height growth model through the two-time extension sample data; S105, inputting real-time collected power transmission line tree barrier hidden danger data into the tree height growth model to obtain tree height prediction data; The one-time extension processing on the power transmission line tree barrier hidden danger sample data specifically comprises: S201, selecting an extension strategy from an extension strategy set; S202, performing one-time extension processing on the power transmission line tree barrier hidden danger sample data according to the selected extension strategy; The two-time extension processing of the one-time extension sample data in the cyclic prediction model specifically comprises the following steps: S501, creating a prediction model and a backstepping model for composing the cyclic prediction model, setting the target of the prediction model as predicting tree barrier growth data according to input tree barrier hidden danger data, and setting the target of the backstepping model as backstepping tree barrier hidden danger data according to input tree barrier growth data; S502, performing multi-round collaborative training on the prediction model and the backstepping model through the one-time extension sample data, generating more accurate tree barrier hidden danger data and tree barrier growth data in each multi-round, and obtaining two-time extension sample data.
2. A method of predicting tree growth on a power line according to claim 1, wherein, The power transmission line tree barrier hidden danger sample data comprises tree species, tree age, canopy density, plant quantity, current tree height, current diameter at breast height, current horizontal distance and vertical distance from the conductor, and growth rate.
3. A method of predicting tree growth on a power line according to claim 1, wherein The selecting of the extension strategy from the extension strategy set specifically comprises the following steps: S301, acquiring a power transmission line section information list, and randomly selecting a section as a target section from the power transmission line section information list; S302, acquiring historical tree barrier removal data of the target section, and selecting an extension strategy from the extension strategy set based on the historical tree barrier removal data.
4. A method of predicting tree growth on a power line according to claim 3, wherein, The selecting of the extension strategy from the extension strategy set based on the historical tree barrier removal data specifically comprises the following steps: S401, generating a historical tree barrier removal data change curve graph according to the historical tree barrier removal data; S402, calculating the average curvature of the curve in the historical tree barrier removal data change curve graph; S403, grading the average curvature, and selecting an extension strategy from the extension strategy set according to the grade to which the average curvature belongs.
5. A method of predicting tree growth on a power line according to claim 1, wherein, The two-time extension processing of the one-time extension sample data in the cyclic prediction model specifically comprises the following steps: S501, creating a prediction model and a backstepping model for composing the cyclic prediction model, setting the target of the prediction model as predicting tree barrier growth data according to input tree barrier hidden danger data, and setting the target of the backstepping model as backstepping tree barrier hidden danger data according to input tree barrier growth data; S502, performing multi-round collaborative training on the prediction model and the backstepping model through the one-time extension sample data, generating more accurate tree barrier hidden danger data and tree barrier growth data in each multi-round, and obtaining two-time extension sample data.
6. A method of predicting tree growth on a power line according to claim 1, wherein The step S502 specifically comprises: S601, extract corresponding tree barrier hidden danger data and tree barrier growth data in the first extended sample data as the solved sample data set; S602, pre-train the prediction model and the back-propagation model through the solved sample data set; S603, extract the tree barrier hidden danger data in the first extended sample data to form a tree barrier hidden danger sample, and extract the tree barrier growth data to form a tree barrier growth sample; S604, cooperatively train the prediction model and the back-propagation model through the tree barrier hidden danger sample and the tree barrier growth sample.
7. A method of predicting tree growth on a power line according to claim 6, wherein, The step S604 specifically comprises: S701, in the first round of cooperative training, input the tree barrier hidden danger sample into the prediction model to obtain the tree barrier growth data corresponding to the tree barrier hidden danger sample, and add the tree barrier hidden danger sample and the corresponding tree barrier growth data to the first training data set; S702, in the first round of cooperative training, input the tree barrier growth sample into the back-propagation model to obtain the tree barrier hidden danger data corresponding to the tree barrier growth sample, and add the tree barrier growth sample and the corresponding tree barrier hidden danger data to the second training data set; S703, in each subsequent round of cooperative training, use the first training data set and the second training data set output in the last round of cooperative training to train the prediction model and the back-propagation model respectively, and update the parameters of the prediction model and the back-propagation model.
Citation Information
Patent Citations
Tree obstacle removing method, device and system based on unmanned aerial vehicle
CN110705526A
Method for predicting tree growth hidden dangers in power transmission line corridor area
CN111931976A