Method, device and equipment for predicting ampacity of overhead transmission line under complex working condition

Abstract

The application relates to an overhead power transmission line ampacity prediction method, device and equipment under complex working conditions, and relates to the technical field of power transmission. The method comprises the following steps: acquiring influence factor data of a target overhead power transmission line under multiple continuous historical time periods and influencing health states; for each historical time period, performing feature extraction on the influence factor data to obtain influence factor features corresponding to the historical time periods; determining health state change information of the target overhead power transmission line according to the influence factor features; the health state change information is used for representing the change relationship between the health state of the target overhead power transmission line and time; and according to the health state change information and meteorological forecast data of at least one future time, determining ampacity prediction data of the target overhead power transmission line at the at least one future time. The method can accurately determine the ampacity prediction data of the target overhead power transmission line at the at least one future time, and improve the efficiency and safety of power transmission.

Description

Technical Field

[0001] This application relates to the field of power transmission technology, and in particular to methods, devices and equipment for predicting the current carrying capacity of overhead transmission lines under complex operating conditions. Background Technology

[0002] In the field of power transmission technology, overhead transmission lines play an indispensable role, and their transmission capacity directly affects the security of power transmission and the stability of the entire power system. Within a power system, the current-carrying capacity of overhead transmission lines directly impacts their transmission capacity. Therefore, predicting the current-carrying capacity of overhead transmission lines is of great significance.

[0003] Traditional methods typically predict the current-carrying capacity of overhead transmission lines based on their age or real-time meteorological data. However, these methods have significant limitations. For example, predicting current-carrying capacity based on the age of the transmission line cannot accurately reflect its deterioration trend, potentially leading to misjudgments. Similarly, methods relying on real-time meteorological data cannot adapt to changes in external operating conditions, causing the predicted current-carrying capacity to lag behind actual conditions. Therefore, these traditional methods cannot accurately predict the current-carrying capacity of overhead transmission lines, impacting the efficiency and safety of power transmission. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, device, and equipment for predicting the current carrying capacity of overhead transmission lines under complex operating conditions, which can accurately predict the current carrying capacity of overhead transmission lines, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions, including:

[0006] Data on factors affecting the health status of the target overhead transmission line were obtained over multiple consecutive historical periods; the duration of each historical period was the same.

[0007] For each historical period, feature extraction is performed on the influencing factor data to obtain the influencing factor features corresponding to the historical period;

[0008] Based on the characteristics of influencing factors, the health status change information of the target overhead transmission line is determined; among which, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time.

[0009] Based on information on changes in health status and meteorological forecast data for at least one future time, determine the predicted current carrying capacity of the target overhead transmission line at at least one future time.

[0010] In one embodiment, determining the health status change information of the target overhead transmission line based on the characteristics of influencing factors includes: inputting the characteristics of influencing factors into a trained latent variable representation model to obtain the latent variables of influencing factors corresponding to historical time periods; the trained latent variable representation model is implemented based on an encoding network and a decoding network; splicing and fusing the latent variables of influencing factors corresponding to multiple historical time periods in chronological order to obtain health status trajectory information; obtaining the rate of change and acceleration of change of health status of the target overhead transmission line based on the latent variables of influencing factors corresponding to multiple historical time periods; and obtaining health status change information based on the health status trajectory information, the rate of change of health status, and the acceleration of change of health status.

[0011] In one embodiment, the latent variable representation model is trained according to the following steps: acquiring data samples of influencing factors affecting the health status of a reference overhead transmission line under multiple consecutive reference historical periods; wherein, each reference historical period has the same duration; for each reference historical period, feature extraction is performed on the influencing factor data samples to obtain the first influencing factor feature samples corresponding to that reference historical period; using an encoding network, feature encoding is performed on the first influencing factor feature samples to obtain the first latent variable distribution features of the first influencing factor feature samples in the target latent variable space, as well as the corresponding latent variable mean and latent variable standard deviation; based on the latent variable mean and latent variable standard deviation, the first latent variable distribution features are resampled to obtain the second latent variable distribution features; using a decoding network, feature decoding is performed on the second latent variable distribution features to obtain the second influencing factor feature samples; based on the first influencing factor feature samples, the second influencing factor feature samples, the latent variable mean, the latent variable standard deviation, and the second latent variable distribution features corresponding to multiple reference historical periods, the pre-constructed latent variable representation model is trained to obtain the latent variable representation model.

[0012] In one embodiment, a pre-constructed latent variable representation model is trained based on first influencing factor feature samples, second influencing factor feature samples, latent variable mean, latent variable standard deviation, and second latent variable distribution characteristics corresponding to multiple reference historical periods. This includes: determining a reconstruction loss based on the first and second influencing factor feature samples corresponding to multiple reference historical periods; determining a relative entropy loss based on the latent variable mean and standard deviation corresponding to multiple reference historical periods; for each reference historical period, taking the second latent variable distribution characteristics corresponding to the next adjacent reference historical period as positive samples and the second latent variable distribution characteristics corresponding to other reference historical periods as negative samples, and determining a time comparison loss corresponding to that reference historical period based on the similarity between the positive and negative samples; and training the pre-constructed latent variable representation model based on the reconstruction loss, relative entropy loss, and time comparison losses corresponding to each reference historical period.

[0013] In one embodiment, the rate of change and acceleration of change of health status of the target overhead transmission line are obtained based on the latent variables of influencing factors corresponding to multiple historical time periods. This includes: for each historical time period, determining the rate of change of health status in that historical time period based on the difference between the latent variables of influencing factors corresponding to the next historical time period and the latent variables of influencing factors corresponding to that historical time period, and the time interval between two adjacent historical time periods; and determining the acceleration of change of health status in that historical time period based on the difference between the rate of change of health status in the next historical time period and the rate of change of health status corresponding to that historical time period, and the time interval.

[0014] In one embodiment, based on health status change information and meteorological forecast data for at least one future time, the current carrying capacity prediction data of the target overhead transmission line at at least one future time is determined, including: inputting the health status change information and meteorological forecast data into a trained current carrying capacity prediction model to obtain the current carrying capacity prediction data of the target overhead transmission line at at least one future time; wherein, the current carrying capacity prediction model is trained based on the health status change information of the reference overhead transmission line, meteorological monitoring data at various historical times, and current carrying capacity monitoring data at the corresponding historical times.

[0015] Secondly, this application also provides a device for predicting the current carrying capacity of overhead transmission lines under complex operating conditions, comprising:

[0016] The acquisition module is used to acquire data on factors affecting the health status of the target overhead transmission line in multiple consecutive historical periods; wherein the duration of each historical period is the same.

[0017] The extraction module is used to extract features from the influencing factor data for each historical period to obtain the influencing factor features corresponding to the historical period.

[0018] The first determining module is used to determine the health status change information of the target overhead transmission line based on the characteristics of influencing factors; wherein, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time.

[0019] The second determining module is used to determine the predicted current carrying capacity of the target overhead transmission line at at least one future time based on information on changes in health status and meteorological forecast data at at least one future time.

[0020] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0021] Data on factors affecting the health status of the target overhead transmission line were obtained over multiple consecutive historical periods; the duration of each historical period was the same.

[0022] For each historical period, feature extraction is performed on the influencing factor data to obtain the influencing factor features corresponding to the historical period;

[0023] Based on the characteristics of influencing factors, the health status change information of the target overhead transmission line is determined; among which, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time.

[0024] Based on information on changes in health status and meteorological forecast data for at least one future time, determine the predicted current carrying capacity of the target overhead transmission line at at least one future time.

[0025] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0026] Data on factors affecting the health status of the target overhead transmission line were obtained over multiple consecutive historical periods; the duration of each historical period was the same.

[0027] For each historical period, feature extraction is performed on the influencing factor data to obtain the influencing factor features corresponding to the historical period;

[0028] Based on the characteristics of influencing factors, the health status change information of the target overhead transmission line is determined; among which, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time.

[0029] Based on information on changes in health status and meteorological forecast data for at least one future time, determine the predicted current carrying capacity of the target overhead transmission line at at least one future time.

[0030] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0031] Data on factors affecting the health status of the target overhead transmission line were obtained over multiple consecutive historical periods; the duration of each historical period was the same.

[0032] For each historical period, feature extraction is performed on the influencing factor data to obtain the influencing factor features corresponding to the historical period;

[0033] Based on the characteristics of influencing factors, the health status change information of the target overhead transmission line is determined; among which, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time.

[0034] Based on information on changes in health status and meteorological forecast data for at least one future time, determine the predicted current carrying capacity of the target overhead transmission line at at least one future time.

[0035] The aforementioned method, apparatus, and equipment for predicting the current carrying capacity of overhead transmission lines under complex operating conditions can acquire data on influencing factors affecting the health status of a target overhead transmission line over multiple consecutive historical periods. For each historical period, feature extraction is performed on the influencing factor data to obtain the corresponding influencing factor features. Based on these features, the health status change information of the target overhead transmission line is determined. Using this health status change information, along with at least one future weather forecast, the predicted current carrying capacity of the target overhead transmission line at at least one future moment is obtained. The health status change information characterizes the relationship between the health status of the target overhead transmission line and time, reflecting its deterioration trend and thus mitigating potential misjudgments of the current carrying capacity. The weather forecast at at least one future moment promptly reflects changes in external operating conditions, further preventing the predicted current carrying capacity from lagging behind actual results. Therefore, based on the health status change information and weather forecast data, the predicted current carrying capacity of the target overhead transmission line at at least one future moment can be determined relatively accurately, improving the efficiency and safety of power transmission. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a diagram illustrating the application environment of a method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions in one embodiment.

[0038] Figure 2 This is a flowchart illustrating a method for predicting the current carrying capacity of an overhead transmission line under complex operating conditions in one embodiment.

[0039] Figure 3 This is a flowchart illustrating the steps for determining health status change information in one embodiment.

[0040] Figure 4 This is a flowchart illustrating the steps for determining traffic volume prediction data in one embodiment.

[0041] Figure 5 This is a flowchart illustrating a method for predicting the current carrying capacity of overhead transmission lines under another complex operating condition in one embodiment.

[0042] Figure 6 This is a structural block diagram of an overhead transmission line current carrying capacity prediction device under complex operating conditions in one embodiment.

[0043] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0045] The method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0046] In one exemplary embodiment, such as Figure 2 As shown, a method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions is provided, and this method is applied to... Figure 1 The following explanation uses servers as an example, including S210 to S240. Among them:

[0047] S210, Obtain data on factors affecting the health status of the target overhead transmission line in multiple consecutive historical periods; wherein, the duration of each historical period is the same.

[0048] The influencing factor data can be understood as the influencing factor data of the target influencing factors. This influencing factor data can include at least one of the following: influencing factor data of meteorological factors and influencing factor data of line electrical factors. Meteorological factors can include the aforementioned target meteorological conditions. Line electrical factors can include grid load, current, voltage, line temperature, etc.

[0049] In one optional embodiment, raw data on the impact of the target overhead transmission line on its health status over multiple consecutive historical periods can be obtained. The raw data is then preprocessed to obtain influencing factor data.

[0050] For example, the raw data may include data such as temperature, humidity, wind speed, radiation intensity, line load, tension, current and voltage of the target overhead transmission line.

[0051] Preprocessing may include at least one of data cleaning and data standardization.

[0052] Data cleaning processes can include filling in missing values, deleting outliers, and denoising data.

[0053] Data standardization processing can be understood as standardizing or normalizing data.

[0054] In one optional embodiment, the aforementioned raw data can be acquired using image acquisition devices, sensors, weather forecasting platforms, etc. For example, image acquisition devices can acquire images, videos, and other data corresponding to the target overhead transmission line over multiple consecutive historical time periods. The images and videos may contain meteorological information, condition information of the target overhead transmission line, etc. For example, sensors can acquire tension data of the target overhead transmission line, etc. For example, weather forecast data can be acquired based on weather forecast data.

[0055] S220: For each historical period, feature extraction is performed on the influencing factor data to obtain the influencing factor features corresponding to the historical period.

[0056] In one optional embodiment, for each type of influencing factor data, the original features corresponding to that influencing factor data are extracted. Then, the original features corresponding to various influencing factor data are aggregated and uniformly encoded to obtain fixed-length influencing factor features corresponding to that historical period.

[0057] Specifically, graph convolutional neural networks can be used to extract features from image or video frames to obtain raw visual features. Raw meteorological features can be obtained based on data on influencing meteorological factors. Raw electrical features can be obtained based on data on influencing electrical factors. Then, the raw visual features, raw meteorological features, and raw electrical features can be aggregated and uniformly encoded to obtain fixed-length influencing factor features corresponding to that historical period.

[0058] S230, Based on the characteristics of influencing factors, determine the health status change information of the target overhead transmission line; wherein, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time.

[0059] Among them, health status change information can be understood as health status at various historical periods, the rate of change of health status within each historical period, and the acceleration of health status.

[0060] S240, based on information on changes in health status and weather forecast data for at least one future time, determine the predicted current carrying capacity of the target overhead transmission line at at least one future time.

[0061] The meteorological forecast data may include forecast data for the target weather. The target weather may include at least one of the following: temperature, humidity, wind speed, radiation intensity, etc.

[0062] In one alternative embodiment, weather forecast data for at least one future moment can be obtained through a weather forecasting platform.

[0063] In one optional embodiment, the predicted carrying capacity data corresponding to the health status change information and the weather forecast data at at least one future time can be determined by using the pre-defined correspondence between health status change information, meteorological monitoring data and carrying capacity data.

[0064] In the above-mentioned method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions, the health status change information characterizes the relationship between the health status of the target overhead transmission line and time, that is, it can reflect the deterioration trend of the overhead transmission line, thereby avoiding misjudgment of the current carrying capacity to a certain extent. At least one future moment's weather forecast data can reflect the changes in external operating conditions in a timely manner, thereby avoiding the situation where the current carrying capacity prediction results lag behind the actual results to a certain extent. Therefore, based on the health status change information and weather forecast data, the current carrying capacity prediction data of the target overhead transmission line at at least one future moment can be determined more accurately, thereby improving the efficiency and safety of power transmission.

[0065] In one exemplary embodiment, such as Figure 3 As shown, the steps for determining health status change information include S310~S340. Wherein:

[0066] S310, input the features of influencing factors into the trained latent variable representation model to obtain the latent variables of influencing factors corresponding to the historical period; the trained latent variable representation model is implemented based on an encoding network and a decoding network.

[0067] Among them, the input to the latent variables represents the influencing factor features in the model, which are the influencing factor features corresponding to multiple historical periods spliced ​​together in chronological order.

[0068] Among them, the latent variables of influencing factors can be understood as the distribution information of the characteristics of influencing factors in the target latent space. The latent variables of influencing factors can include the distribution characteristics of the latent variables, as well as the mean and standard deviation of the latent variables corresponding to the distribution characteristics.

[0069] The encoding network can be used to convert the input influencing factor features into latent influencing factor variables. The decoding network can decode the latent influencing factor variables into reconstructed influencing factor features. Optionally, the latent variable representation model can be a variational autoencoder network.

[0070] In practice, the encoding network in the trained latent variable representation model can be used to encode the characteristics of influencing factors to obtain the latent variables of influencing factors.

[0071] S320, in chronological order, splices and merges the latent variables of influencing factors corresponding to multiple historical periods to obtain health status trajectory information.

[0072] Here, chronological order can be understood as the temporal sequence of multiple historical periods. Merging and combining the latent variables of influencing factors corresponding to multiple historical periods can be understood as combining and combining the distribution characteristics of the latent variables to obtain the health status trajectory.

[0073] S330, based on the latent variables of influencing factors corresponding to multiple historical periods, obtains the rate of change and acceleration of the health status of the target overhead transmission line.

[0074] Among them, the rate of change in health status can be understood as the latent variable of influencing factors per unit time.

[0075] Among them, the acceleration of change in health status can be understood as the rate of change of health status per unit time.

[0076] Optionally, the rate of change and acceleration of change in health status can be obtained for each historical time period.

[0077] S340: Based on the health status trajectory information, the rate of change of health status, and the acceleration of change of health status, the information on changes in health status is obtained.

[0078] In this embodiment, the latent variable representation model is used to transform the features of influencing factors to obtain the latent variables of influencing factors corresponding to each historical period. This not only achieves effective compression and noise reduction of high-dimensional features (i.e., features of influencing factors), but also successfully extracts the main implicit features reflecting changes in health status, thereby enabling the more accurate acquisition of information on changes in health status.

[0079] In one exemplary embodiment, the latent variable representation model includes an encoding network and a decoding network.

[0080] The latent variable representation model is trained according to the following steps: Data samples of influencing factors affecting the health status of a reference overhead transmission line under multiple consecutive reference historical periods are obtained; wherein each reference historical period has the same duration; for each reference historical period, features are extracted from the influencing factor data samples to obtain the first influencing factor feature samples corresponding to that reference historical period; using an encoding network, the first influencing factor feature samples are feature-encoded to obtain the first latent variable distribution features of the first influencing factor feature samples in the target latent variable space, as well as the corresponding latent variable mean and standard deviation; based on the latent variable mean and standard deviation, the first latent variable distribution features are resampled to obtain the second latent variable distribution features; using a decoding network, the second latent variable distribution features are feature-decoded to obtain the second influencing factor feature samples; the pre-constructed latent variable representation model is trained based on the first influencing factor feature samples, the second influencing factor feature samples, the latent variable mean, the latent variable standard deviation, and the second latent variable distribution features corresponding to multiple reference historical periods.

[0081] The reference overhead transmission line can be understood as an overhead transmission line with the same or similar line attributes and operating conditions as the target overhead transmission line, or it can be the target overhead transmission line. The line attributes can include at least one of the following: line topology, number of lines, line type, etc.

[0082] The reference historical period can be understood as a historical period with the same or similar time attribute characteristics as the target historical period. These time attribute characteristics may include at least one of the following: the season in which it belongs, the duration of each historical period, and the time interval between two adjacent historical periods.

[0083] The second latent variable distribution feature can be understood as the same probability distribution feature as the first latent variable distribution feature, but with different latent variable mean and standard deviation.

[0084] In one alternative implementation, the distribution characteristics of the first latent variable can be resampled using reparameterization techniques. Specifically, the distribution characteristics of the second latent variable can be obtained by resampling from the latent variable mean and standard deviation.

[0085] In one alternative implementation, a loss function can be constructed using the first influencing factor feature samples, the second influencing factor feature samples, the latent variable mean, the latent variable standard deviation, and the second latent variable distribution characteristics corresponding to multiple reference historical periods, and the pre-constructed latent variable representation model can be trained based on the loss function.

[0086] In this embodiment, by using data samples of influencing factors affecting health status under multiple consecutive reference historical periods of reference overhead transmission lines, the latent variable representation model is trained. The trained latent variable representation model can compress the target latent variable space of the first influencing factor feature samples corresponding to the influencing factor data samples, and map the high-dimensional features (first influencing factor feature samples) to low-dimensional features (first latent variable distribution features), thereby extracting the main implicit features reflecting changes in health status.

[0087] In an exemplary embodiment, a pre-constructed latent variable representation model is trained based on first influencing factor feature samples, second influencing factor feature samples, latent variable mean, latent variable standard deviation, and second latent variable distribution characteristics corresponding to multiple reference historical periods. This includes: determining a reconstruction loss based on the first and second influencing factor feature samples corresponding to multiple reference historical periods; determining a relative entropy loss based on the latent variable mean and standard deviation corresponding to multiple reference historical periods; for each reference historical period, using the second latent variable distribution characteristics corresponding to the next adjacent reference historical period as positive samples and the second latent variable distribution characteristics corresponding to other reference historical periods (excluding the next adjacent reference historical period) as negative samples, and determining a time comparison loss corresponding to that reference historical period based on the similarity between the positive and negative samples; and training the pre-constructed latent variable representation model based on the reconstruction loss, relative entropy loss, and time comparison losses corresponding to each reference historical period.

[0088] In one alternative embodiment, the reconstruction loss can be constructed based on the sum of the mean squared errors between the first influencing factor feature samples and the second influencing factor feature samples corresponding to each reference historical period.

[0089] In an alternative embodiment, the KL divergence (KLD), i.e., relative entropy loss, can be constructed based on the latent variable mean and standard deviation corresponding to multiple reference historical periods.

[0090] In one alternative embodiment, the time contrast loss for each reference historical period can be determined based on the cosine similarity between positive and negative samples.

[0091] In one optional embodiment, the reconstruction loss, relative entropy loss, and time comparison loss corresponding to each reference historical period are weighted and summed to obtain the total loss, and the pre-built latent variable representation model is trained based on the total loss.

[0092] In this embodiment, the reconstruction capability of the latent variable representation model can be optimized by constructing a reconstruction loss, the rationality of the latent space structure of the latent variable representation model can be optimized by constructing a relative entropy loss, and the time comparison loss can be constructed by using positive and negative samples, enabling the latent variable representation model to perform self-supervised learning without manual annotation, strengthening the time perception capability of the latent variable representation model, and thus enabling it to autonomously learn the continuous change law of health degradation.

[0093] In an exemplary embodiment, the rate of change and acceleration of change of health status of the target overhead transmission line are obtained based on the latent variables of influencing factors corresponding to multiple historical time periods. This includes: for each historical time period, determining the rate of change of health status in that historical time period based on the difference between the latent variables of influencing factors corresponding to the next historical time period and the latent variables of influencing factors corresponding to that historical time period, and the time interval between two adjacent historical time periods; and determining the acceleration of change of health status in that historical time period based on the difference between the rate of change of health status in the next historical time period and the rate of change of health status corresponding to that historical time period, and the time interval.

[0094] In one optional embodiment, for each historical period, the rate of change of health status in that historical period can be determined by the ratio of the difference between the latent variables of influencing factors corresponding to the subsequent historical period and the latent variables of influencing factors corresponding to that historical period, to the time interval between two adjacent historical periods. The acceleration of change of health status in that historical period can be determined by the ratio of the difference between the rate of change of health status in the subsequent historical period and the rate of change of health status corresponding to that historical period, to the time interval between two adjacent historical periods.

[0095] In this embodiment, by determining the rate of change and acceleration of health status in each historical period, the characteristics of the health status of the target overhead transmission line changing over time can be intuitively reflected.

[0096] In one exemplary embodiment, see Figure 4 As shown, S240 includes S410. Wherein:

[0097] S410, input the health status change information and the meteorological forecast data for at least one future time into the trained current carrying capacity prediction model to obtain the current carrying capacity prediction data of the target overhead transmission line at at least one future time; wherein, the current carrying capacity prediction model is trained based on the health status change information of the reference overhead transmission line, the meteorological monitoring data for each historical time, and the current carrying capacity monitoring data for the corresponding historical time.

[0098] Optionally, the load prediction model can be a lightweight multilayer perceptron (MLP) or a temporal network.

[0099] The future timeframe can be preset. For example, a traffic flow prediction model can be used to predict traffic flow data for the next 24 hours to 30 days.

[0100] In this embodiment, the current carrying capacity prediction model can be used to fit the health status changes of the target overhead transmission line and the impact of weather forecast data on the current carrying capacity, thereby obtaining more accurate current carrying capacity prediction data for the target overhead transmission line.

[0101] In one exemplary embodiment, such as Figure 5 As shown, another method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions is provided, including S510~S590. Among them,

[0102] S510: Obtain raw data on the health status of the target overhead transmission line under multiple consecutive historical periods; wherein, the duration of each historical period is the same.

[0103] The raw data includes meteorological data, such as temperature, humidity, and wind speed; load data, such as power grid load and electrical status of transmission lines; and equipment data, such as current, voltage, and temperature of overhead lines.

[0104] S520 preprocesses the raw data to obtain the influencing factor data for each historical period.

[0105] Preprocessing can include data cleaning and data standardization. Data cleaning can include filling in missing values, removing outliers, denoising the data, and detecting outliers. Data standardization can include standardizing or normalizing the input data.

[0106] The influencing factor data may include at least one of the following: influencing factor data of meteorological factors and influencing factor data of line electrical factors. Meteorological factors may include the aforementioned target weather conditions. Line electrical factors may include grid load, current, voltage, line temperature, etc.

[0107] In one alternative embodiment, data on multiple influencing factors can be combined into a time series dataset:

[0108]

[0109] in, This represents time series data consisting of all influencing factors within a total time period spanning multiple historical periods; t represents the timestamp. This represents the visual data (images or video frames) collected during the historical time period t. This represents the tension data collected by the tension sensor during the historical time period t; This represents the current data collected by the current sensor during the historical time period t; This represents meteorological monitoring data collected during historical time period t; This represents temperature data for a historical time period t; This represents wind speed data within a historical time period t; This represents the radiation intensity data for historical time period t.

[0110] S530 extracts features from the influencing factor data for each historical period to obtain the influencing factor features corresponding to the historical period.

[0111] Here, graph convolutional neural networks can be used to extract features from image or video frames to obtain raw visual features. Raw meteorological features can be obtained based on data on the influencing factors of meteorological factors. Raw electrical features can be obtained based on data on the influencing factors of line electrical factors. Then, the raw visual features, raw meteorological features, and raw electrical features can be aggregated and uniformly encoded to obtain the fixed-length influencing factor features corresponding to that historical period. The influencing factor features can be represented as follows:

[0112]

[0113] Where l represents the duration of the historical period; This represents the time range of the i-th historical period; This represents the influencing factor data corresponding to the i-th historical time period; This represents the characteristics of the influencing factors corresponding to the i-th historical period; This represents a feature extraction network.

[0114] S540: Input the characteristics of influencing factors into the trained latent variable representation model to obtain the latent variables of influencing factors corresponding to historical periods.

[0115] The latent variable representation model can be a variational autoencoder (VAE), which includes an encoding network and a decoding network.

[0116] The latent variable representation model is trained according to the following steps:

[0117] Data samples of influencing factors affecting the health status of a reference overhead transmission line were obtained under multiple consecutive reference historical periods, where each reference historical period had the same duration. For each reference historical period, features were extracted from the influencing factor data samples to obtain the first influencing factor feature samples corresponding to that period. Using an encoding network, the first influencing factor feature samples were feature-encoded to obtain the first latent variable distribution characteristics of the first influencing factor feature samples in the target latent variable space, as well as the corresponding latent variable mean and standard deviation. Based on the latent variable mean and standard deviation, the first latent variable distribution characteristics were resampled to obtain the second latent variable distribution characteristics. Using a decoding network, the second latent variable distribution characteristics were feature-decoded to obtain the second influencing factor feature samples. Based on the feature samples of the first and second influencing factors corresponding to multiple reference historical periods, the reconstruction loss is determined; based on the mean and standard deviation of the latent variables corresponding to multiple reference historical periods, the relative entropy loss is determined; for each reference historical period, the distribution features of the second latent variables corresponding to the next reference historical period adjacent to the previous reference historical period are taken as positive samples, and the distribution features of the second latent variables corresponding to other reference historical periods are taken as negative samples. Based on the similarity between the positive and negative samples, the time contrast loss corresponding to the reference historical period is determined; based on the reconstruction loss, the relative entropy loss, and the time contrast loss corresponding to each reference historical period, the pre-constructed latent variable representation model is trained.

[0118] In practice, the influencing factor features input into the latent variable representation model are the influencing factor features corresponding to multiple historical time periods spliced ​​together in chronological order.

[0119] The encoding network transforms the input influencing factor features into the first latent variable distribution feature z in the target latent variable space. The encoding network then outputs the latent variable mean μ and the latent variable standard deviation σ corresponding to z.

[0120] In practice, the reparameterization technique can be used to resample the distribution characteristics z of the first latent variable. Specifically, the distribution characteristics of the second latent variable can be obtained by resampling from the latent variable mean μ and the latent variable standard deviation σ. ;in, It is standard normally distributed noise. This indicates element-wise multiplication.

[0121] The decoding network can perform feature decoding on the distribution characteristics of the second latent variable to obtain feature samples of the second influencing factor: .

[0122] The reconstruction loss, relative entropy loss, and time comparison loss for each reference historical period can be obtained from the following formulas:

[0123]

[0124]

[0125]

[0126]

[0127] in, Indicates the reconstruction loss; This represents the relative entropy loss; This represents the time-comparison loss; N represents the number of reference time periods; This represents the feature sample of the first influencing factor corresponding to the i-th reference historical period; This represents the feature sample of the second influencing factor corresponding to the i-th reference historical period; This represents the distribution characteristics of the second latent variable corresponding to the i-th reference historical period; This represents the distribution characteristics of the second latent variable corresponding to the (i+1)th reference historical period; This represents the distribution characteristics of the second latent variable corresponding to the i-th reference historical period other than the (i+1)-th reference historical period; This indicates that the cosine similarity between two vectors is calculated. Indicates temperature parameter; This represents the mean of the latent variables corresponding to the i-th reference historical period; Let represent the standard deviation of the latent variable corresponding to the i-th reference historical period.

[0128] After obtaining the reconstruction loss, relative entropy loss, and time comparison loss for each reference historical period, they can be weighted and summed to obtain the total loss:

[0129]

[0130] in, Indicates the total loss; This represents the weighting coefficients corresponding to the reconstruction loss; This represents the weighting coefficient corresponding to the relative entropy loss; This represents the weighting coefficient corresponding to the time-comparison loss.

[0131] During the training of a pre-built latent variable representation model using the total loss, the Adaptive Moment Estimation (Adam) optimizer can be used to optimize the parameters in the latent variable representation model using the total loss according to the following formula:

[0132]

[0133] in, This indicates the value of the parameter at time t; This indicates the value of the parameter at time t+1; This represents the learning rate, used to control the step size of the parameters. This represents the gradient of the total loss with respect to the parameters.

[0134] S550, in chronological order, splices and merges the latent variables of influencing factors corresponding to multiple historical periods to obtain health status trajectory information.

[0135] S560, based on the latent variables of influencing factors corresponding to multiple historical periods, obtains the rate of change and acceleration of the health status of the target overhead transmission line.

[0136] In one optional embodiment, for each historical period, the rate of change of health status in a historical period can be determined based on the difference between the latent variables of influencing factors corresponding to the next historical period and the latent variables of influencing factors corresponding to the previous historical period, as well as the time interval between two adjacent historical periods.

[0137] The rate of change in health status can be obtained using the following formula:

[0138]

[0139] In the above formula, Indicates the rate of change in health status; This represents the distribution characteristics of the second latent variable corresponding to historical time period t; This represents the distribution characteristics of the second latent variable corresponding to the historical time period t+△t; It represents the time interval between two adjacent historical periods.

[0140] In one optional embodiment, for each historical period, the acceleration of health status change in a historical period can be determined based on the difference between the rate of change of health status in the subsequent historical period and the rate of change of health status in the corresponding historical period, as well as the time interval.

[0141] The acceleration of changes in health status can be obtained from the following formula:

[0142]

[0143] In the above formula, Indicates the acceleration of changes in health status; This represents the rate of change in health status corresponding to historical time period t; It represents the rate of change in health status corresponding to the historical period t+Δt; Δt represents the time interval between two adjacent historical periods.

[0144] S570: Based on the health status trajectory information, the rate of change of health status, and the acceleration of change of health status, the information on changes in health status is obtained.

[0145] S580, acquire weather forecast data for at least one future time.

[0146] S590, input the health status change information and meteorological forecast data into the trained current carrying capacity prediction model to obtain the current carrying capacity prediction data of the target overhead transmission line at at least one future time; wherein, the current carrying capacity prediction model is trained based on the health status change information of the reference overhead transmission line, the meteorological monitoring data at each historical time, and the current carrying capacity monitoring data at the corresponding historical time.

[0147] In practical implementation, the following formula can be used to obtain the traffic flow prediction data for at least one future time point:

[0148]

[0149] in, Indicates future time The following is the predicted load factor data; This represents the capacity prediction model; Indicates future time The following weather forecast data; It indicates information about changes in health status.

[0150] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0151] Based on the same inventive concept, this application also provides a device for predicting the current carrying capacity of overhead transmission lines under complex operating conditions, used to implement the aforementioned method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the device for predicting the current carrying capacity of overhead transmission lines under complex operating conditions provided below can be found in the limitations of the method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions described above, and will not be repeated here.

[0152] In one exemplary embodiment, such as Figure 6 As shown, a device for predicting the current carrying capacity of overhead transmission lines under complex operating conditions is provided, comprising: an acquisition module 610, an extraction module 620, a first determination module 630, and a second determination module 640, wherein:

[0153] The acquisition module 610 is used to acquire data on the factors affecting the health status of the target overhead transmission line in multiple consecutive historical periods; wherein the duration of each historical period is the same.

[0154] The extraction module 620 is used to extract features from the influencing factor data for each historical period to obtain the influencing factor features corresponding to the historical period.

[0155] The first determining module 630 is used to determine the health status change information of the target overhead transmission line based on the characteristics of influencing factors; wherein, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time.

[0156] The second determining module 640 is used to determine the predicted current carrying capacity of the target overhead transmission line at at least one future time based on information on changes in health status and meteorological forecast data at at least one future time.

[0157] In one embodiment, the first determining module 630 is specifically used for:

[0158] The influencing factor features are input into the trained latent variable representation model to obtain the influencing factor latent variables corresponding to the historical period; the trained latent variable representation model is implemented based on an encoding network and a decoding network;

[0159] According to the time sequence, the latent variables of influencing factors corresponding to multiple historical periods are spliced ​​and fused to obtain health status trajectory information;

[0160] Based on the latent variables of influencing factors corresponding to multiple historical periods, the rate of change and acceleration of change of the health status of the target overhead transmission line are obtained.

[0161] Based on health status trajectory information, health status change rate, and health status change acceleration, health status change information is obtained.

[0162] In one embodiment, the latent variable representation model includes an encoding network and a decoding network;

[0163] The latent variable representation model is trained according to the following steps:

[0164] Data samples of factors affecting the health status of reference overhead transmission lines were obtained under multiple consecutive reference historical periods; wherein, the duration of each reference historical period is the same;

[0165] For each reference historical period, feature extraction is performed on the data samples of influencing factors to obtain the first influencing factor feature sample corresponding to that reference historical period;

[0166] Using an encoding network, feature encoding is performed on the feature samples of the first influencing factor to obtain the distribution characteristics of the first latent variable of the feature samples of the first influencing factor in the target latent variable space, as well as the corresponding latent variable mean and latent variable standard deviation;

[0167] The distribution characteristics of the first latent variable are resampled based on the mean and standard deviation of the latent variable to obtain the distribution characteristics of the second latent variable;

[0168] By using a decoding network, the distribution characteristics of the second latent variable are decoded to obtain the feature samples of the second influencing factor;

[0169] Based on the first influencing factor feature samples, the second influencing factor feature samples, the latent variable mean, the latent variable standard deviation, and the distribution characteristics of the second latent variable corresponding to multiple reference historical periods, the pre-constructed latent variable representation model is trained, and the latent variable representation model is obtained after training.

[0170] In one embodiment, a pre-constructed latent variable representation model is trained based on first influencing factor feature samples, second influencing factor feature samples, latent variable mean, latent variable standard deviation, and second latent variable distribution characteristics corresponding to multiple reference historical periods, including:

[0171] The reconstruction loss is determined based on the first and second influencing factor feature samples corresponding to multiple reference historical periods;

[0172] The relative entropy loss is determined based on the mean and standard deviation of latent variables corresponding to multiple reference historical periods;

[0173] For each reference historical period, the distribution features of the second latent variable corresponding to the next reference historical period adjacent to this reference historical period are taken as positive samples, and the distribution features of the second latent variable corresponding to other reference historical periods other than the next reference historical period are taken as negative samples. Based on the similarity between the positive samples and the negative samples, the time comparison loss corresponding to this reference historical period is determined.

[0174] The pre-built latent variable representation model is trained based on reconstruction loss, relative entropy loss, and time comparison loss corresponding to each reference historical period.

[0175] In one embodiment, based on latent variables of influencing factors corresponding to multiple historical time periods, the rate of change and acceleration of change of the health status of the target overhead transmission line are obtained, including:

[0176] For each historical period, the rate of change of health status in that historical period is determined by the difference between the latent variables of influencing factors corresponding to the next historical period and the latent variables of influencing factors corresponding to that historical period, as well as the time interval between two adjacent historical periods.

[0177] The acceleration of health status change in a given historical period is determined by the difference between the rate of change of health status in the next historical period and the rate of change of health status in the corresponding historical period, as well as the time interval.

[0178] In one embodiment, the second determining module 640 includes:

[0179] By inputting information on changes in health status and meteorological forecast data for at least one future moment into a trained current carrying capacity prediction model, the current carrying capacity prediction data of the target overhead transmission line at at least one future moment can be obtained.

[0180] The current carrying capacity prediction model is trained based on the health status change information of the reference overhead transmission line, meteorological monitoring data at various historical moments, and current carrying capacity monitoring data at the corresponding historical moments.

[0181] Each module in the aforementioned overhead transmission line current carrying capacity prediction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0182] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores relevant data for the overhead transmission line current-carrying capacity prediction method. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements an overhead transmission line current-carrying capacity prediction method.

[0183] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0184] In one exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the overhead transmission line current-carrying capacity prediction method under complex operating conditions provided in any of the above embodiments.

[0185] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the overhead transmission line current-carrying capacity prediction method under complex operating conditions provided in any of the above embodiments.

[0186] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the overhead transmission line current-carrying capacity prediction method under complex operating conditions provided in any of the above embodiments.

[0187] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0188] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0189] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for predicting the current carrying capacity of overhead transmission lines under complex operating conditions, characterized in that, The method includes: Data on factors affecting the health status of a target overhead transmission line in multiple consecutive historical periods are obtained; wherein, the duration of each historical period is the same; the data on factors affecting the health status includes at least one of the following: data on factors affecting meteorological factors and data on factors affecting line electrical factors. For each historical period, feature extraction is performed on the influencing factor data to obtain the influencing factor features corresponding to the historical period; The influencing factor features are input into a trained latent variable representation model to obtain the influencing factor latent variables corresponding to the historical period. The trained latent variable representation model is implemented based on an encoding network and a decoding network. The influencing factor latent variables are the distribution information of the influencing factor features in the target latent space. The influencing factor latent variables include the latent variable distribution features, as well as the latent variable mean and latent variable standard deviation corresponding to the latent variable distribution features. According to the time sequence, the latent variables of influencing factors corresponding to multiple historical periods are spliced ​​and fused to obtain health status trajectory information; Based on the latent variables of influencing factors corresponding to multiple historical periods, the rate of change and acceleration of change of the health status of the target overhead transmission line are obtained. The health status change information is obtained based on the health status trajectory information, the health status change rate, and the health status change acceleration; wherein, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time. Based on the health status change information and the weather forecast data for at least one future time, determine the predicted current carrying capacity of the target overhead transmission line at the at least one future time. The latent variable representation model is trained according to the following steps: Data samples of factors affecting the health status of a reference overhead transmission line are obtained under multiple consecutive reference historical periods; wherein, the duration of each of the reference historical periods is the same; For each reference historical period, feature extraction is performed on the influencing factor data sample to obtain the first influencing factor feature sample corresponding to the reference historical period; Using the coding network, feature encoding is performed on the first influencing factor feature sample to obtain the first latent variable distribution feature of the first influencing factor feature sample in the target latent variable space, as well as the corresponding latent variable mean and latent variable standard deviation; Based on the mean and standard deviation of the latent variables, the distribution characteristics of the first latent variable are resampled to obtain the distribution characteristics of the second latent variable; Using the decoding network, the distribution characteristics of the second latent variable are decoded to obtain the feature samples of the second influencing factor; The reconstruction loss is determined based on the first influencing factor feature samples and the second influencing factor feature samples corresponding to the plurality of reference historical periods; The relative entropy loss is determined based on the mean and standard deviation of the latent variables corresponding to the plurality of reference historical periods; For each reference historical period, the distribution features of the second latent variable corresponding to the next reference historical period adjacent to the reference historical period are taken as positive samples, and the distribution features of the second latent variable corresponding to other reference historical periods besides the next reference historical period are taken as negative samples. The time comparison loss corresponding to the reference historical period is determined based on the similarity between the positive samples and the negative samples. The pre-built latent variable representation model is trained based on the reconstruction loss, the relative entropy loss, and the time comparison loss corresponding to each of the reference historical periods.

2. The method according to claim 1, characterized in that, The step of obtaining the rate of change and acceleration of change of the health status of the target overhead transmission line based on the latent variables of influencing factors corresponding to multiple historical time periods includes: For each historical period, the rate of change of health status in the historical period is determined based on the difference between the latent variables of influencing factors corresponding to the next historical period and the latent variables of influencing factors corresponding to the historical period, as well as the time interval between two adjacent historical periods. The acceleration of health status change in the historical period is determined based on the difference between the rate of change of health status in the subsequent historical period and the rate of change of health status in the corresponding historical period, as well as the time interval.

3. The method according to claim 1 or 2, characterized in that, The step of determining the predicted current-carrying capacity of the target overhead transmission line at the at least one future time based on the health status change information and weather forecast data for at least one future time includes: The health status change information and the meteorological forecast data at the least one future time are input into the trained current carrying capacity prediction model to obtain the current carrying capacity prediction data of the target overhead transmission line at the at least one future time. The current carrying capacity prediction model is trained based on the health status change information of the reference overhead transmission line, meteorological monitoring data at various historical moments, and current carrying capacity monitoring data at the corresponding historical moments.

4. A device for predicting the current carrying capacity of overhead transmission lines under complex operating conditions, characterized in that, The device includes: The acquisition module is used to acquire data on factors affecting the health status of a target overhead transmission line in multiple consecutive historical periods; wherein, the duration of each historical period is the same; the data on factors affecting the health status includes at least one of the following: data on factors affecting meteorological factors and data on factors affecting line electrical factors; The extraction module is used to extract features from the influencing factor data for each historical period to obtain the influencing factor features corresponding to the historical period. The first determining module is used to input the influencing factor features into a trained latent variable representation model to obtain the influencing factor latent variables corresponding to the historical time periods; the trained latent variable representation model is implemented based on an encoding network and a decoding network; the influencing factor latent variables are the distribution information of the influencing factor features in the target latent space; the influencing factor latent variables include latent variable distribution features, and the latent variable mean and latent variable standard deviation corresponding to the latent variable distribution features; according to the time sequence, the influencing factor latent variables corresponding to multiple historical time periods are spliced ​​and fused to obtain health status trajectory information; based on the influencing factor latent variables corresponding to multiple historical time periods, the rate of change of health status and the acceleration of change of health status of the target overhead transmission line are obtained; based on the health status trajectory information, the rate of change of health status, and the acceleration of change of health status, the health status change information is obtained; wherein, the health status change information is used to characterize the relationship between the health status of the target overhead transmission line and time. The second determining module is used to determine the predicted current carrying capacity of the target overhead transmission line at the at least one future time based on the health status change information and the meteorological forecast data at at least one future time. The latent variable representation model is trained according to the following steps: acquiring data samples of influencing factors affecting the health status of a reference overhead transmission line under multiple consecutive reference historical periods; wherein, the duration of each reference historical period is the same; for each reference historical period, performing feature extraction on the influencing factor data samples to obtain the first influencing factor feature samples corresponding to the reference historical period; using the encoding network, performing feature encoding on the first influencing factor feature samples to obtain the first latent variable distribution features of the first influencing factor feature samples in the target latent variable space, as well as the corresponding latent variable mean and latent variable standard deviation; resampling the first latent variable distribution features based on the latent variable mean and latent variable standard deviation to obtain the second latent variable distribution features; using the decoding network, performing feature decoding on the second latent variable distribution features to obtain... The process involves obtaining feature samples of the second influencing factor; determining a reconstruction loss based on the feature samples of the first and second influencing factors corresponding to the plurality of reference historical periods; determining a relative entropy loss based on the mean and standard deviation of the latent variables corresponding to the plurality of reference historical periods; for each reference historical period, taking the distribution features of the second latent variables corresponding to the next reference historical period adjacent to the reference historical period as positive samples, and taking the distribution features of the second latent variables corresponding to other reference historical periods besides the next reference historical period as negative samples, and determining the time contrast loss corresponding to the reference historical period based on the similarity between the positive samples and the negative samples; and training a pre-constructed latent variable representation model based on the reconstruction loss, the relative entropy loss, and the time contrast loss corresponding to each reference historical period.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

7. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 3.

Images