Method, device and equipment for predicting galloping position of power transmission line and storage medium

Abstract

This disclosure provides a method, apparatus, device, and storage medium for predicting the galloping position of a transmission line. The method includes: acquiring location information and meteorological information of the transmission line at each sampling time point within a preset historical time window; constructing a joint state vector for each sampling time point based on the location and meteorological information; constructing a historical joint state vector sequence based on the joint state vectors of each sampling time point; and predicting the galloping position of the transmission line at multiple future sampling time points using a pre-constructed prediction model for the galloping position of the transmission line, based on the historical joint state vector sequence and the sampling interval between each sampling time point. Since the joint state vector at each sampling time point includes location information and the meteorological information driving its galloping, the prediction of the galloping position at future moments fully considers the physical mechanism of transmission line galloping, thereby improving the accuracy of the transmission line galloping position prediction.

Description

Technical Field

[0001] This disclosure relates to the field of power system safety detection technology, and in particular to a method, apparatus, equipment and storage medium for predicting the galloping position of transmission lines. Background Technology

[0002] Transmission line galloping is a low-frequency, large-amplitude oscillation phenomenon caused by overhead transmission lines under specific meteorological conditions (such as icing and strong winds), which seriously threatens the safe operation of the power grid. Currently, the predicted galloping locations are often based on historical galloping positions. However, this traditional method results in significant discrepancies between the predicted and actual galloping locations, failing to meet the requirements of line safety monitoring. Therefore, improving the accuracy of transmission line galloping location prediction has become a pressing technical problem that needs to be solved in this field. Summary of the Invention

[0003] In view of this, the present disclosure proposes a method, apparatus, device and storage medium for predicting the galloping position of transmission lines, which can improve the accuracy of transmission line galloping position prediction.

[0004] According to a first aspect of this disclosure, a method for predicting the galloping position of a transmission line is provided, comprising: Acquire location information and meteorological information of monitoring points on transmission lines at each sampling time point within a preset historical time window; Based on the location information and meteorological information of each sampling time point, a joint state vector of each sampling time point is constructed; Based on the joint state vector at each of the aforementioned sampling time points, a sequence of historical joint state vectors is constructed; Based on the historical joint state vector sequence and the sampling interval between each sampling time point, a pre-constructed prediction model for the galloping position of the transmission line is used to predict the galloping position of the monitoring point of the transmission line at multiple future sampling time points.

[0005] In one possible implementation, after acquiring the location information and meteorological information of the transmission line at each sampling time point within a preset historical time window, the method further includes: The location information and meteorological information at each sampling time point are subjected to at least one of the following data preprocessing operations: data cleaning, normalization, and timestamp alignment.

[0006] In one possible implementation, the prediction model is constructed based on multiple Gaussian process models, wherein the number of Gaussian process models is equal to the number of future sampling time points.

[0007] In one possible implementation, when constructing the prediction model based on multiple Gaussian process models, the following is included: For each of the multiple future sampling time points, an initial Gaussian process model is constructed. Acquire and construct training sample sets corresponding to each initial Gaussian process model based on the location information and meteorological information of the transmission line at each sampling time point within a preset historical time period; The initial Gaussian process model is trained using the training sample set corresponding to each initial Gaussian process model to obtain the Gaussian process model corresponding to each future sampling time point. The Gaussian process model corresponding to each future sampling time point is used to predict the galloping position of the monitoring point of the transmission line at each future sampling time point based on the historical joint state vector sequence and the sampling interval between each sampling time point.

[0008] The prediction model is constructed based on the Gaussian process model corresponding to each future sampling time point.

[0009] In one possible implementation, when constructing the training sample set corresponding to each initial Gaussian process model based on the location information and meteorological information of the transmission line at each sampling time point within a preset historical time period, the implementation is based on the multi-stage Markov model assumption.

[0010] In one possible implementation, when predicting the galloping position of the monitoring point of the transmission line at multiple future sampling time points using a pre-built prediction model of the transmission line galloping position based on the historical joint state vector sequence and the sampling interval between each sampling time point, the method includes: Based on the historical joint state vector sequence and the sampling interval, the input vector of the prediction model is constructed; Based on the input vector, the prediction model is used to predict the position deviation of the monitoring point of the transmission line at multiple future sampling time points relative to the last sampling time point within the historical time window. Based on the position deviation of the monitoring point of the transmission line at multiple future sampling time points, and combined with the position information of the last sampling time point within the historical time window, the galloping position of the monitoring point of the transmission line at multiple future sampling time points is calculated.

[0011] In one possible implementation, after predicting the galloping positions of the monitoring points on the transmission line at multiple future sampling time points, the method further includes: Calculate the confidence level of the dancing position at multiple future sampling time points.

[0012] According to a second aspect of this disclosure, a device for predicting the galloping position of a transmission line is provided, comprising: The data acquisition module is used to acquire the location information and meteorological information of the monitoring points of the transmission line at each sampling time point within a preset historical time window; A joint state vector construction module is used to construct a joint state vector for each of the sampling time points based on the location information and meteorological information of each sampling time point; The historical joint state vector sequence construction module is used to construct a historical joint state vector sequence based on the joint state vectors at each of the sampling time points; The prediction module is used to predict the galloping position of the monitoring point of the transmission line at multiple future sampling time points based on the historical joint state vector sequence and the sampling interval between each sampling time point, using a pre-built prediction model of the galloping position of the transmission line.

[0013] According to a third aspect of this disclosure, a device for predicting the galloping position of a transmission line is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to perform the method described in the first aspect of this disclosure.

[0014] According to a fourth aspect of this disclosure, a non-volatile computer-readable storage medium is provided that stores computer program instructions thereon, wherein the computer program instructions, when executed by a processor, implement the method described in the first aspect of this disclosure.

[0015] This disclosure provides a method, apparatus, device, and storage medium for predicting the galloping position of a transmission line. The method includes: acquiring location information and meteorological information of the transmission line at each sampling time point within a preset historical time window; constructing a joint state vector for each sampling time point based on the location and meteorological information; constructing a historical joint state vector sequence based on the joint state vectors of each sampling time point; and predicting the galloping position of the transmission line at multiple future sampling time points using a pre-constructed prediction model for the galloping position of the transmission line, based on the historical joint state vector sequence and the sampling interval between each sampling time point. Since the joint state vector at each sampling time point includes location information and the meteorological information driving its galloping, the prediction of the galloping position at future moments fully considers the physical mechanism of transmission line galloping, thereby improving the accuracy of the transmission line galloping position prediction.

[0016] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0017] The accompanying drawings, which are included in and form part of this specification, illustrate exemplary embodiments, features, and aspects of this disclosure together with the specification and serve to explain the principles of this disclosure.

[0018] Figure 1 A flowchart illustrating a method for predicting the galloping position of a transmission line according to an embodiment of the present disclosure is shown. Figure 2 A schematic block diagram of a transmission line galloping position prediction device according to an embodiment of the present disclosure is shown. Figure 3 A schematic block diagram of a transmission line galloping position prediction device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0019] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0020] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0021] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0022] <Method Implementation> Figure 1 A flowchart illustrating a method for predicting the galloping position of a transmission line according to an embodiment of the present disclosure is shown. Figure 1 As shown, the method includes steps S1100-S1400: S1100 acquires the location information and meteorological information of the monitoring points of the transmission line at each sampling time point within a preset historical time window.

[0023] First, it should be noted that for the transmission line to be monitored, a monitoring point is set at preset intervals, and a high-precision position sensor (such as a high-precision GNSS sensor) is deployed at each monitoring point. Each position sensor collects the position information of each monitoring point within a preset historical time window at a uniform sampling frequency (not less than 10 Hz). This allows the acquisition of the position information of each monitoring point at each sampling time point within the preset historical time window. The position information can be the three-dimensional position coordinates of the monitoring point directly collected by the position sensor. For simplified calculation, this position information can also be the lateral displacement of the monitoring point perpendicular to the line direction calculated based on the three-dimensional position coordinates directly collected by the position sensor. No specific limitation is made here.

[0024] In one possible implementation, the preset historical time window is determined by the current sampling time point. and the current sampling time point It is formed by p-1 adjacent sampling time points. The value of p is determined by the assumptions of the multi-stage Markov model. For example, according to the assumptions of the multi-stage Markov model, the three consecutive sampling time points to be predicted... , and The dancing position depends on the three most recent consecutive sampling time points. , and If the location and weather information are available, then the value of P is set to 3. When p=3, the historical time window is determined by... , and These sampling time points constitute the data. In this example, the location information of each monitoring point at each sampling time point within a preset historical time window is obtained, that is, the location information of each monitoring point at each sampling time point is obtained. , and Location information for these three sampling time points.

[0025] Furthermore, it should be noted that for the transmission line to be monitored, a miniature meteorological sensor can be deployed simultaneously at each monitoring point. Each miniature meteorological sensor collects meteorological data from each monitoring point within the same preset historical time window, using the same sampling frequency as the sensors at each location. In this way, meteorological information at each monitoring point at each sampling time point within the preset historical time window can be obtained. This meteorological information includes at least one meteorological factor that can drive the transmission line to gallop, such as wind speed (numerical value), wind direction (angle relative to the direction of the transmission line), temperature, and relative humidity.

[0026] In another possible implementation, a miniature weather sensor can be deployed on each tower connected to the transmission line. In this embodiment, each miniature weather sensor collects meteorological data at its deployment location within the same preset historical time window, using the same sampling frequency as the sensors at each location. Each monitoring point uses the meteorological information from the miniature weather sensor on its nearest tower at each sampling time point within the preset historical time window as the meteorological information for each monitoring point at each sampling time point within the preset historical time window.

[0027] This disclosure utilizes miniature meteorological sensors deployed at each monitoring point or on each tower connected to the transmission line to accurately acquire local micro-meteorological data within a preset distance range at each monitoring point. Compared to traditional methods that use regional meteorological station data with lower spatial resolution as meteorological data for each monitoring point, the meteorological data acquisition method of this disclosure effectively avoids the deficiency in galloping excitation identification caused by the smoothing of local micro-meteorological data (such as local meteorological data of the transmission line corridor) by large-scale averaged data. Therefore, it significantly improves the accuracy of identifying transmission line galloping excitation conditions and further enhances the prediction accuracy of future unknown galloping events.

[0028] It should be further explained here that this disclosure is actually based on the location information and meteorological information of each monitoring point at each sampling time point within a preset historical time window, to predict the agglomeration position of each monitoring point at multiple future sampling time points. Since the prediction method for the agglomeration position of each monitoring point is the same, the following section uses one monitoring point as an example to explain in detail the method for predicting the agglomeration position of each monitoring point at multiple future sampling time points.

[0029] In one possible implementation, after acquiring the location information and meteorological information of the monitoring points of the transmission line at each sampling time point within a preset historical time window, the method further includes: performing at least one data preprocessing operation on the location information and meteorological information at each sampling time point, including data cleaning, normalization, and timestamp alignment. Specifically, for the location information at each sampling time point, outlier identification and replacement are first performed to remove singular values ​​caused by signal interference; then, missing values ​​are identified and supplemented to obtain the cleaned location information at each sampling time point; next, the cleaned location information at each sampling time point is normalized so that the mean of the location information at each sampling time point is 0 and the variance is 1, in order to accelerate the rapid convergence of the prediction model training process and improve data stability. Then, the same method is used to perform data cleaning operations such as outlier identification and replacement, missing value identification and supplementation, and normalization on the meteorological values ​​of each meteorological factor at each sampling time point in the meteorological information, thereby obtaining the cleaned meteorological values ​​of each meteorological factor at each sampling time point. Finally, the sampling time points of location information and meteorological values ​​of various meteorological factors are strictly synchronized to obtain a time series sequence of location information and a time series sequence of meteorological values ​​of various meteorological factors with a unified sampling time interval.

[0030] S1200 constructs a joint state vector for each sampling time point based on location and meteorological information within a historical time window. Specifically, for each sampling time point within the historical time window... :Will Meteorological values ​​of multiple meteorological factors collected at different times are spliced ​​together in a specified order to obtain the values ​​of each sampling time point. Corresponding micro-meteorological vector For example, in an embodiment where meteorological information includes wind speed, wind direction, temperature, and relative humidity, each sampling time point... Corresponding micro-meteorological vector Then, for each sampling time point Corresponding location information and micro-meteorological vectors By splicing them together in the specified order, each sampling time point can be obtained. The corresponding joint state vector , where the joint state vector .

[0031] S1300: Based on the joint state vectors at each sampling time point, a historical joint state vector sequence is constructed. Specifically, the joint state vectors at each sampling time point are concatenated in chronological order to obtain the historical joint state vector sequence of the monitoring point within a preset historical time window. For example, within the historical time window... , and These sampling time points constitute the implementation example. First, the joint state vector for each sampling time point is constructed. , , Then, the joint state vectors at each sampling time point are concatenated in chronological order to obtain the historical joint state vector sequence within that historical time window. .

[0032] S1400, based on the historical joint state vector sequence within a preset historical time window and the sampling interval between each sampling time point, uses a pre-constructed prediction model for the galloping position of the transmission line to predict the galloping position of the monitoring points of the transmission line at multiple future sampling time points. Specifically, this may include the following steps: First, the input vector for the prediction model is constructed based on the historical joint state vector sequence and sampling interval within the historical time window. Specifically, the joint state vector at each sampling time point in the historical joint state vector sequence within the historical time window is flattened and compared with the sampling interval. By concatenating the vectors in sequence, the input vector for the prediction model can be obtained. In the historical joint state vector In the embodiment, the input vector .

[0033] Second, based on the input vector, a pre-built prediction model is used to predict the deviation of the galloping position of the monitoring points of the transmission line from the last sampling time point within the historical time window at multiple future sampling time points. Specifically, the input vector... The data is input into a pre-built prediction model, which automatically predicts the positional deviation of the monitoring points on the power line at multiple future sampling times relative to the last sampling time point within the historical time window. The vector obtained by sequentially concatenating the positional deviations at multiple future sampling times is the output vector of the prediction model. (Input vector...) To predict the next three consecutive sampling time points , and In the embodiment of the dancing position, the prediction model outputs that the monitoring point is in , and The positional deviations of these three future sampling time points relative to the last sampling time point within the historical time window are respectively denoted as: , as well as The output vector is formed by concatenating the positional deviations of the three future sampling time points. ,in, , For monitoring points at This location information for future sampling time points The last sampling time point within the historical time window Location information of monitoring points collected in real time. For monitoring points at Location information of this future sampling time point Relative to the last sampling time point within the historical time window Positional deviation. , And so on, which will not be elaborated upon here. It should be noted that... The formula only serves as an example to explain what the positional deviation of a future sampling time point is relative to the last sampling time point within the historical time window. Throughout the prediction process, In fact, it is directly predicted by a pre-built prediction model.

[0034] Third, based on the positional deviations of the monitoring points of the transmission line predicted by the prediction model at multiple future sampling time points, and combined with the positional information of the last sampling time point within the historical time window, the galloping positions of the monitoring points of the transmission line at multiple future sampling time points are calculated. Specifically, the sum of the positional deviations of the monitoring points of the transmission line at each future sampling time point and the positional information of the last sampling time point within the historical time window is calculated to obtain the galloping positions at each future sampling time point. Continuing from the previous example, for the future sampling time points... The corresponding dancing position of the monitoring point = + ,in, For the output of the prediction model The location information of this future sampling time point relative to the last sampling time point within the historical time window. Positional deviation of location information The last sampling time point within the historical time window actually measured by the position sensor. The actual location information. For future sampling time points. The corresponding dancing position of the monitoring point = + ,in, For the output of the prediction model The location information of this future sampling time point relative to the last sampling time point within the historical time window. The location deviation of the location information. For future sampling time points. The corresponding dancing position of the monitoring point = + ,in, For the output of the prediction model The location information of this future sampling time point relative to the last sampling time point within the historical time window. The positional deviation of the positional information.

[0035] In one possible implementation, after predicting the galloping positions of the monitoring points of the transmission line at multiple future sampling time points, the method further includes: calculating the confidence level of the galloping positions at multiple future sampling time points. In this way, if the confidence level of the predicted galloping position at a certain sampling time point does not meet the preset requirements, an "excessive prediction uncertainty" warning can be automatically triggered, prompting maintenance personnel to make a comprehensive judgment based on other information.

[0036] In one possible implementation, the prediction model described above can be constructed based on multiple Gaussian process models, where the number of Gaussian process models is equal to the number of future sampling time points.

[0037] In this embodiment, the following steps may be included when constructing a prediction model based on multiple Gaussian process models: First, for each of the multiple future sampling time points, construct an initial Gaussian process model. For example, if the number of future sampling time points is k, then k corresponding initial Gaussian process models need to be constructed. The initial Gaussian process model constructed for the kth future sampling time point is denoted as... Its mean function is set to 0, and the covariance function uses the radial basis function (RBF) kernel: In the formula, and Represents any two input vectors, Standard deviation, For length, both are The hyperparameters of the model Let be the covariance between any two input vectors.

[0038] Second, based on the location information and meteorological information of the transmission line at each sampling time point within a preset historical time period, a training sample set corresponding to each initial Gaussian process model is constructed. The location information and meteorological information at each sampling time point within the preset historical time period are all actual measured values ​​from the sensors.

[0039] In one possible implementation, the training sample set corresponding to each initial Gaussian process model is constructed based on the assumption of a multi-stage Markov model.

[0040] For example, based on the assumptions of a multi-stage Markov model: the dancing position at the next k (k=3) consecutive sampling time points depends on the location and meteorological information of the most recent p (p=3) consecutive sampling time points. Therefore, setting the historical window length p=3, a training sample is constructed for each sampling time point within a preset historical time period. The following uses sampling time points... Taking this as an example, the process of constructing training samples for each historical sampling time point is explained.

[0041] First, construct the input vector for the prediction model. Referring to the above, first construct the sampling time points. Corresponding historical joint state vector sequence Then, the sequence of historical joint state vectors The joint state vector at each sampling time point is flattened and compared with the sampling interval. By concatenating the data in sequence, the sampling time points can be obtained. corresponding input vector ,Right now .

[0042] Secondly, the output vector of the prediction model is constructed. To enhance the generalization ability of the prediction model, the output vector of the prediction model is defined as... , and The dancing positions of these k (k=3) future sampling time points relative to the last sampling time within the historical time window The position deviation sequence, i.e., the output vector .in, , , In this calculation process , , , , respectively , and The actual galloping positions of the transmission lines were collected at these historical sampling time points.

[0043] At this point, the sampling time point can be obtained. Corresponding training samples .

[0044] By using a sliding time window (length P=3, step size equal to the sampling interval between each sampling time point), traversing the entire historical dataset, and referring to the method described above, a large number of datasets can be generated. Training samples, all of which constitute the training sample set. .in, For the first The input vector corresponding to each historical sampling time point For the first The output vector corresponding to each historical sampling time point This represents the total number of samples in the training sample set.

[0045] After obtaining the training sample set Then, based on this training sample set, we continue to construct the initial Gaussian process model corresponding to the k-th future sampling time. The training sample set. Specifically, iterate through each training sample in the training sample set, and for the current training sample encountered ( Extract the input vector and output vector The positional deviation corresponding to the k-th sampling time point is then extracted from the input vector. and output vector The positional deviation corresponding to the kth sampling time point (i.e., the output vector) The k-th dimension output value ) combined into sample pairs ( This is the initial Gaussian process model. This corresponds to a single training sample. After the iteration is complete, all the recombined training samples are used as the initial Gaussian process model. The corresponding training sample set. The construction process of the training sample set for the initial Gaussian process model corresponding to other sampling time points is the same as above, and will not be repeated here.

[0046] Second, the training sample sets corresponding to each initial Gaussian process model are used to train each initial Gaussian process model to obtain the Gaussian process model corresponding to each future sampling time point. Among them, the Gaussian process model corresponding to each future sampling time point is used to predict the galloping position of the monitoring point of the transmission line at each future sampling time point based on the historical joint state vector sequence and the sampling interval between each sampling time point.

[0047] Since the training process of the initial Gaussian process model is the same for all future sampling time points, the following section takes the k-th future sampling time point as an example to analyze its corresponding initial Gaussian process model. The training will be explained in detail.

[0048] Specifically, the initial Gaussian process model is used. The corresponding training sample set is optimized by maximizing the logarithm of the marginal likelihood. The optimization problem can be expressed as: In the formula, It is the first of all training samples A vector composed of dimensional output values. The covariance matrix is ​​composed of the covariances between any two input vectors in the training sample set. Specifically, iterate through any two input vectors in the training sample set. For each pair of input vectors encountered, use the covariance function mentioned above to calculate the covariance between them. During the training phase, the hyperparameters in the covariance function are unknowns, and the calculated covariance between the two input vectors is represented by the two unknown hyperparameters of the model. After the iteration is complete, the covariance between any two input vectors in the training sample set can be obtained. Organizing all the calculated covariances into a matrix form gives the covariance matrix. , It is the noise variance added for numerical stability (which is also an optimizable hyperparameter). It is the identity matrix. The optimal hyperparameters can be calculated by solving the above optimization problem using the conjugate gradient method or the L-BFGS algorithm. , to optimize hyperparameters Assigned to the initial Gaussian process model This yields the Gaussian process model for the k-th future sampling time point. Simultaneously, it is necessary to save the training data input matrix used during the training process of this Gaussian process model. and the corresponding dimension of the output vector This is used for subsequent prediction calculations. The training data input matrix... It consists of the input vectors of all training samples. For the first training sample The vector formed by the output values ​​of the dimension is equivalent to .

[0049] By referring to the above training method, we can train the Gaussian process model, training data input matrix, and output vector of the corresponding dimension for each future sampling time point.

[0050] Third, a prediction model is constructed based on the Gaussian process models corresponding to each future sampling time point. Specifically, the combination of Gaussian process models corresponding to each future sampling time point constitutes the constructed prediction model.

[0051] After constructing the aforementioned prediction model, the galloping positions of the transmission line monitoring points at multiple future sampling time points can be predicted based on this model. In an embodiment where the constructed prediction model is composed of Gaussian process models corresponding to each future sampling time point, the input vector of the prediction model is constructed based on the historical joint state vector sequence within the historical time window and the sampling interval. Then, iterate through each future sampling time point. For the k-th future sampling time point reached, based on the input vector... The training data input matrix corresponding to the k-th future sampling time point The k-th future sampling time point corresponds to the k-th training sample. The vector formed by the dimensional output values Using the Gaussian process model corresponding to the k-th future sampling time point, the conditional distribution of the predicted position deviation corresponding to the k-th future sampling time point is calculated. Based on the properties of Gaussian processes, the conditional distribution is Gaussian, and its mean (predicted value) and variance (uncertainty) are calculated as follows: In the formula, It is the covariance matrix of the training samples themselves. Specifically, it is calculated using the Gaussian process model corresponding to the k-th future sampling time point to obtain the training data input matrix. The covariance between itself and the variable is K; It is the training sample and the current input vector The covariance vector between them, specifically, is calculated from the Gaussian process model corresponding to the k-th future sampling time point to obtain the training data input matrix. and the current input vector The covariance between them, that covariance is ; It is the covariance of the current input vector itself. Specifically, it is calculated using the Gaussian process model corresponding to the k-th future sampling time point. The covariance between itself and itself is called the covariance. ; To estimate the level of noise, It is the identity matrix. The mean of the Gaussian distribution described above represents the deviation of the transmission line's galloping position at the k-th future sampling time point from the last sampling time point within the historical time window. for Corresponding variance is used to characterize The confidence level is then determined. After the traversal is complete, the positional deviation of the transmission line's galloping position at each future sampling time point relative to the last sampling time point within the historical time window, along with the corresponding confidence level for each positional deviation, can be obtained. Then, the positional deviation corresponding to each future sampling time point is summed with the measured position information of the last sampling time point within the historical time window to obtain the galloping position of the transmission line's monitoring point at multiple future sampling time points.

[0052] The above describes a method for predicting the future galloping position of transmission lines equipped with position sensors and micro-meteorological sensors. Since the prediction model in this scheme is trained using the position deviation of the transmission line's galloping position at each future sampling time point relative to the last sampling time point within the historical time window as the training objective, this prediction model can be transferred to predict the galloping positions of multiple future time points at various monitoring points on transmission lines without sensors. The specific prediction steps are as follows: First, set the initial state and acquire real-time weather data. Specifically, for sensorless target lines, maintenance personnel should set an initial estimate of the galloping position based on experience or line logs. (For example, the initial position can be assumed to be 0, i.e., a static equilibrium position). Simultaneously, real-time micrometeorological data of the vicinity of the railway corridor are acquired. (This information can be obtained by downscaling from nearby weather stations or numerical weather forecasts).

[0053] Second, iterative multi-step state deduction. The specific steps are as follows: Step 1: Construct an initial hypothetical historical joint state vector sequence using the set initial location and real-time meteorological data. For example, the location information in the joint state vectors at the first three times t-2, t-1, and t is... All meteorological information is Then, referring to the method above, we can obtain the historical joint state vector sequence: .

[0054] Step 2: Based on and sampling time interval The system constructs an input vector and inputs the constructed output vector into a trained prediction model, thereby predicting the next time step. Corresponding position offset and location information .

[0055] Step 3: Based on the predicted location information For the historical joint state vector sequence Perform rolling state updates. Specifically, use the predicted state... and the latest moment A new state is formed, which, together with the previous partial joint state vectors, constructs a new sequence of historical joint state vectors: Then, input the data into the model again to predict the next step. By iterating in this way, a continuous time series of dancing displacement data, similar to that obtained by a virtual sensor, can be generated.

[0056] It should be noted that for target lines without installed sensors, it is necessary to first select a prediction model of another sensor-deployed output circuit that matches the target line based on similarity in structural parameters, geographical location, and micro-meteorological characteristics, and then combine the initial state, structural parameters, and available verification data of the target line to perform bias correction or parameter correction on the source model to obtain a transfer model suitable for the target line. Based on this transfer model, an iterative prediction process for the future galloping position of the target line is then carried out.

[0057] This method is applicable to the pre-assessment of galloping risk for newly built lines or lines not covered by monitoring.

[0058] This disclosure provides a method for predicting the galloping position of a transmission line, comprising: acquiring location information and meteorological information of the transmission line at each sampling time point within a preset historical time window; constructing a joint state vector for each sampling time point based on the location information and meteorological information; constructing a historical joint state vector sequence based on the joint state vector of each sampling time point; and predicting the galloping position of the transmission line at multiple future sampling time points using a pre-constructed prediction model for the galloping position of the transmission line, based on the historical joint state vector sequence and the sampling interval between each sampling time point. Since the joint state vector at each sampling time point includes location information and the meteorological information driving its galloping, the prediction of the galloping position at future moments fully considers the physical mechanism of transmission line galloping, thereby improving the accuracy of the prediction.

[0059] Furthermore, in one embodiment of this disclosure, a galloping position prediction model based on a multi-stage Markov framework and a Gaussian process learner is proposed. This prediction model defines the galloping state of a transmission line as a joint state of "position-weather" and assumes that the probability distribution of its future state depends on the joint state sequence of several recent consecutive historical moments, thereby overcoming the shortcomings of the "memoryless" assumption of the classical Markov model in describing the inertial characteristics of the galloping dynamics process.

[0060] <Device Embodiment> Figure 2 A schematic block diagram of a transmission line galloping position prediction device according to an embodiment of the present disclosure is shown. Figure 2 As shown, the device 100 includes: The data acquisition module 110 is used to acquire the location information and meteorological information of the monitoring points of the transmission line at each sampling time point within a preset historical time window; The joint state vector construction module 120 is used to construct the joint state vector for each sampling time point based on the location information and meteorological information at each sampling time point; The historical joint state vector sequence construction module 130 is used to construct a historical joint state vector sequence based on the joint state vector at each sampling time point; The prediction module 140 is used to predict the galloping position of the monitoring points of the transmission line at multiple future sampling time points based on the historical joint state vector sequence and the sampling interval between each sampling time point, using a pre-built prediction model of the galloping position of the transmission line.

[0061] <Equipment Example> Figure 3 A schematic block diagram of a transmission line galloping position prediction device according to an embodiment of the present disclosure is shown. Figure 3 As shown, the transmission line galloping position prediction device 200 includes a processor 210 and a memory 220 for storing executable instructions of the processor 210. The processor 210 is configured to implement any of the aforementioned transmission line galloping position prediction methods when executing the executable instructions.

[0062] It should be noted here that the number of processors 210 can be one or more. Furthermore, the transmission line galloping position prediction device 200 in this embodiment may also include an input device 230 and an output device 240. The processors 210, memory 220, input device 230, and output device 240 can be connected via a bus or other means, without specific limitations here.

[0063] The memory 220, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and various modules, such as the program or module corresponding to the transmission line galloping position prediction method of this disclosure embodiment. The processor 210 executes various functional applications and data processing of the transmission line galloping position prediction device 200 by running the software program or module stored in the memory 220.

[0064] Input device 230 can be used to receive input digital numbers or signals. These signals may include key signals related to user settings and function control of the device / terminal / server. Output device 240 may include a display device such as a screen.

[0065] <Storage Medium Examples> According to a fourth aspect of this disclosure, a non-volatile computer-readable storage medium is also provided, on which computer program instructions are stored, which, when executed by processor 210, implement the method for predicting the galloping position of any of the preceding methods.

[0066] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical applications, or technical improvements to the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for predicting the galloping position of a transmission line, characterized in that, include: Acquire location information and meteorological information of monitoring points on transmission lines at each sampling time point within a preset historical time window; Based on the location information and meteorological information of each sampling time point, a joint state vector of each sampling time point is constructed; Based on the joint state vector at each of the aforementioned sampling time points, a sequence of historical joint state vectors is constructed; Based on the historical joint state vector sequence and the sampling interval between each sampling time point, a pre-constructed prediction model for the galloping position of the transmission line is used to predict the galloping position of the monitoring point of the transmission line at multiple future sampling time points.

2. The method according to claim 1, characterized in that, After acquiring the location information and meteorological information of the transmission line at each sampling time point within a preset historical time window, the process also includes: The location information and meteorological information at each sampling time point are subjected to at least one of the following data preprocessing operations: data cleaning, normalization, and timestamp alignment.

3. The method according to claim 1, characterized in that, The prediction model is constructed based on multiple Gaussian process models, wherein the number of Gaussian process models is equal to the number of future sampling time points.

4. The method according to claim 3, characterized in that, When constructing the prediction model based on multiple Gaussian process models, the following are included: For each of the multiple future sampling time points, an initial Gaussian process model is constructed. Acquire and construct training sample sets corresponding to each initial Gaussian process model based on the location information and meteorological information of the transmission line at each sampling time point within a preset historical time period; The initial Gaussian process model is trained using the training sample set corresponding to each initial Gaussian process model to obtain the Gaussian process model corresponding to each future sampling time point. The Gaussian process model corresponding to each future sampling time point is used to predict the galloping position of the monitoring point of the transmission line at each future sampling time point based on the historical joint state vector sequence and the sampling interval between each sampling time point. The prediction model is constructed based on the Gaussian process model corresponding to each future sampling time point.

5. The method according to claim 4, characterized in that, When constructing the training sample set corresponding to each initial Gaussian process model based on the location information and meteorological information of the transmission line at each sampling time point within a preset historical time period, the assumption of a multi-stage Markov model is implemented.

6. The method according to claim 1, characterized in that, When predicting the galloping position of the monitoring point of the transmission line at multiple future sampling time points based on the historical joint state vector sequence and the sampling interval between each sampling time point, using a pre-built prediction model for the galloping position of the transmission line, the method includes: Based on the historical joint state vector sequence and the sampling interval, the input vector of the prediction model is constructed; Based on the input vector, the prediction model is used to predict the position deviation of the monitoring point of the transmission line at multiple future sampling time points relative to the last sampling time point within the historical time window. Based on the position deviation of the monitoring point of the transmission line at multiple future sampling time points, and combined with the position information of the last sampling time point within the historical time window, the galloping position of the monitoring point of the transmission line at multiple future sampling time points is calculated.

7. The method according to claim 1, characterized in that, After predicting the galloping positions of the monitoring points on the transmission line at multiple future sampling time points, the method further includes: Calculate the confidence level of the dancing position at multiple future sampling time points.

8. A device for predicting the galloping position of a transmission line, characterized in that, include: The data acquisition module is used to acquire the location information and meteorological information of the monitoring points of the transmission line at each sampling time point within a preset historical time window; A joint state vector construction module is used to construct a joint state vector for each of the sampling time points based on the location information and meteorological information of each sampling time point; The historical joint state vector sequence construction module is used to construct a historical joint state vector sequence based on the joint state vectors at each of the sampling time points; The prediction module is used to predict the galloping position of the monitoring point of the transmission line at multiple future sampling time points based on the historical joint state vector sequence and the sampling interval between each sampling time point, using a pre-built prediction model of the galloping position of the transmission line.

9. A device for predicting the galloping position of a transmission line, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to implement the method of any one of claims 1 to 7 when executing the executable instructions.

10. A non-volatile computer-readable storage medium storing computer program instructions thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images