Load prediction method and device for power transmission line, storage medium, processor and program product

Through hidden variable conversion and enhancement processing, combined with mean and variance calculation network, the problem of uneven sample distribution in transmission line load prediction is solved, and the prediction accuracy and efficiency are improved.

CN118970899APending Publication Date: 2024-11-15GUANGDONG POWER GRID CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411007946.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing transmission line load prediction methods fail to effectively consider the problem of uneven distribution of training samples, resulting in low prediction accuracy.

Method used

By obtaining the historical load data of the transmission line, performing hidden variable conversion and enhancement processing, a short-term load prediction model based on hidden variable transformation and enhancement is constructed, and a mean calculation network and variance calculation network are used to perform hidden variable sampling and transformation, combining Gaussian distribution and important Sex sampling technology improves the uneven distribution of training samples.

Benefits of technology

It improves the accuracy of power transmission line load prediction, enhances the prediction ability of categories with small sample size, reduces training time, and achieves end-to-end prediction results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118970899A_ABST
    Figure CN118970899A_ABST
Patent Text Reader

Abstract

The invention discloses a load prediction method and device for a power transmission line, a storage medium, a processor and a program product. The method comprises the steps that multiple pieces of initial load data of a power transmission line in a first historical period are acquired, and the first historical period comprises multiple sub-periods; converting at least one piece of initial load data corresponding to each sub-period in the plurality of pieces of initial load data to obtain a first hidden variable corresponding to each sub-period; enhancing the first hidden variable to obtain an enhanced hidden variable; and predicting the plurality of enhanced hidden variables corresponding to the plurality of sub-periods to obtain a plurality of target load data of the power transmission line in the future period. The technical problem of low prediction accuracy of the load of the power transmission line is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of power transmission lines, and in particular to a method, device, storage medium, processor and program product for load prediction of power transmission lines. Background Art

[0002] At present, transmission lines are widely distributed, and it is of great significance to conduct accurate short-term load forecasting on the load data of transmission lines.

[0003] In the related art, the load data of transmission lines is usually predicted based on deep learning models, which extracts time-correlation features from the load data to achieve the purpose of predicting the load data of transmission lines. However, this method does not take into account the impact of the uneven distribution of training samples on the prediction results, and there is a problem of low prediction accuracy of the load of power lines.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] Embodiments of the present invention provide a method, device, storage medium, processor and program product for load prediction of a power transmission line, so as to at least solve the technical problem of low prediction accuracy of the load of the power transmission line.

[0006] According to one aspect of an embodiment of the present invention, a method for load prediction of a transmission line is provided, which may include: obtaining multiple initial load data of the transmission line in a first historical period, wherein the first historical period includes multiple sub-periods; converting at least one initial load data corresponding to each sub-period among the multiple initial load data to obtain a first latent variable corresponding to each sub-period; performing enhancement processing on the first latent variable to obtain an enhanced latent variable; and predicting multiple enhanced latent variables corresponding to multiple sub-periods to obtain multiple target load data of the transmission line in future periods.

[0007] Optionally, at least one initial load data corresponding to each sub-period among multiple initial load data is converted to obtain a first latent variable corresponding to each sub-period, including: determining a first mean and a first variance value corresponding to at least one initial load data to obtain a first mean and a first variance value corresponding to each sub-period; constructing a first Gaussian distribution of multiple first means and multiple first variance values; sampling the first Gaussian distribution to obtain a first latent variable corresponding to each sub-period in the first historical period to obtain multiple first latent variables.

[0008] Optionally, the first latent variable is enhanced to obtain an enhanced latent variable, including: determining the total load of a sub-period based on at least one initial load data corresponding to the sub-period; obtaining a total load set of the first historical period based on the total loads of multiple sub-periods; dividing the total load set to obtain multiple sub-load sets; respectively determining target loads in multiple sub-load sets to obtain multiple target loads; determining target latent variables corresponding to multiple target loads in the first latent variable to obtain multiple target latent variables; and converting multiple target latent variables to obtain multiple enhanced latent variables.

[0009] Optionally, the total load set is divided to obtain multiple sub-load sets, including: determining the maximum value and the minimum value in the total load set; and dividing the total load set based on the maximum value and the minimum value to obtain multiple sub-load sets.

[0010] Optionally, target loads in multiple sub-load sets are determined respectively to obtain multiple target loads, including: determining a first value and a second value in the sub-load set, wherein the first value is greater than the second value; and using the first value and the second value to filter out the target loads in the sub-load set.

[0011] Optionally, multiple target latent variables are transformed to obtain multiple enhanced latent variables, including: determining the second mean corresponding to the target latent variable and the covariance corresponding to the target latent variable based on the number of initial load data in the sub-load set; obtaining a flowing mean value set and a flowing covariance set; using a Gaussian filter to perform a convolution operation on the flowing mean value and the flowing covariance to obtain a sliding mean value and a sliding covariance; determining the difference between the sliding mean value and the second mean value as a transformed mean value; determining the difference between the sliding covariance and the covariance as a transformed covariance; constructing a second Gaussian distribution of the transformed mean value and the transformed covariance; and sampling the second Gaussian distribution to obtain enhanced latent variables.

[0012] Optionally, multiple enhanced latent variables corresponding to multiple sub-periods are predicted to obtain multiple target load data of the transmission line in future time periods, including: using a prediction model to predict multiple enhanced latent variables to obtain target load data, wherein the prediction model is trained based on multiple historical load data of the transmission line in a second historical period, and the second historical period is earlier than the first historical period.

[0013] Optionally, the method may also include: obtaining multiple historical load data of the transmission line; constructing an input feature data set and a prediction data set based on the multiple historical load data; obtaining meteorological data of the transmission line in a second historical period; constructing the meteorological data and the input feature data set to obtain an input data set; and training a prediction model using the input data set and the prediction data set.

[0014] Optionally, a prediction model is obtained by training the input data set and the prediction data set, including: extracting the input data set using a long short-term memory network to obtain multiple second latent variables; determining the third latent variable of the input data set using a mean calculation network and a variance calculation network; enhancing the third latent variable to obtain a fourth latent variable; constructing a target equation based on the second latent variable, the third latent variable and the prediction data set; determining the model parameters of the prediction model using the target equation; and adjusting the initial model parameters of the sub-prediction model according to the model parameters to obtain the prediction model.

[0015] According to another aspect of an embodiment of the present invention, a load prediction device for a transmission line is also provided, and the device may include: an acquisition unit, used to acquire multiple initial load data of the transmission line in a first historical period, wherein the first historical period includes multiple sub-periods; a conversion unit, used to convert at least one initial load data corresponding to each sub-period among the multiple initial load data, to obtain a first latent variable corresponding to each sub-period; a processing unit, used to enhance the first latent variable to obtain an enhanced latent variable; and a prediction unit, used to predict the multiple enhanced latent variables corresponding to the multiple sub-periods, to obtain multiple target load data of the transmission line in future periods.

[0016] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the load prediction method for the transmission line of an embodiment of the present invention.

[0017] According to another aspect of an embodiment of the present invention, a processor is further provided. The processor is used to run a program, wherein when the program is run by the processor, the method for load prediction of a power transmission line according to an embodiment of the present invention is executed.

[0018] According to another aspect of an embodiment of the present invention, a program product is further provided. The program product includes computer instructions, wherein the computer instructions, when executed by a processor, implement the load forecasting method for a power transmission line according to an embodiment of the present invention.

[0019] In an embodiment of the present invention, a plurality of initial load data of a transmission line in a first historical period are obtained, wherein the first historical period includes a plurality of sub-periods; at least one initial load data corresponding to each sub-period among the plurality of initial load data is converted to obtain a first latent variable corresponding to each sub-period; the first latent variable is enhanced to obtain an enhanced latent variable; and the plurality of enhanced latent variables corresponding to the plurality of sub-periods are predicted to obtain a plurality of target load data of the transmission line in future periods. That is, in an embodiment of the present invention, when predicting the load of the transmission line at a future moment, a plurality of initial load data in the first historical period can be obtained, a plurality of first latent variables corresponding to the plurality of initial load data can be determined, and the plurality of first latent variables can be enhanced to obtain a plurality of enhanced latent variables. By predicting the plurality of latent variables, a plurality of target loads in future periods can be obtained, thereby achieving the technical effect of improving the accuracy of load prediction of the transmission line and solving the technical problem of low prediction accuracy of the load of the transmission line. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0021] Figure 1 is a flow chart of a method for load prediction of a power transmission line according to an embodiment of the present invention;

[0022] Figure 2 is a flow chart of constructing a short-term load forecasting model for a power transmission line based on latent variable transformation and enhancement according to an embodiment of the present invention;

[0023] Figure 3 is a schematic diagram of an overall framework of a transmission line short-term load forecasting model based on latent variable transformation and enhancement according to an embodiment of the present invention;

[0024] Figure 4 is a schematic diagram of a load prediction device for a power transmission line according to an embodiment of the present invention;

[0025] Figure 5 is a structural block diagram of a computer terminal according to an embodiment of the present invention;

[0026] Figure 6 It is a block diagram of an electronic device according to a method for detecting performance of an operating system according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0029] Example 1

[0030] According to an embodiment of the present invention, an embodiment of a method for load forecasting of a transmission line is provided, and the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and, although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0031] In this embodiment, a method for load prediction of a transmission line is proposed. When predicting the load of the transmission line at a future moment, the method can obtain multiple initial load data in a first historical period, determine multiple first latent variables corresponding to the multiple initial load data, and enhance the multiple first latent variables to obtain multiple enhanced latent variables. By predicting the multiple latent variables, multiple target loads in the future period are obtained, thereby achieving the technical effect of improving the accuracy of load prediction of the transmission line and solving the technical problem of low prediction accuracy of the load of the transmission line.

[0032] Figure 1 FIG. 1 is a flow chart of a method for load prediction of a transmission line according to an embodiment of the present invention. Figure 1 As shown, the method may include the following steps:

[0033] Step S102: Acquire a plurality of initial load data of the transmission line in a first historical period, wherein the first historical period includes a plurality of sub-periods.

[0034] In the technical solution provided in the above step S102 of the present invention, multiple initial loads of the transmission line in the first historical period can be obtained. The above first historical period can be one day, two days or one week. Multiple sub-periods can be multiple hours in a day, every day of a week, etc. This is only an example, and there is no specific restriction on the length of the first historical period and the length of the sub-period. The initial load data can be historical load data, which can be used to characterize the amount of load carried on the output line, can be a transmission line load, can be used to determine the load demand of the power supply system, and can include different types of load data such as industrial electricity, commercial electricity, and household electricity.

[0035] Optionally, before performing load data prediction of a transmission line, a plurality of initial load data of the transmission line in a first historical period may be obtained.

[0036] Step S104: convert at least one initial load data corresponding to each sub-period among the multiple initial load data to obtain a first latent variable corresponding to each sub-period.

[0037] In the technical solution provided in step S104 of the present invention, each sub-period contains at least one initial load data. The at least one initial load data can be converted to obtain a first latent variable corresponding to each sub-period. The first latent variable can be represented by z′ or z′ (1) It is indicated that this is only an example and there is no specific limitation on the expression form of the first latent variable.

[0038] For example, the initial load data for July can be obtained to obtain multiple initial load data. Each day in July corresponds to a sub-period, so there are 31 sub-periods in total, each of which can include 24 initial load data, or can include only one. There is no specific limit on the number of initial load data in each sub-period. Based on the conversion of at least one initial load data corresponding to each sub-period, the first latent variable corresponding to each sub-period can be obtained.

[0039] Step S106, performing enhancement processing on the first latent variable to obtain an enhanced latent variable.

[0040] In the technical solution provided in step S106 of the present invention, the first latent variable may be enhanced to obtain an enhanced latent variable. The enhanced latent variable may be a transformation enhanced latent variable, which may be The low-dimensional vector space of the initial load data can be used to represent , and can be used to capture the potential features and structures in the initial load data. It should be noted that this is only an example, and there is no specific limitation on the method of enhancing latent variables.

[0041] Optionally, this embodiment can transform and enhance the first latent variable. On the one hand, it can effectively calibrate the input features (i.e., initial load data) from the latent space, and on the other hand, it can add latent variables of categories with small sample sizes in the latent space, thereby improving the adverse effects of the imbalanced distribution of training samples on short-term load forecasting results, thereby achieving the technical effect of improving the accuracy of load forecasting for transmission lines and solving the technical problem of low accuracy in load forecasting for transmission lines.

[0042] Step S108, predicting multiple enhanced latent variables corresponding to multiple sub-periods to obtain multiple target load data of the transmission line in the future period.

[0043] In the technical solution provided in step S108 of the present invention, after obtaining multiple enhanced latent variables, multiple enhanced latent variables corresponding to multiple sub-periods can be predicted to obtain multiple target load data of the transmission line in the future period. The future period can be a week, a day or a month, which is only for example, and the size of the future period is not specifically limited.

[0044] For example, the obtained multiple initial load data x * (that is, load characteristics) are input into the mean calculation network and the variance calculation network to obtain the first mean μ * And the first variance value σ * , construct a first Gaussian distribution of multiple first means and multiple first variance values, and obtain S first latent variables z′={z′ based on the Bayesian rule and importance sampling (1) , z′ (2) , ..., z′ (s)}. The first latent variable can be transformed to obtain the enhanced latent variable Based on the trained prediction model (i.e., predictor), the enhanced latent variables are processed to obtain the target load data (ie, predicted value).

[0045] Through the above-mentioned step S102 and step S108 of the present invention, a plurality of initial load data of the transmission line in the first historical period are obtained, wherein the first historical period includes a plurality of sub-periods; at least one initial load data corresponding to each sub-period in the plurality of initial load data is converted to obtain a first latent variable corresponding to each sub-period; the first latent variable is enhanced to obtain an enhanced latent variable; the plurality of enhanced latent variables corresponding to the plurality of sub-periods are predicted to obtain a plurality of target load data of the transmission line in the future period. That is to say, in the embodiment of the present invention, when predicting the load of the transmission line at a future moment, a plurality of initial load data in the first historical period can be obtained, a plurality of first latent variables corresponding to the plurality of initial load data can be determined, and a plurality of first latent variables can be enhanced to obtain a plurality of enhanced latent variables, and a plurality of target loads in the future period can be obtained by predicting the plurality of latent variables, thereby achieving the technical effect of improving the accuracy of load prediction of the transmission line and solving the technical problem of low prediction accuracy of the load of the transmission line.

[0046] The above method of this embodiment is further introduced below.

[0047] As an optional implementation, step S104, converts at least one initial load data corresponding to each sub-period among multiple initial load data to obtain a first latent variable corresponding to each sub-period, including: determining a first mean and a first variance value corresponding to at least one initial load data to obtain a first mean and a first variance value corresponding to each sub-period; constructing a first Gaussian distribution of multiple first means and multiple first variance values; sampling the first Gaussian distribution to obtain a first latent variable corresponding to each sub-period in the first historical period, and obtaining multiple first latent variables.

[0048] In this embodiment, during the test phase, at least one initial load data may be processed by a pre-trained mean calculation network to determine a first mean corresponding to the at least one initial load data, and at least one initial load data may be processed by a variance calculation network to determine a first variance value corresponding to the at least one initial load data. In the above manner, a first mean and a first variance value corresponding to each sub-period may be obtained, and a first Gaussian distribution of multiple first variance values ​​and multiple first means may be further constructed. The first latent variable corresponding to each sub-period in the first historical period may be obtained by sampling the first Gaussian distribution, thereby obtaining multiple first latent variables.

[0049] In the training phase, the input feature x and target value y in the input feature data set can be input into the mean calculation network to obtain the mean μ = C μ(x, y; φ1). φ1 can be used to characterize the parameters of the mean calculation network, which can be a pre-set value or a randomly generated value. There is no specific restriction on the determination method of the above parameters. The input feature x and the target value y can be input into the variance calculation network to obtain the variance σ = C σ (x, y; φ2). φ2 can be the parameter of the variance calculation network. Furthermore, based on the mean μ and variance σ, the Bayesian rule and importance sampling can be used , to get the third latent variable

[0050] After training the mean calculation network and the variance calculation network, in the test phase, multiple initial load data can be obtained and transferred to the trained mean calculation network and the variance calculation network to obtain the first mean and the first variance value corresponding to each sub-period, and obtain multiple first means and multiple first variance values. A first Gaussian distribution of multiple first means and multiple first variance values ​​is constructed, and the first Gaussian distribution can be processed using the Bayesian rule and importance sampling to obtain a first latent variable.

[0051] As an optional implementation, step S106, enhancing the first latent variable to obtain an enhanced latent variable, including: determining the total load of a sub-period based on at least one initial load data corresponding to the sub-period; obtaining a total load set of the first historical period based on the total loads of multiple sub-periods; dividing the total load set to obtain multiple sub-load sets; respectively determining the target loads in the multiple sub-load sets to obtain multiple target loads; determining the target latent variables corresponding to the multiple target loads in the first latent variable to obtain multiple target latent variables; converting the multiple target latent variables to obtain multiple enhanced latent variables.

[0052] In this embodiment, the total load of the sub-period can be determined based on at least one initial load data corresponding to the sub-period to obtain the total load of multiple sub-periods. The total load can be the sum of at least one initial load data and can be determined by the following formula:

[0053]

[0054] in, It can be used to represent the total load at the i-th sub-moment. (i) It can be used to characterize the initial load data at the i-th moment.

[0055] Optionally, based on the total loads of multiple sub-times, the total load set of the first historical period can be obtained: The total load set can be divided to obtain multiple sub-load sets; the target loads (yl ) to obtain multiple target loads; determine the target latent variables (z′) corresponding to the multiple target loads in the first latent variable l ), and obtain multiple target latent variables; transform the multiple target latent variables to obtain multiple enhanced latent variables.

[0056] As an optional implementation, the total load set is divided to obtain multiple sub-load sets, including: determining the maximum value and the minimum value in the total load set; and dividing the total load set based on the maximum value and the minimum value to obtain multiple sub-load sets.

[0057] In this embodiment, the maximum value and the minimum value in the total load set are determined, and the total load set can be divided based on the maximum value and the minimum value to obtain multiple sub-load sets. The sub-load sets can be partitions of the target value set during the training process.

[0058] Optionally, determine the total load set (y s ) and minimum value , calculate the interval size of the sub-load set partition , and divide the total load set into L parts, that is, the size of each interval after the total load set is divided is .

[0059] For example, for the total load set: y s ={25, 28, 29, 30, 32, 29, 36}, where the maximum value is 36 and the minimum value is 20. r = 16. If L is set to 4, that is, the size of each interval is 16 / 4 = 4, then the multiple sub-load sets can be y 1 ={20},y 2 ={25, 28}, y 3 ={39, 29, 30, 32}, y 4 ={36}.

[0060] As an optional implementation, target loads in multiple sub-load sets are determined separately to obtain multiple target loads, including: determining a first value and a second value in the sub-load set, wherein the first value is greater than the second value; and using the first value and the second value to filter out the target loads in the sub-load set.

[0061] In this embodiment, a first value and a second value in the sub-load set are determined, wherein the first value may be an upper limit value in the sub-load set, which may be referred to as the upper limit. The second value may be a lower limit value in the sub-load set, which may be referred to as the lower limit. The first value may be greater than the second value. Using the first value and the second value, a target load in the sub-load set may be determined.

[0062] Optionally, determine the upper and lower limits of each sub-load set, and the lower limit of the lth sub-load set is Upper limit is Based on the upper and lower limits, the target loads distributed in each sub-load set are counted separately. Where l∈{1, 2, ..., L}.

[0063] As an optional implementation, multiple target latent variables are transformed to obtain multiple enhanced latent variables, including: determining the second mean corresponding to the target latent variable and the covariance corresponding to the target latent variable based on the number of initial load data in the sub-load set; obtaining a flowing mean value set and a flowing covariance set; using a Gaussian filter to perform a convolution operation on the flowing mean value and the flowing covariance to obtain a sliding mean value and a sliding covariance; determining the difference between the sliding mean value and the second mean value as a transformed mean value; determining the difference between the sliding covariance and the covariance as a transformed covariance; constructing a second Gaussian distribution of the transformed mean value and the transformed covariance; and sampling the second Gaussian distribution to obtain enhanced latent variables.

[0064] In this embodiment, the target load in each sub-load set is determined, and the first latent variable (z′) corresponding to the target load is determined. l ), and determine the number of samples in each sub-load set (N l ). Calculate the first latent variable corresponding to at least one target load in the sub-load set to obtain at least one first latent variable. Determine the number of initial load data (N) in the sub-load set l ), based on the number of initial load data, the second mean (m l ), and the covariance corresponding to the target latent variable (c l ), the second mean (also called average value) and covariance can be calculated by the following formula:

[0065]

[0066]

[0067] Optionally, a moving average set can be set in advance ( ) and the flow covariance set ( ), where t is the current iteration number, the initial moving average m0 and the moving covariance c0 can be randomly generated or preset, and there is no specific restriction on the determination method of the initial moving average and the moving covariance. After obtaining the initial load data, the moving average set and the moving covariance set can be obtained; during the iteration process, the moving average ( ) and flow covariance ( ) is calculated as follows:

[0068]

[0069]

[0070] in, It can be pre-generated randomly. L can be the lth interval. t can be the iteration here. α can be a random number in [0, 1].

[0071] Optionally, this embodiment may set a Gaussian filter (G σ (u, v)), as shown below:

[0072]

[0073] Where u and v can be integers in [-z, ..., z],

[0074] Optionally, the moving mean and the moving covariance can be convolved using a Gaussian filter to obtain a sliding mean and sliding covariance , as shown below:

[0075]

[0076]

[0077] Optionally, after determining the sliding mean and sliding covariance, the transformed mean belonging to the lth interval can be obtained , and the transformed covariance , and construct a Gaussian distribution . Determine the number of samples in each time interval {N1, N2, ..., N l The maximum value among them is N max , based on Bayes' rule and using importance sampling from a Gaussian distribution In the above example, N max Hidden variables Complete the latent variable enhancement belonging to the lth interval.

[0078] Optionally, repeat the above steps for L times to obtain the enhanced latent variable

[0079] For example, the load set y s ={25,28,29,30,32,20,29,36}, the maximum value is 36, the minimum value is 20, and the interval size r = 16. Assuming L is set to 4, that is, the size of each interval is 16 / 4 = 4, then the sub-load sets are , we can further determine the number of samples in each sub-load set {N1, N2, N3, N4} = {1, 2, 4, 1}, and then sample 4 latent variables from the Gaussian distribution corresponding to each interval, and then complete the latent variable enhancement. It should be noted that the size of the above numbers and letters are only for example and can be changed according to actual conditions, and no specific restrictions are made here.

[0080] As an optional implementation, step S108 predicts multiple enhanced latent variables corresponding to multiple sub-periods to obtain multiple target load data of the transmission line in the future period, including: using a prediction model to predict multiple enhanced latent variables to obtain target load data, wherein the prediction model is trained based on multiple historical load data of the transmission line in the second historical period, and the second historical period is earlier than the first historical period.

[0081] In this embodiment, a prediction model can be used to predict multiple enhanced latent variables to obtain target load data. The prediction model can be trained based on multiple historical load data of the power line in the second historical period and can be a predictor. The second historical period is earlier than the first historical period.

[0082] Optionally, a prediction model may be trained in advance based on a plurality of historical load data, and after the enhanced latent variables are acquired, the prediction model may be used to predict the plurality of enhanced latent variables to obtain a plurality of target load data.

[0083] As an optional implementation, the method may also include: obtaining multiple historical load data of the transmission line; constructing an input feature data set and a prediction data set based on the multiple historical load data; obtaining meteorological data of the transmission line in a second historical period; constructing the meteorological data and the input feature data set to obtain an input data set; and training a prediction model using the input data set and the prediction data set.

[0084] In this embodiment, the historical load data of the transmission line recorded in the second historical period can be collected. ,in, Can be used to represent the number of days recorded. Can be used to characterize daily load data. The resolution that can be used to characterize the load data can be, for example, 24, 48, or 96.

[0085] Optionally, an input feature data set and a prediction data set can be constructed based on multiple historical load data. The prediction data set can be a target value (y (i) =L (i+1) ),in , which can be used to determine the forecast value for the next day.

[0086] Optionally, meteorological data for the second historical period may also be collected, and the meteorological data may include the maximum temperature of each day. , minimum temperature , average temperature , daily relative humidity , daily rainfall , etc. It should be noted that this is only an example and there is no specific restriction on the type and form of expression of meteorological data.

[0087] Optionally, the input dataset can be constructed using the meteorological data and the input feature dataset , the input data set can be the input feature.

[0088] For example, assuming that the resolution of the historical load data is 24, and the second historical period can be from July 7 to July 8, then the load value for 24 hours on July 7 is obtained to obtain the input feature data set, and the meteorological data of the transmission line in the second historical period, such as the highest temperature, the lowest temperature, the daily average temperature, the daily relative humidity, the daily rainfall and other data, can be obtained. The input data set can be constructed based on the input feature data set and the meteorological data. The load value for 24 hours on July 8 can be determined as the prediction data set. The input data set is used as the input data, and the prediction data set is used as the target value (that is, the output data). The above data is used to train the prediction model.

[0089] As an optional implementation, a prediction model is obtained by training an input data set and a prediction data set, including: extracting the input data set using a long short-term memory network to obtain multiple second latent variables; determining the third latent variable of the input data set using a mean calculation network and a variance calculation network; enhancing the third latent variable to obtain a fourth latent variable; constructing a target equation based on the second latent variable, the fourth latent variable and the prediction data set; determining the model parameters of the prediction model using the target equation; and adjusting the initial model parameters of the sub-prediction model according to the model parameters to obtain the prediction model.

[0090] In this embodiment, the long short-term memory network may be a three-layer long short-term memory (LSTM) neural network. The number of neurons in the first layer of the neural network may be 16, the number of neurons in the second layer may be 32, and the number of neurons in the third layer may be 64. It should be noted that the size of the above number of neurons is only for example and is not specifically limited here.

[0091] Optionally, the above-mentioned long short-term memory network can also be replaced by a neural network such as a recurrent neural network, a gated recurrent neural network, and a variant of a recurrent neural network. There is no specific limitation on the model for obtaining the second latent variable.

[0092] The prior latent variable extraction model three-layer long short-term memory neural network can be used to obtain the second latent variable, and a three-layer long short-term memory neural network can be pre-built. The number of neurons in the first layer of the neural network can be 16, the number of neurons in the second layer can be 32, and the number of neurons in the third layer can be 64.

[0093] Optionally, after the input data set is constructed, the input data set can be extracted through a long short-term memory network to obtain multiple second latent variables. For example, the LSTM neural network can obtain multiple second latent variables (z) through the following formula:

[0094] z=LSTM(x;θ LSTM )

[0095] in, can be used to characterize the second latent variable, can be used to characterize the input data set, θ LSTM It can be used to characterize the parameters of LSTM neural network.

[0096] Optionally, the LSTM neural network can be replaced by a recurrent neural network, a gated recurrent neural network, a variant of a recurrent neural network, and the like. No specific limitation is imposed on the model for obtaining the second latent variable.

[0097] In this embodiment, a two-layer fully connected neural network can also be established, wherein the number of neurons in the first layer of the two-layer fully connected neural network can be 64, and the number of neurons in the second layer can be the resolution of the load data. Furthermore, the load forecast value can be obtained by using the above fully connected neural network ( ):

[0098]

[0099] Among them, θ FC It can be used to characterize the parameters of the fully connected neural network FC.

[0100] Furthermore, based on the load forecast value, the three-layer long short-term memory neural network can be iteratively updated using the stochastic gradient descent optimization algorithm (Adam) optimizer and the target value y, and finally the optimal LSTM neural network parameters are obtained. , the prior latent variable extraction model obtained by training can be used to accurately obtain the second latent variable.

[0101] For example, using the Adam optimization algorithm to optimize the model can With the predicted value For comparison, the model parameters θ LSTM and θ FC Update to obtain the trained prior latent variable extraction model.

[0102] In this embodiment, the mean calculation network and the variance calculation network can be used to determine the third latent variable of the input data set (it can also be used as denoted), may include the following steps: the input feature x and the target value y in the input feature data set may be input into the mean calculation network to obtain the mean μ=C μ (x, y; φ1). φ1 can be used to characterize the parameters of the mean calculation network, which can be a pre-set value or a randomly generated value. The method for determining the above parameters is not specifically limited here. The mean calculation network can include but is not limited to LSTM neural networks, recurrent neural networks, gated recurrent neural networks, and variants of recurrent neural networks. The type of mean calculation network is not specifically limited here. The input feature x and the target value y can be input into the variance calculation network to obtain the variance σ=C σ (x, y; φ2). φ2 may be a parameter of a variance calculation network, and the variance calculation network may include but is not limited to a LSTM neural network, a recurrent neural network, a gated recurrent neural network, and a variant of a recurrent neural network. The type of the variance calculation network is not specifically limited here.

[0103] Furthermore, based on the mean μ and variance σ, we can use the Bayesian rule and importance sampling to perform a multi-mean and multi-variance Gaussian distribution ( ) is sampled to obtain the third latent variable

[0104] In this embodiment, the third latent variable can be enhanced by the enhancement method of the first latent variable described above to obtain the fourth latent variable.

[0105] For example, determine the target load in each sub-load set of the input data set, determine the third latent variable corresponding to the target load, and determine the number of samples (N) in each sub-load set. l ). Calculate the third latent variable corresponding to at least one target load in the sub-load set to obtain at least one third latent variable. Determine the number of historical load data in the sub-load set. Based on the number of historical load data, the mean corresponding to the third latent variable and the covariance corresponding to the third latent variable can be determined. A moving average set and a moving covariance set can be set in advance. A Gaussian filter can be used to perform a convolution operation on the moving average and the moving covariance to obtain a sliding average. and sliding covariance , further, we can get the transformed mean and transformed covariance belonging to the lth interval, and construct a Gaussian distribution based on the transformed mean and transformed covariance . Determine the number of samples in each time interval {N1, N2, ..., N L The maximum value among them is N max , based on Bayes' rule and using importance sampling from a Gaussian distribution In the above example, N max The latent variable belonging to the lth interval is enhanced by a latent variable, and the sampling is repeated L times according to the above steps to obtain the fourth latent variable.

[0106] In this embodiment, based on the second latent variable, the fourth latent variable and the prediction data set, a target equation can be constructed; using the target equation, the model parameters of the prediction model are determined; according to the model parameters, the initial model parameters of the sub-prediction model are adjusted to obtain the prediction model. The above target equation can be constructed based on multiple optimization problems.

[0107] Optionally, a discriminator D = {D1, D2, ..., D L}, by constructing the discriminator, the second latent variable (z) and the fourth latent variable ( ) to distinguish the second latent variable (z) and the fourth latent variable ( ) come from the same distribution. , the discriminator D can be a two-layer fully connected neural network, and the activation function can be ReLU. Furthermore, based on the discriminator, an optimization problem can be constructed to adjust the distribution of latent variables. The optimization problem for the latent variable belonging to the lth interval is defined as follows:

[0108]

[0109] Among them, φ={φ1, φ2}.

[0110] Optionally, considering that the optimization of L discriminators will become a multi-objective optimization problem, which increases the difficulty and time of solving, therefore, in order to reduce the difficulty of solving, the target value of each interval is set to y int ={1, 2, ..., L}, then y int Convert to one-hot encoding , the above multi-objective optimization problem can be transformed into a single-objective optimization problem as shown below:

[0111]

[0112] in,

[0113] In this embodiment, the generator f can be constructed x , used to generate samples . Among them, the generator f x It can be a two-layer fully connected neural network, and the activation function can be ReLU. For the generator f x Parameters; Based on the generator, the following optimization problem can be constructed to further adjust the distribution of the fourth latent variable :

[0114]

[0115] In this embodiment, the predictor f can also be constructed y , used to make predictions . Among them, the predictor f y It can be used to predict the fourth latent variable of the input. It can be a two-layer fully connected neural network, and the activation function of the predictor can be ReLU. The predictor f y Parameters. Furthermore, the following optimization problem can be constructed for the predictor to reduce the prediction error and realize a predictor with strong stability:

[0116]

[0117] Optionally, based on the above three optimization problems, the final objective equation can be derived:

[0118]

[0119] in,

[0120] In this embodiment, the model parameters of the predictor (i.e., the sub-prediction model) can be quickly determined by the above-mentioned target equation to obtain a trained prediction model, and the acquired enhanced latent variables are predicted using the trained prediction model to obtain accurate multiple target load data. Among them, the generator and the discriminator are mainly used to generate the target equation to complete the rapid confirmation of the model parameters of the sub-prediction model.

[0121] Optionally, this embodiment may set a maximum step length T. Fixing the parameters determined by the objective equation Furthermore, the gradient of the discriminator D can be obtained based on the gradient descent method (e.g., mini-batch) and the Adam optimization algorithm. , and update θ D . The gradient of the discriminator D can be determined by the following formula:

[0122]

[0123] in, , which can be used to represent the target value in each interval and can be in coded form.

[0124] Optionally, fix , based on the mini-batch gradient descent method, the gradients of the remaining models are obtained , to update Among them, the gradient As shown below:

[0125]

[0126] Repeat the above gradient calculation process T times, or , we can get the optimal parameter θ * . Among them, ∈ = 1e-6 is the setting error. θ * It can be the updated θ.

[0127] Optionally, the updated model parameters can be obtained through the above steps, and the initial model parameters of the sub-prediction model can be adjusted according to the model parameters to obtain the prediction model.

[0128] In this embodiment, through the transformation and enhancement of the designed latent variables, on the one hand, the input features can be effectively calibrated from the latent space, and on the other hand, latent variables of small sample categories can be added in the latent space to improve the adverse effects of the imbalanced distribution of training samples on the short-term load forecasting results. At the same time, the latent variable transformation and enhancement and prediction are integrated into the same model to achieve the goal of end-to-end prediction from samples to prediction results, avoid the expenditure of additional computing costs, and reduce training time, thereby achieving the technical effect of improving the accuracy of load prediction of transmission lines and solving the technical problem of low prediction accuracy of load of transmission lines.

[0129] In an embodiment of the present invention, when predicting the load of a transmission line at a future moment, multiple initial load data in a first historical period can be obtained, multiple first latent variables corresponding to the multiple initial load data can be determined, and the multiple first latent variables can be enhanced to obtain multiple enhanced latent variables. By predicting the multiple latent variables, multiple target loads in the future period can be obtained, thereby achieving the technical effect of improving the accuracy of load prediction of the transmission line and solving the technical problem of low prediction accuracy of the load of the transmission line.

[0130] Example 2

[0131] The technical solution of the embodiment of the present invention is illustrated below in conjunction with preferred implementation modes.

[0132] Since accurate short-term load forecasting for transmission lines can, on the one hand, ensure the safe and stable operation of the power system, thereby improving social benefits; on the other hand, it can accelerate the development of the power market and improve economic benefits. Therefore, it is of great significance to design an efficient and accurate short-term load forecasting model.

[0133] At present, deep learning models are usually used to improve prediction accuracy by mining deep features in massive load data. For example, the Long Short-Term Memory (LSTM) neural network extracts time-related features from load data to reduce prediction errors. Alternatively, the prediction model can be optimized through swarm intelligence optimization algorithms to improve the generalization and robustness of the model.

[0134] Furthermore, due to the influence of weather and regional electricity consumption, the feature distributions corresponding to different load values ​​are often different, resulting in an imbalanced distribution of training samples. However, current prediction models have failed to effectively consider the impact of the imbalanced distribution of training samples on the prediction results, resulting in the model performing worse than the prediction of the category with a small number of load value samples when predicting a category with a sufficient number of load value samples, resulting in a technical problem of low prediction accuracy of the load of the transmission line.

[0135] In order to solve the above problems, in this embodiment, a short-term load forecasting model considering the problem of unbalanced sample distribution is proposed, and the purpose of improving the accuracy of short-term load forecasting is achieved by utilizing the model.

[0136] Optionally, the above model can be a short-term load forecasting model for transmission lines based on latent variable transformation and enhancement. In this embodiment, a model training set based on the transmission line load is established; a priori latent variable extraction model is established; a mean calculation network and a variance calculation network are established to perform latent variable sampling; the latent variables are transformed and enhanced; a target equation is constructed; based on the mini-batch gradient descent method and the target equation, the established model is trained to obtain a trained prediction model, and the prediction model can be used to perform load prediction on new input features, thereby obtaining target load data with high accuracy, thereby achieving the technical effect of improving the load prediction accuracy of the transmission line and solving the technical problem of low prediction accuracy of the load of the transmission line.

[0137] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments; the size of the numbers in the embodiments is only for illustration and is not specifically limited here. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention and are not specifically limited here.

[0138] Figure 2 is a flow chart of constructing a short-term load forecasting model for a transmission line based on latent variable transformation and enhancement according to an embodiment of the present invention, such as Figure 2 As shown, the prediction model can be constructed by following the steps below:

[0139] Step S201: construct a model training set based on the historical load of the transmission line.

[0140] In this embodiment, the historical load data of the transmission line recorded in the second historical period can be collected. ,in, Can be used to represent the number of days recorded. Can be used to characterize daily load data. The resolution that can be used to characterize the load data can be, for example, 24, 48, or 96.

[0141] Optionally, an input feature data set and a prediction data set can be constructed based on multiple historical load data. The prediction data set can be a target value (y (i) =L (i+1) ),in Can be used to determine the forecast value for the next day.

[0142] Optionally, meteorological data for the second historical period may also be collected, and the meteorological data may include the maximum temperature of each day. , minimum temperature , average temperature , daily relative humidity , daily rainfall , etc. It should be noted that this is only an example and there is no specific restriction on the type and form of expression of meteorological data.

[0143] Optionally, the input dataset can be constructed using the meteorological data and the input feature dataset , the input data set can be the input feature.

[0144] For example, assuming that the resolution of the historical load data is 24, and the second historical period can be from July 7 to July 8, then the load value for 24 hours on July 7 is obtained to obtain the input feature data set, and the meteorological data of the transmission line in the second historical period, such as the highest temperature, the lowest temperature, the daily average temperature, the daily relative humidity, the daily rainfall and other data, can be obtained. The input data set can be constructed based on the input feature data set and the meteorological data. The load value for 24 hours on July 8 can be determined as the prediction data set. The input data set is used as the input data, and the prediction data set is used as the target value (that is, the output data). The above data is used to train the prediction model.

[0145] Step S202: training a priori latent variable extraction model.

[0146] In this embodiment, the prior latent variable extraction model three-layer long short-term memory neural network can be used to obtain the second latent variable, and a three-layer long short-term memory neural network can be pre-built. The number of neurons in the first layer of the neural network can be 16, the number of neurons in the second layer can be 32, and the number of neurons in the third layer can be 64.

[0147] Optionally, after the input data set is constructed, the input data set can be extracted through a long short-term memory network to obtain multiple second latent variables. For example, the LSTM neural network can obtain the second latent variable through the following formula:

[0148] z=LSTM(x;θ LSTM )

[0149] in, can be used to characterize the second latent variable, can be used to characterize the input data set, θ LSTM It can be used to characterize the parameters of LSTM neural network.

[0150] Optionally, the LSTM neural network can be replaced by a recurrent neural network, a gated recurrent neural network, a variant of a recurrent neural network, and the like. No specific limitation is imposed on the model for obtaining the second latent variable.

[0151] In this embodiment, a two-layer fully connected neural network can also be established, wherein the number of neurons in the first layer of the two-layer fully connected neural network can be 64, and the number of neurons in the second layer can be the resolution of the load data. Furthermore, the load forecast value can be obtained by using the above fully connected neural network

[0152]

[0153] Among them, θ FC It can be used to characterize the parameters of the fully connected neural network FC.

[0154] Furthermore, based on the load forecast value, the three-layer long short-term memory neural network can be iteratively updated using the stochastic gradient descent optimization algorithm (Adam) optimizer and the target value y, and finally the optimal LSTM neural network parameters can be obtained. , the prior latent variable extraction model obtained by training can be used to accurately obtain the second latent variable.

[0155] For example, using the Adam optimization algorithm to optimize the model can With the predicted value For comparison, the model parameters θ LSTM and φ FC to update.

[0156] Optionally, the prior latent variable extraction model can be used only for extracting the second latent variable during the model training process, and is not involved in the testing process. It should be noted that the size of the above numbers is only for example and is not specifically limited here.

[0157] Step S203, establishing a mean calculation network and a variance calculation network, performing latent variable sampling, and obtaining a third latent variable.

[0158] In this embodiment, the input feature x and the target value y in the input feature data set can be input into the mean calculation network to obtain the mean μ=C μ (x, y; φ1). φ1 can be used to characterize the parameters of the mean calculation network, which can be a pre-set value or a randomly generated value. The method for determining the above parameters is not specifically limited here. The mean calculation network can include but is not limited to LSTM neural networks, recurrent neural networks, gated recurrent neural networks, and variants of recurrent neural networks, etc. The type of the mean calculation network is not specifically limited here.

[0159] In this embodiment, the input feature x and the target value y can be input into the variance calculation network to obtain the variance σ=C σ (x, y; φ2). φ2 may be a parameter of a variance calculation network, and the variance calculation network may include but is not limited to a LSTM neural network, a recurrent neural network, a gated recurrent neural network, and a variant of a recurrent neural network. The type of the variance calculation network is not specifically limited here.

[0160] Furthermore, based on the mean μ and variance σ, we can use the Bayesian rule and importance sampling to obtain the third latent variable .

[0161] Step S204, transform and enhance the third latent variable.

[0162] In this embodiment, the parameters L, α, s, and σ can be preset. (i) , we can sum all the elements to get , and then obtain a new target value set

[0163] For example, the 24 load data on July 7, July 8, and July 9 can be summed to obtain the total load on July 7, July 8, and July 9, respectively, thereby obtaining the total load set for the first historical period (ie, the target value set).

[0164] In this embodiment, the total load set (y s ) and minimum value , calculate the target value set partition interval size , and divide the interval into L parts, that is, the size of each interval is

[0165] For example, we can sum the load data from July 7 to July 14 to get the total load set y for the first historical period. s ={25, 28, 29, 30, 32, 20, 29, 36}. Further, it can be determined that the maximum value of the total load set is 36, the minimum value is 20, and the interval size r=16. Then, assuming that L is set to 4, the size of each interval is 16 / 4=4. Then the multiple sub-load sets can be y 1 ={20},y 2 ={25, 28}, y 3 ={29, 29, 30, 32}, y 4 ={36}

[0166] In this embodiment, the upper and lower limits of each interval are determined, and the lower limit of the lth interval is Upper limit is ; Based on the upper and lower limits, the target values ​​distributed in each sub-load set are counted separately , where l∈{1, 2, ..., L}, and the corresponding third latent variable z′ l At the same time, the number of samples N in each sub-load set can be counted l .

[0167] In this embodiment, the third latent variable z can be calculated l The corresponding average value m l and the covariance c l , as shown below:

[0168]

[0169]

[0170] In this embodiment, during the model iteration process, the moving average and the moving covariance can be set to obtain the moving average set and the flow covariance , where t can be used to represent the current iteration number, and the initial flowing mean m0 and flowing covariance c0 can be randomly generated. The flowing mean and flowing covariance belonging to the lth interval during the iteration process can be calculated by the following formula:

[0171]

[0172]

[0173] in, It can be pre-generated randomly. L can be the lth interval. t can be the iteration here. α can be a random number in [0, 1].

[0174] In this embodiment, the Gaussian filter can be set as follows:

[0175]

[0176] where u and v are integers in [-z, ..., z],

[0177] In this embodiment, a Gaussian filter is used to filter the running mean m t+1 and flow covariance c t+1 Perform convolution operation to obtain the sliding average and sliding covariance , as shown below:

[0178]

[0179]

[0180] Optionally, the transformed mean value belonging to the lth interval can be obtained as , and the transformed covariance , and construct a Gaussian distribution . Determine the number of samples in each time interval {N1, N2, ..., N L The maximum value among them is N max , based on Bayes' rule and using importance sampling from a Gaussian distribution In the above example, N max Hidden variables Complete the latent variable enhancement belonging to the lth interval.

[0181] Optionally, repeat the above steps for L sampling times to obtain the transformation enhanced latent variable .

[0182] For example, the load set y s ={25, 28, 29, 30, 32, 20, 29, 36}, the maximum value is 36, the minimum value is 20, and the interval size r = 16. Assuming L is set to 4, that is, the size of each interval is 16 / 4 = 4, then the sub-load sets are , we can further determine the number of samples in each sub-load set {N1, N2, N3, N4} = {1, 2, 4, 1}, and then sample 4 latent variables from the Gaussian distribution corresponding to each interval to complete the latent variable enhancement.

[0183] Step S205, constructing the target equation.

[0184] In this embodiment, a discriminator D={D1, D2, ..., D L}, through the constructed discriminator, the second latent variable and the fourth latent variable Processing is performed to determine the second latent variable (z) and the fourth latent variable Are they from the same distribution? , the discriminator D can be a two-layer fully connected neural network, and the activation function can be ReLU. Furthermore, based on the discriminator, an optimization problem can be constructed to adjust the distribution of latent variables. The optimization problem for the latent variable belonging to the lth interval is defined as follows:

[0185]

[0186] Among them, φ={φ1, φ2}.

[0187] In this embodiment, the second latent variable is obtained by using the LSTM model, and the second latent variable and the fourth latent variable are input into the discriminator together. In the iterative process, the discriminator cannot determine which one is the real latent variable, thereby making the second latent variable and the fourth latent variable more similar.

[0188] Optionally, considering that the optimization of L discriminators will become a multi-objective optimization problem, which increases the difficulty and time of solving, therefore, in order to reduce the difficulty of solving, the target value of each interval is set to y int ={1, 2, ..., L}, then y int Convert to one-hot encoding , the above multi-objective optimization problem can be transformed into a single-objective optimization problem as shown below:

[0189]

[0190] in,

[0191] In this embodiment, the generator f can be constructed x , used to generate samples . Among them, the generator f x It can be a two-layer fully connected neural network, and the activation function can be ReLU. For the generator f x Parameters; Based on the generator, the following optimization problem can be constructed to further adjust the distribution of the fourth latent variable

[0192]

[0193] In this embodiment, the predictor f can also be constructed y , used to make predictions . Among them, the predictor f y It can be used to predict the fourth latent variable of the input. It can be a two-layer fully connected neural network, and the activation function of the predictor can be ReLU. The predictor f y Parameters. Furthermore, the following optimization problem can be constructed for the predictor to reduce the prediction error and realize a predictor with strong stability:

[0194]

[0195] Optionally, based on the above three optimization problems, the final objective equation can be derived:

[0196]

[0197] in,

[0198] In this embodiment, the model parameters of the predictor (i.e., the sub-prediction model) can be quickly determined by the above-mentioned target equation to obtain a trained prediction model, and the acquired enhanced latent variables are predicted using the trained prediction model to obtain accurate multiple target load data. Among them, the generator and the discriminator are mainly used to generate the target equation to complete the rapid confirmation of the model parameters of the sub-prediction model.

[0199] Figure 3 is a schematic diagram of an overall framework of a transmission line short-term load forecasting model based on latent variable transformation and enhancement according to an embodiment of the present invention, such as Figure 3As shown, an input data set 301 and a prediction data set 302 are obtained. The mean value calculation network 303 processes the load data in the input data set 301 to obtain a mean value (μ = C μ (x, y; φ1)), the variance calculation network 304 processes the load data in the input data set 301 to obtain the variance value (σ=C σ (x, y; φ2)), the variance and mean are sampled as latent variables, and the third latent variable of the input data set can be obtained , enhance the third latent variable and get the fourth latent variable The optimization problem is constructed using the generator 305, the predictor 306 and the discriminator 307 to obtain the target equation. The initial model parameters can be determined through the target equation. Further based on the mini-batch gradient descent method and the target equation, the final model parameters of the sub-prediction model can be obtained. The initial model parameters of the sub-prediction model can be adjusted according to the model parameters to obtain the prediction model.

[0200] Optionally, the predictor can be used to generate a target value, which can be compared with data in the prediction data set. The discriminator can be used to compare the similarity between the second latent variable and the fourth latent variable. The generator can be used to generate input features.

[0201] Step S206, based on the gradient descent method and the target equation, the established model is trained to obtain the final prediction model.

[0202] In this embodiment, a maximum step size T can be set. The parameters determined by the objective equation are fixed. Furthermore, the gradient of the discriminator D can be obtained based on the mini-batch gradient descent method and the Adam optimization algorithm. And update θ D . The gradient of the discriminator D can be determined by the following formula:

[0203]

[0204] Optionally, fix , based on the mini-batch gradient descent method, the gradients of the remaining models are obtained , to update Among them, the gradient As shown below:

[0205]

[0206] Repeat the above gradient calculation process T times, or , we can get the optimal parameter θ * . Among them, ∈ = 1e-6 is the setting error. θ* It can be the updated θ.

[0207] Step S207, testing the trained prediction model.

[0208] In this embodiment, the optimal parameter θ * Take out φ, ,in, is the predictor f y φ={φ1,φ2},φ1 can be the parameter of the mean calculation network, φ2 can be the parameter of the variance calculation network. * (that is, load characteristics) are input into the mean calculation network and the variance calculation network to obtain the first mean μ * And the first variance value σ * , construct a first Gaussian distribution of multiple first means and multiple first variance values, and obtain S first latent variables z′={z′ based on the Bayesian rule and importance sampling (1) , z′ (2) , ..., z′ (S)}. The first latent variable can be transformed to obtain the enhanced latent variable Based on the trained prediction model (i.e., predictor), the enhanced latent variables are processed to obtain the target load data (ie, predicted value).

[0209] In this embodiment, through the transformation and enhancement of the designed latent variables, on the one hand, the input features can be effectively calibrated from the latent space, and on the other hand, latent variables of small sample categories can be added in the latent space to improve the adverse effects of the imbalanced distribution of training samples on the short-term load forecasting results. At the same time, the latent variable transformation and enhancement and prediction are integrated into the same model to achieve the goal of end-to-end prediction from samples to prediction results, avoid the expenditure of additional computing costs, and reduce training time, thereby achieving the technical effect of improving the accuracy of load prediction of transmission lines and solving the technical problem of low prediction accuracy of load of transmission lines.

[0210] Example 3

[0211] According to an embodiment of the present invention, a load prediction device for a power transmission line is also provided. It should be noted that the load prediction device for a power transmission line of this embodiment can be used to execute the load prediction method for a power transmission line of embodiment 1 of the present invention.

[0212] Figure 4 Schematic diagram of a load prediction device for a power transmission line according to an embodiment of the present invention. Figure 4As shown, the load prediction device 40 for the transmission line may include: an acquisition unit 402 , a conversion unit 404 , a processing unit 406 and a prediction unit 408 .

[0213] The acquisition unit 402 is used to acquire a plurality of initial load data of the transmission line in a first historical period, wherein the first historical period includes a plurality of sub-periods.

[0214] The conversion unit 404 is used to convert at least one initial load data corresponding to each sub-period among the multiple initial load data to obtain a first latent variable corresponding to each sub-period.

[0215] The processing unit 406 is used to perform enhancement processing on the first latent variable to obtain an enhanced latent variable.

[0216] The prediction unit 408 is used to predict multiple enhanced latent variables corresponding to multiple sub-periods to obtain multiple target load data of the transmission line in the future period.

[0217] The load prediction device for the transmission line of this embodiment obtains multiple initial load data of the transmission line in a first historical period through an acquisition unit, wherein the first historical period includes multiple sub-periods; converts at least one initial load data corresponding to each sub-period among the multiple initial load data through a conversion unit to obtain a first latent variable corresponding to each sub-period; enhances the first latent variable through a processing unit to obtain an enhanced latent variable; and predicts multiple enhanced latent variables corresponding to multiple sub-periods through a prediction unit to obtain multiple target load data of the transmission line in future periods, thereby achieving the technical effect of improving the accuracy of load prediction of the transmission line and solving the technical problem of low accuracy of load prediction of the transmission line.

[0218] Example 4

[0219] The embodiment of the present invention can provide a computer terminal, which can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal can also be replaced by a terminal device such as a mobile terminal.

[0220] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

[0221] In this embodiment, the computer terminal can execute the steps in the performance detection method of the operating system.

[0222] Optionally, Figure 5 is a structural block diagram of a computer terminal according to an embodiment of the present invention. Figure 5As shown, the computer terminal 508 may include: one or more (only one is shown in the figure) processors 502 , a memory 504 and a transmission device 506 .

[0223] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the performance detection method and device of the operating system in the embodiment of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the performance detection method of the above-mentioned operating system. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 508 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0224] It can be understood by those skilled in the art that Figure 5 The structure shown is for illustration only, and the computer terminal 508 may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (MID for short), a PAD, or other terminal devices. Figure 5 The structure of the computer terminal 508 is not limited. For example, the computer terminal 508 may also include Figure 5 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 5 Different configurations are shown.

[0225] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0226] Example 5

[0227] According to an embodiment of the present invention, there is further provided a computer-readable storage medium, the computer-readable storage medium comprising a stored program, wherein the program executes the load prediction method for the power transmission line in Embodiment 1.

[0228] Optionally, in this embodiment, the computer-readable storage medium may be located in any one of the computer terminals in a computer terminal group in a computer network, or in any one of the mobile terminals in a mobile terminal group.

[0229] Optionally, in this embodiment, the computer-readable storage medium is configured to store steps for executing a method for load prediction of a power transmission line.

[0230] Example 6

[0231] According to an embodiment of the present invention, a processor is further provided, the processor being used to run a program, wherein when the program is run by the processor, the load prediction method for the power transmission line in Embodiment 1 is executed.

[0232] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of the computer network.

[0233] In this embodiment, the computer terminal can execute the program code of the steps in the method for load prediction of a power transmission line.

[0234] Wherein, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the multilingual translation method and device in the embodiment of the present invention, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned multilingual translation method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0235] By adopting the embodiment of the present invention, when predicting the load of a transmission line at a future moment, multiple initial load data in a first historical period can be obtained, multiple first latent variables corresponding to the multiple initial load data can be determined, and the multiple first latent variables can be enhanced to obtain multiple enhanced latent variables. By predicting the multiple latent variables, multiple target loads in the future time period can be obtained, thereby achieving the technical effect of improving the accuracy of load prediction of the transmission line and solving the technical problem of low prediction accuracy of the load of the transmission line.

[0236] Example 7

[0237] According to an embodiment of the present invention, a computer program product is further provided. The computer program product includes computer instructions, wherein the computer instructions implement the load prediction method for the transmission line in Embodiment 1 when executed by a processor.

[0238] Example 8

[0239] An embodiment of the present application may provide an electronic device, which may include a memory and a processor.

[0240] Figure 6 It is a block diagram of an electronic device according to a performance detection method of an operating system according to an embodiment of the present application. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0241] like Figure 6 As shown, the device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 to a random access memory (RAM) 603. In RAM 603, various programs and data required for the operation of the device 600 can also be stored. The computing unit 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0242] A number of components in the device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 604, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0243] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as a verification method for data. For example, in some embodiments, the verification method for data may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the verification method for data described above may be executed. Alternatively, in other embodiments, the computing unit 601 may be configured to execute the data verification method in any other appropriate manner (for example, by means of firmware).

[0244] According to an embodiment of the present application, a performance detection method for an operating system is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0245] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that may be executed and / or interpreted on a programmable system including at least one programmable processor that may be a dedicated or general purpose programmable processor that may receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0246] The program code for implementing the method of the present application can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0247] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0248] To provide interaction with a user, the systems and techniques described herein may be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or an LCD (liquid crystal display, monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a path ball) through which the user can provide input to the computer. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0249] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a Local Area Network (LAN), a Wide Area Network (WAN), and the Internet.

[0250] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0251] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0252] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0253] In the several embodiments provided by the present invention, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic, for example, the division of units can be a logical function division, and there can be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0254] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0255] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0256] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0257] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for load forecasting of a transmission line, characterized in that: include: Acquire a plurality of initial load data of the transmission line in a first historical period, wherein the first historical period includes a plurality of sub-periods; Convert at least one of the plurality of initial load data corresponding to each of the sub-periods to obtain a first latent variable corresponding to each of the sub-periods; Performing enhancement processing on the first latent variable to obtain an enhanced latent variable; Predicting the multiple enhanced latent variables corresponding to the multiple sub-periods, to obtain multiple target load data of the transmission line in future time periods.

2. The method according to claim 1, characterized in that: Converting at least one of the initial load data corresponding to each of the sub-periods among the multiple initial load data to obtain the first latent variable corresponding to each of the sub-periods includes: Determine a first mean value and a first variance value corresponding to at least one of the initial load data, and obtain the first mean value and the first variance value corresponding to each of the sub-periods; constructing a first Gaussian distribution of a plurality of the first means and a plurality of the first variance values; The first Gaussian distribution is sampled to obtain a first latent variable corresponding to each sub-period in the first historical period, and a plurality of first latent variables are obtained.

3. The method according to claim 1, characterized in that Performing enhancement processing on the first latent variable to obtain an enhanced latent variable includes: Determining the total load of the sub-period based on at least one initial load data corresponding to the sub-period; Based on the total loads of the multiple sub-periods, obtaining a total load set of the first historical period; Dividing the total load set to obtain multiple sub-load sets; respectively determining target loads in the plurality of sub-load sets to obtain a plurality of target loads; Determine target latent variables corresponding to the plurality of target loads in the first latent variables to obtain the plurality of target latent variables; The multiple target latent variables are transformed to obtain multiple enhanced latent variables.

4. The method according to claim 3, characterized in that The total load set is divided to obtain a plurality of sub-load sets, including: Determine the maximum and minimum values ​​in the total load set; Based on the maximum value and the minimum value, the total load set is divided to obtain multiple sub-load sets.

5. The method according to claim 3, characterized in that: Determining the target loads in the plurality of sub-load sets respectively to obtain the plurality of target loads comprises: determining a first value and a second value in the subload set, wherein the first value is greater than the second value; The target load in the sub-load set is screened out by using the first value and the second value.

6. The method according to claim 3, characterized in that The plurality of target latent variables are transformed to obtain a plurality of enhanced latent variables, including: Determining a second mean corresponding to the target latent variable and a covariance corresponding to the target latent variable based on the number of initial load data in the sub-load set; Get the moving mean set and the moving covariance set; The moving mean and the moving covariance are convolved using a Gaussian filter to obtain a sliding mean and a sliding covariance; Determine the difference between the sliding average value and the second mean value as a transformed average value; Determine the difference between the sliding covariance and the covariance as the transformed covariance; constructing a second Gaussian distribution of the transformed mean and the transformed covariance; The second Gaussian distribution is sampled to obtain the enhanced latent variable.

7. The method according to claim 1, characterized in that Predicting a plurality of enhanced latent variables corresponding to a plurality of the sub-periods to obtain a plurality of the target load data of the transmission line in the future period includes: The target load data is obtained by predicting the multiple enhanced latent variables using a prediction model, wherein the prediction model is trained based on multiple historical load data of the transmission line in a second historical period, and the second historical period is earlier than the first historical period.

8. The method according to claim 7, characterized in that The method further comprises: Acquire a plurality of historical load data of the transmission line; Based on the plurality of historical load data, construct an input feature data set and a prediction data set; Acquiring meteorological data of the power transmission line in the second historical period; Constructing the input data set by using the meteorological data and the input feature data set; The prediction model is obtained by training using the input data set and the prediction data set.

9. The method according to claim 8, characterized in that The prediction model is obtained by training the input data set and the prediction data set, comprising: The input data set is extracted using a long short-term memory network to obtain multiple second latent variables; Determine a third latent variable of the input data set by using a mean calculation network and a variance calculation network; Performing enhancement processing on the third latent variable to obtain a fourth latent variable; Based on the second latent variable, the fourth latent variable and the prediction data set, construct a target equation; Determining model parameters of the prediction model using the objective equation; According to the model parameters, the initial model parameters of the sub-prediction model are adjusted to obtain the prediction model.

10. A load prediction device for a transmission line, characterized in that: include: An acquisition unit, configured to acquire a plurality of initial load data of a transmission line in a first historical period, wherein the first historical period includes a plurality of sub-periods; A conversion unit, configured to convert at least one of the initial load data corresponding to each of the sub-periods among the plurality of initial load data to obtain a first latent variable corresponding to each of the sub-periods; A processing unit, configured to perform enhancement processing on the first latent variable to obtain an enhanced latent variable; The prediction unit is used to predict the multiple enhanced latent variables corresponding to the multiple sub-periods to obtain multiple target load data of the transmission line in the future period.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 9.

12. A processor, characterized in that: The processor is used to run a program, wherein the program executes the method according to any one of claims 1 to 9 when being run by the processor.

13. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the method according to any one of claims 1 to 9.