Distribution line icing quality prediction method and device based on neural network, equipment and medium
By using a neural network-based approach and combining conductor characteristics and environmental parameters, an icing prediction model was constructed. This solved the problem that conductor characteristics and environmental changes have a significant impact on existing technologies, and achieved high-precision prediction and adaptive improvement of icing quality.
Patent Information
- Application Number
- CN202511573929.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-30
- Publication Date
- 2026-02-03
AI Technical Summary
Existing icing prediction methods fail to fully consider the influence of conductor characteristic parameters and environmental changes, resulting in large prediction deviations, poor adaptability, and inability to meet the needs of accurate early warning.
A neural network-based approach was adopted to construct a training dataset by determining the characteristic parameters of the conductor, environmental climate parameters, and geographical topography parameters, and then performing fitting training. The neural network model based on transfer learning was used to predict icing quality, thereby reducing the impact of conductor and environmental changes.
It improves the accuracy and generalization ability of icing prediction, enhances adaptability in complex environments, and meets the needs of precise early warning.
Smart Images
Figure CN121457700A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, equipment, and medium for predicting the icing quality of power distribution lines based on neural networks. Background Technology
[0002] Icing on power distribution lines is a major natural hazard threatening the safe operation of the power grid in winter. Under icing conditions, conductors are prone to ice shedding, jumping, and galloping, which can lead to short circuits or even serious accidents such as line breaks and tower collapses, causing widespread power outages and posing a serious threat to the safe and stable operation of the power system. Therefore, accurately predicting the quality of icing on power distribution lines before it develops into an ice disaster, thereby understanding the development trend of icing and rationally planning de-icing and removal work, has crucial engineering value.
[0003] Currently, existing icing prediction methods largely rely on monitoring and physical modeling of environmental atmospheric parameters. However, these methods typically consider a limited number of parameters and fail to adequately account for the impact of conductor characteristics (such as conductor thickness, surface hydrophobicity, power distribution type, and appearance) on the icing process and quality. Different meteorological conditions can lead to different types of icing, such as rime and hoarfrost, with varying densities, adhesion, and degrees of damage to power lines. Existing technologies often fail to differentiate and model icing types, resulting in discrepancies between predicted and actual icing weights. Models based on purely physical theories are poorly adaptable to complex and variable real-world environments. Their measurement and prediction accuracy is easily affected by various environmental factors and dynamic changes in conductors under icing conditions, leading to low prediction reliability and failing to meet the needs of accurate early warning.
[0004] As can be seen from the above, how to reduce the impact of the characteristics of the conductor itself and environmental changes on icing prediction, and improve the generalization ability and adaptability of icing prediction, is an urgent problem to be solved. Summary of the Invention
[0005] In view of this, the purpose of this invention is to provide a method, apparatus, device, and medium for predicting icing quality of power distribution lines based on neural networks, which can reduce the impact of conductor characteristics and environmental changes on icing prediction, and improve the generalization ability and adaptability of icing prediction. The specific solution is as follows:
[0006] Firstly, this application provides a method for predicting the icing quality of power distribution lines based on neural networks, including:
[0007] The target parameters affecting icing conditions are determined, and data acquisition operations are performed based on the target parameters to obtain icing data corresponding to the target parameters. The target parameters include conductor characteristic parameters, environmental climate parameters, and geographical topographic parameters. The icing data corresponding to the target parameters include icing type and icing mass change data of the icing type per unit time. The icing type includes rime ice type and hoarfrost type.
[0008] A training dataset is constructed based on the target parameters and the corresponding icing data, and the preset neural network model is fitted and trained using the training dataset to obtain the target neural network model.
[0009] The target neural network model is transferred to the target prediction area using icing data collected in the target prediction area, and the transferred-learned target neural network model is used to predict the icing quality of the power distribution lines in the target prediction area to obtain the corresponding prediction results.
[0010] Optionally, the conductor characteristic parameters are parameters generated based on conductor thickness, conductor surface hydrophobicity, power distribution type, power distribution voltage, and conductor appearance; the environmental climate parameters include ambient temperature, ambient wind speed, liquid water content, and median volume diameter of water droplets; the geographical topography parameters include elevation data of the power distribution tower and height difference data between towers.
[0011] Optionally, the training dataset consists of sample point data under several different geographical and climatic conditions, and each sample point data contains several sets of icing data sequences collected over time.
[0012] Optionally, the target neural network model consists of an input layer, a hidden layer, and an output layer; the input of the target neural network model includes fixed conductor feature parameters, altitude data, and elevation difference data, as well as environmental temperature, environmental wind speed, liquid water content, median volume diameter of water droplets, and initial icing mass that change over time; the output of the target neural network model is the change in mass of rime and the change in mass of hoarfrost per unit time.
[0013] Optionally, the step of using the training dataset to fit and train a preset neural network model to obtain a target neural network model includes:
[0014] The preset neural network model is fitted and trained using the training dataset to obtain a prediction function between the target parameters and the ice type and the change in ice mass.
[0015] The model parameters of the preset neural network model are adjusted based on the prediction function to obtain the target neural network model.
[0016] Optionally, the neural network-based method for predicting the icing quality of power distribution lines further includes:
[0017] Based on the laws of climate change, the range of variation and the accuracy of the values of the environmental climate parameters are set, and the range of variation and the accuracy of the values are used as the boundary constraints of the prediction function.
[0018] Optionally, the transfer learning of the target neural network model using icing data collected from the target prediction area includes:
[0019] Icing data of the target prediction area is collected based on a preset data collection cycle, and a regional icing dataset is constructed based on the icing data.
[0020] The model parameters of the target neural network model are adjusted using the regional icing dataset to complete the transfer learning of the target neural network model.
[0021] Secondly, this application provides a neural network-based device for predicting the icing quality of power distribution lines, comprising:
[0022] The data acquisition module is used to determine the target parameters affecting icing conditions and to perform data acquisition operations based on the target parameters to obtain icing data corresponding to the target parameters. The target parameters include conductor characteristic parameters, environmental climate parameters, and geographical topography parameters. The icing data corresponding to the target parameters includes icing type and the icing mass change data of the icing type per unit time. The icing type includes rime ice type and hoarfrost type.
[0023] The model building module is used to build a training dataset based on the target parameters and the icing data corresponding to the target parameters, and to use the training dataset to fit and train a preset neural network model to obtain the target neural network model.
[0024] The prediction module is used to perform transfer learning on the target neural network model using icing data collected in the target prediction area, and to use the target neural network model after transfer learning to predict the icing quality of the power distribution lines in the target prediction area, so as to obtain the corresponding prediction results.
[0025] Thirdly, this application provides an electronic device, comprising:
[0026] Memory, used to store computer programs;
[0027] A processor is used to execute the computer program to implement the aforementioned neural network-based method for predicting the icing quality of power distribution lines.
[0028] Fourthly, this application provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned neural network-based method for predicting the icing quality of power distribution lines.
[0029] This application provides a neural network-based method for predicting the icing quality of power distribution lines. First, target parameters affecting icing are determined, and data collection is performed based on these parameters to obtain corresponding icing data. The target parameters include conductor characteristic parameters, environmental climate parameters, and geographical topography parameters. The icing data corresponding to the target parameters includes icing type and the icing quality change data of each icing type per unit time. The icing types include rime ice and hoarfrost. Then, a training dataset is constructed based on the target parameters and the corresponding icing data. A preset neural network model is fitted and trained using this training dataset to obtain a target neural network model. Finally, icing data collected in the target prediction area is used to perform transfer learning on the target neural network model, and the transferred-learned target neural network model is used to predict the icing quality of power distribution lines in the target prediction area to obtain the corresponding prediction results.
[0030] As can be seen from the above, this application starts with the icing type of atmospheric structures, and improves the accuracy of power line icing prediction by utilizing real-time atmospheric parameters of the freezing environment and conductor characteristic parameters. Referring to the past variation patterns of atmospheric parameters in the freezing environment, a neural network algorithm is used to improve prediction accuracy, achieving prediction of power line icing quality in freezing environments. This reduces the impact of conductor characteristics and environmental changes on icing prediction, and improves the generalization ability and adaptability of icing prediction. Attached Figure Description
[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0032] Figure 1 This is a flowchart of a neural network-based method for predicting the icing quality of power distribution lines disclosed in this application.
[0033] Figure 2 This is a schematic diagram of a neural network model architecture disclosed in this application;
[0034] Figure 3 This is a schematic diagram of a neural network-based power distribution line icing quality prediction device disclosed in this application.
[0035] Figure 4 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] Icing on power distribution lines is a major natural hazard threatening the safe operation of the power grid in winter. Under icing conditions, conductors are prone to ice shedding, jumping, and galloping, which can lead to short circuits or even serious accidents such as line breaks and tower collapses, causing widespread power outages and posing a serious threat to the safe and stable operation of the power system. Therefore, accurately predicting the quality of icing on power distribution lines before it develops into an ice disaster, thereby understanding the development trend of icing and rationally planning de-icing and removal work, has crucial engineering value.
[0038] Currently, existing icing prediction methods largely rely on monitoring and physical modeling of environmental atmospheric parameters. However, these methods typically consider a limited number of parameters, failing to adequately account for the impact of conductor characteristics (such as conductor thickness, surface hydrophobicity, distribution type, and appearance) on the icing process and quality. Different meteorological conditions can lead to different types of icing, such as rime and hoarfrost, with varying densities, adhesion, and degrees of damage to power lines. Existing technologies often fail to differentiate and model icing types, resulting in discrepancies between predicted and actual icing weights. Models based on purely physical theories are poorly adaptable to complex and variable real-world environments, and their measurement and prediction accuracy is easily affected by various environmental factors and conductor dynamics under icing conditions, leading to low prediction reliability and failing to meet the need for accurate early warning. Therefore, this application provides a neural network-based icing quality prediction scheme for power distribution lines, which can reduce the impact of conductor characteristics and environmental changes on icing prediction, improving the generalization ability and adaptability of icing prediction.
[0039] See Figure 1 As shown in the figure, this application discloses a method for predicting the icing quality of power distribution lines based on neural networks, including:
[0040] Step S11: Determine the target parameters that affect icing conditions, and perform data acquisition operations based on the target parameters to obtain icing data corresponding to the target parameters.
[0041] In this embodiment, the power distribution line icing quality prediction model requires several environmental parameters, including conductor thickness, conductor surface hydrophobicity, conductor icing characteristic parameters influenced by power distribution type and voltage, tower height difference and altitude, ambient temperature, ambient wind speed, liquid water content, median volume diameter of water droplets, and aerodynamic viscosity coefficient. Since some parameters have a relatively small impact on the prediction results, simplifications can be made in some cases. For example, the aerodynamic viscosity coefficient can be approximated as a constant value if the boundary layer qualitative temperature change caused by the power distribution line temperature rise is not considered. For example, in some specific embodiments, the target parameters include conductor characteristic parameters k, environmental climate parameters, and geographical topography parameters; the icing data corresponding to the target parameters includes icing type and the icing mass change data of the icing type per unit time; the icing type includes rime ice type and hoarfrost type; the conductor characteristic parameter k is a parameter generated based on conductor thickness r, conductor surface hydrophobicity a, power distribution type b, power distribution voltage U, and conductor appearance c; the environmental climate parameters include the ambient temperature T(t), ambient wind speed v(t), liquid water content w(t), and median volume diameter of water droplets a(t) at the current time t; the geographical topography parameters include the elevation data H of the power distribution tower and the height difference data h between the towers.
[0042] Furthermore, based on the variation patterns among meteorological parameters, the range of variation, accuracy, and step size of each parameter are pre-defined as boundary constraints for the prediction function. For example, in some specific implementations, the range of ambient temperature T is set to -15 to 5. The accuracy of the value is 0.5. The wind speed v is set to a range of 0~20. The accuracy of the value is 0.5. The liquid water content (w) is set to a range of 0.2~2.6%. The accuracy of the value is 0.2. The median volume diameter 'a' of the water droplet ranges from 10 to 100. Precision of value is 5 The target power distribution line conductor diameter D is set to a range of 4~50mm, with a value accuracy of 1mm. The single-line current of a single conductor or split conductor is set to 0~800A, with a value accuracy of 10A. Setting boundary conditions can improve the model's prediction efficiency.
[0043] Step S12: Construct a training dataset based on the target parameters and the icing data corresponding to the target parameters, and use the training dataset to fit and train a preset neural network model to obtain the target neural network model.
[0044] In this embodiment, different target data required for prediction can be obtained by combining different measuring devices. For example, in one specific implementation, data is collected on a one-year cycle from power distribution lines covering most geographical scenes and climatic characteristics to create a dataset for model training. The measuring devices include, but are not limited to, multi-cylinder volume ice detectors, six-element meteorological instruments, cloud radar, and Rayleigh microwave radiometers. The training dataset consists of sample point data under several different geographical and climatic conditions, and each sample point data contains several sets of icing data sequences collected over time.
[0045] See Figure 2 As shown, in this embodiment, a three-layer preset neural network is designed. The input parameters include fixed parameters for a single sample, such as the conductor icing feature parameter k, the tower height difference h, and the altitude H, as constant inputs. The ambient temperature T(t), ambient wind speed v(t), liquid water content w(t), and median volume diameter of water droplets a(t) are used as one-dimensional input information, combined in groups with the corresponding time. And the final output and As a complete dataset, each sample location contains j sets of data along the time axis (the sample point is sampled j times at different times). Effective training samples are then obtained for n different geographical and climatic conditions of the towers. Group (3 types of fixed input parameters, 4 types of one-dimensional time axis signals, and 2 types of output rime and frost weight variation).
[0046] Furthermore, the preset neural network model is fitted and trained using the training dataset to obtain the target neural network model; the icing increment per unit length of the target power distribution line is calculated. The icing weight per unit length of power distribution line per unit hour was obtained. Based on the ambient temperature T(t) and ambient wind speed v(t), the types of icing are discussed, where rime ice is denoted as... Record of Rime Ice .
[0047] ;
[0048] ;
[0049] in, The output of the preset neural network model; The current icing mass; The prediction function is obtained by fitting and training a preset neural network model using the training dataset, and is used to represent the nonlinear relationship between the target parameters and the ice type and the change in ice mass.
[0050] Step S13: Use the icing data collected in the target prediction area to perform transfer learning on the target neural network model, and use the target neural network model after transfer learning to predict the icing quality of the power distribution lines in the target prediction area to obtain the corresponding prediction results.
[0051] In this embodiment, transfer learning is performed on a small dataset created for actual single-unit situations. Specifically, the transfer learning of the target neural network model using icing data collected from the target prediction region may include: collecting icing data of the target prediction region based on a preset data collection period, and constructing a regional icing dataset based on the icing data; adjusting the model parameters of the target neural network model using the regional icing dataset to complete the transfer learning of the target neural network model. That is, in order to achieve the best possible model base fit for the weights, laying the foundation for subsequent transfer learning. After completing the initial training and testing of the model using the dataset, it is also necessary to record data for a complete period in the prediction region to perform transfer learning on the model and correct the prediction results.
[0052] As can be seen from the above, the embodiments of this application utilize a neural network method for atmospheric parameters in freezing environments to construct a prediction function for atmospheric parameters in freezing environments, thereby achieving the prediction of icing quality of power distribution lines under freezing conditions. By referring to the past variation patterns of atmospheric parameters in freezing environments and employing a neural network algorithm, the prediction accuracy is improved, meeting the needs of practical engineering for icing quality prediction. The power distribution line icing quality prediction method based on the neural network method for atmospheric parameters in freezing environments does not rely on the physical changes of power distribution lines during icing, but rather starts from the icing patterns of atmospheric structures, using real-time atmospheric parameters in freezing environments to improve the accuracy of power distribution line icing prediction. This reduces the influence of conductor characteristics and environmental changes on icing prediction, improving the generalization ability and adaptability of icing prediction.
[0053] See Figure 3 As shown in the figure, this application discloses a neural network-based device for predicting the icing quality of power distribution lines, comprising:
[0054] The data acquisition module 11 is used to determine the target parameters affecting icing conditions and to perform data acquisition operations based on the target parameters to obtain icing data corresponding to the target parameters. The target parameters include conductor characteristic parameters, environmental climate parameters, and geographical topography parameters. The icing data corresponding to the target parameters includes icing type and the icing mass change data of the icing type per unit time. The icing type includes rime ice type and hoarfrost type. The conductor characteristic parameters are parameters generated based on conductor thickness, conductor surface hydrophobicity, power distribution type, power distribution voltage, and conductor appearance. The environmental climate parameters include environmental temperature, environmental wind speed, liquid water content, and median volume diameter of water droplets. The geographical topography parameters include the elevation data of the power distribution towers and the height difference data between towers.
[0055] The model building module 12 is used to construct a training dataset based on the target parameters and the corresponding icing data, and to use the training dataset to fit and train a preset neural network model to obtain a target neural network model. The training dataset consists of sample point data under several different geographical and climatic conditions, and each sample point data contains several sets of icing data sequences collected over time. The target neural network model consists of an input layer, a hidden layer, and an output layer. The input of the target neural network model includes fixed conductor feature parameters, altitude data, and elevation difference data, as well as environmental temperature, environmental wind speed, liquid water content, median volume diameter of water droplets, and initial icing mass that change over time. The output of the target neural network model is the change in the mass of rime ice and the change in the mass of hoarfrost per unit time.
[0056] The prediction module 13 is used to perform transfer learning on the target neural network model using the icing data collected in the target prediction area, and to use the target neural network model after transfer learning to predict the icing quality of the power distribution lines in the target prediction area, so as to obtain the corresponding prediction results.
[0057] In some specific embodiments, the model building module 12 may specifically include:
[0058] The prediction function generation unit is used to fit and train a preset neural network model using the training dataset to obtain a prediction function between the target parameters and the ice type and the change in ice mass.
[0059] The neural network model training unit is used to adjust the model parameters of the preset neural network model based on the prediction function to obtain the target neural network model.
[0060] In some specific embodiments, the prediction module 13 may specifically include:
[0061] The data acquisition unit is used to collect icing data of the target prediction area based on a preset data acquisition cycle, and to construct a regional icing dataset based on the icing data.
[0062] The model parameter adjustment unit is used to adjust the model parameters of the target neural network model using the regional icing dataset, so as to complete the transfer learning of the target neural network model.
[0063] In some specific embodiments, the neural network-based distribution line icing quality prediction device may further include:
[0064] The boundary constraint determination unit is used to set the variation range and accuracy of the environmental climate parameters based on the climate change law, and to use the variation range and accuracy as the boundary constraint conditions of the prediction function.
[0065] Furthermore, embodiments of this application also disclose an electronic device, Figure 4 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the neural network-based power distribution line icing quality prediction method disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be a computer.
[0066] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.
[0067] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.
[0068] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the neural network-based power line icing quality prediction method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.
[0069] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned neural network-based method for predicting the icing quality of power distribution lines. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.
[0070] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.
[0071] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0072] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0073] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0074] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for predicting the icing quality of power distribution lines based on neural networks, characterized in that, include: Determine the target parameters that affect icing conditions, and perform data acquisition operations based on the target parameters to obtain icing data corresponding to the target parameters; The target parameters include conductor characteristic parameters, environmental climate parameters, and geographical topographic parameters; the icing data corresponding to the target parameters include icing type and the icing mass change data of the icing type per unit time; the icing type includes rime ice type and hoarfrost type; A training dataset is constructed based on the target parameters and the corresponding icing data, and the preset neural network model is fitted and trained using the training dataset to obtain the target neural network model. The target neural network model is transferred to the target prediction area using icing data collected in the target prediction area, and the transferred-learned target neural network model is used to predict the icing quality of the power distribution lines in the target prediction area to obtain the corresponding prediction results.
2. The method for predicting the icing quality of power distribution lines based on neural networks according to claim 1, characterized in that, The conductor characteristic parameters are generated based on conductor thickness, conductor surface hydrophobicity, power distribution type, power distribution voltage, and conductor appearance; the environmental climate parameters include ambient temperature, ambient wind speed, liquid water content, and median volume diameter of water droplets; the geographical topography parameters include the elevation data of the power distribution towers and the height difference data between towers.
3. The method for predicting the icing quality of power distribution lines based on neural networks according to claim 1, characterized in that, The training dataset consists of sample point data under several different geographical and climatic conditions, and each sample point data contains several sets of icing data sequences collected over time.
4. The method for predicting the icing quality of power distribution lines based on neural networks according to claim 1, characterized in that, The target neural network model consists of an input layer, a hidden layer, and an output layer. The inputs of the target neural network model include fixed conductor feature parameters, altitude data, and elevation difference data, as well as environmental temperature, environmental wind speed, liquid water content, median volume diameter of water droplets, and initial icing mass that change over time. The outputs of the target neural network model are the mass change of rime and the mass change of hoarfrost per unit time.
5. The method for predicting the icing quality of power distribution lines based on neural networks according to claim 1, characterized in that, The step of fitting and training a preset neural network model using the training dataset to obtain a target neural network model includes: The preset neural network model is fitted and trained using the training dataset to obtain a prediction function between the target parameters and the ice type and the change in ice mass. The model parameters of the preset neural network model are adjusted based on the prediction function to obtain the target neural network model.
6. The method for predicting the icing quality of power distribution lines based on neural networks according to claim 5, characterized in that, Also includes: Based on the laws of climate change, the range of variation and the accuracy of the values of the environmental climate parameters are set, and the range of variation and the accuracy of the values are used as the boundary constraints of the prediction function.
7. The method for predicting the icing quality of power distribution lines based on neural networks according to any one of claims 1 to 6, characterized in that, The transfer learning of the target neural network model using icing data collected from the target prediction area includes: Icing data of the target prediction area is collected based on a preset data collection cycle, and a regional icing dataset is constructed based on the icing data. The model parameters of the target neural network model are adjusted using the regional icing dataset to complete the transfer learning of the target neural network model.
8. A neural network-based device for predicting the icing quality of power distribution lines, characterized in that, include: The data acquisition module is used to determine the target parameters affecting icing conditions and to perform data acquisition operations based on the target parameters to obtain icing data corresponding to the target parameters. The target parameters include conductor characteristic parameters, environmental climate parameters, and geographical topography parameters. The icing data corresponding to the target parameters includes icing type and the icing mass change data of the icing type per unit time. The icing type includes rime ice type and hoarfrost type. The model building module is used to build a training dataset based on the target parameters and the icing data corresponding to the target parameters, and to use the training dataset to fit and train a preset neural network model to obtain the target neural network model. The prediction module is used to perform transfer learning on the target neural network model using icing data collected in the target prediction area, and to use the target neural network model after transfer learning to predict the icing quality of the power distribution lines in the target prediction area, so as to obtain the corresponding prediction results.
9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the neural network-based method for predicting the icing quality of power distribution lines as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the neural network-based method for predicting the icing quality of power distribution lines as described in any one of claims 1 to 7.