Power system load data super-resolution reconstruction method and system

Abstract

The invention discloses a power system load data super-resolution reconstruction method and system, and relates to the technical field of power system data quality improvement, and the method comprises the steps: carrying out the weighting of weather influence factors under different weather conditions, calculating the correlation between load data and weather influence factor data, obtaining key influence factor data, and carrying out the calculation of the correlation between the load data and the weather influence factor data; performing dynamic range normalization processing by adopting a sliding window; carrying out super-resolution reconstruction on the normalized data by adopting a trained time sequence generation model to obtain high-resolution load data; the time sequence generation model adjusts the weight of a loss function according to the change of the current training loss during training, and after training is completed, super-resolution reconstruction is carried out by extracting combined features of normalized load data and key influence factor data. The defect of insufficient resolution of low-quality sampling data is overcome, the quality of reconstructed data is improved through meteorological factor constraints, and the problem that training or using requirements cannot be met due to the fact that the resolution of the sampling data is low is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of improving power system data quality, and particularly to a method and system for super-resolution reconstruction of power system load data. Background Art

[0002] The statements in this section merely provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] With the continuous change of the global energy structure, the power system is undergoing unprecedented changes. On the one hand, the unexpected installation of renewable energy brings strong randomness and intermittency to the operation of the power system, which requires the system to have high-frequency data to assist power flow scheduling and operation optimization; on the other hand, the large-scale deployment of new loads such as electric vehicles and data centers requires high-quality power supply and real-time coordination ability with the power system, which requires power system operators to have high-quality and real-time system operation data to meet the additional needs of users.

[0004] However, the existing sampled data has the following problems: on the one hand, due to reasons such as power system sensor anomalies and transmission line failures, there are a large number of missing data in the power data; on the other hand, due to factors such as sampling equipment and transmission bandwidth, the data sampling frequency of some regional power grids is relatively low. For example, in existing monitoring and data acquisition systems, most meters in smart grids are traditional sensors deployed many years ago, and traditional meters collect measurement values at a relatively low frequency, resulting in a small amount of effective data.

[0005] Currently, the solutions proposed by experts and scholars for the problems of limited effective data and low sampling frequency can be roughly divided into two categories: one is to recover missing data, but this does not fundamentally solve the problem of low sampling frequency; the other is data enhancement based on super-resolution reconstruction, but most of this is based on generative adversarial networks, and traditional generative adversarial network models have certain limitations in multi-dimensional data processing and cannot efficiently process the current complex power system time series data.

[0006] Power system data are all time series data. Affected by the integration of a high proportion of new energy, the current difference between high-frequency data and low-frequency data is mainly affected by weather conditions such as temperature, relative humidity, and wind speed. However, existing methods mainly start from the single time series characteristics of power system operation data and perform super-resolution reconstruction according to the time series characteristics, without considering the influence of external characteristics such as weather factors on data characteristics; moreover, traditional methods do not capture the long-distance dependence relationships contained in the time series data sampled for a long time, resulting in low-quality generated data. Summary of the Invention

[0007] To solve the above problems, the present invention proposes a method and system for super-resolution reconstruction of power system load data. Considering the influence of weather factors, analyzing the data of key weather influencing factors most relevant to load changes, performing super-resolution reconstruction of the data based on an improved time series generation model, and adjusting the model parameters according to the training process to achieve super-resolution reconstruction of low-resolution load data in low-voltage power distribution areas.

[0008] To achieve the above object, the present invention adopts the following technical solutions:

[0009] In a first aspect, the present invention provides a method for super-resolution reconstruction of power system load data, including:

[0010] Obtain load data and weather influencing factor data. After weighting each weather influencing factor under different weather conditions, calculate the correlation between the load data and the weather influencing factor data, thereby obtaining key influencing factor data;

[0011] Use a sliding window to perform normalization processing on the dynamic range of the load data and the key influencing factor data;

[0012] For the normalized load data and key influencing factor data, perform super-resolution reconstruction using a trained time series generation model to obtain high-resolution load data, and after denormalizing the high-resolution load data, obtain high-resolution load data that conforms to the actual dimension;

[0013] When training, the time series generation model adjusts the weight of the loss function according to the change of the current training loss. After training, super-resolution reconstruction is performed by extracting the combined features of the normalized load data and the key influencing factor data.

[0014] As an alternative implementation, the correlation between the load data and the weather influencing factor data is:

[0015]

[0016] In the formula, r is the correlation coefficient, w k is the weight of the kth weather influencing factor, x i and y i are the i-th data value of the load data sample and the i-th data value of the weather influencing factor data sample under the kth weather influencing factor, n is the total number of data in the sample, and are the average values.

[0017] As an alternative implementation, the weather influencing factors include temperature, light, relative humidity, and wind speed. The process of weighting each weather influencing factor under different weather conditions includes:

[0018] The temperature weight w1 is: w1 = α TE k T k S k E,TE ;

[0019] The light weight w2 is: w2 = α L k T k W ;

[0020] The relative humidity weight w3 is: w3 = α H k T k S ;

[0021] The wind speed weight w4 is: w4 = α W k WS k E,WS ;

[0022] Among them, α TE is the temperature reference weight; k E，TE is the extreme temperature correction coefficient; α L is the light reference weight; k W is the sunny and cloudy coefficient; α H is the humidity reference weight; α W is the wind speed reference weight; k WS is the wind speed strength coefficient; k E，WS is the extreme weather correction coefficient; k T is the time coefficient, k S is the season coefficient.

[0023] As an alternative implementation, the time series generation model includes a generator and a discriminator; the generator uses a double-layer long short-term memory network to extract the combined features of the normalized load data and the key influencing factor data, and adds a self-attention mechanism to each long short-term memory network layer, and then uses a transposed convolutional layer to expand the normalized load data into high-resolution load data and transmit the generated high-resolution load data to the discriminator.

[0024] As an alternative implementation, the total loss function of the time series generation model is composed of a generator loss function, a discriminator loss function, and a time series consistency loss function through weighting, and the weight of each loss function is adjusted according to the change of the current training loss.

[0025] As an alternative implementation, the weight x1 of the generator loss function is:

[0026] Among them, x 10is the generator loss benchmark weight; Δgen_loss is the difference between the previous round of generator loss and the current loss; λ is the adjustment coefficient; gen_loss(t - 1) is the previous round of generator loss;

[0027] The discriminator loss function weight x2 is:

[0028] where x 20 is the discriminator loss benchmark weight; Δdisc_loss is the difference between the previous round of discriminator loss and the current loss; disc_loss(t - 1) is the previous round of discriminator loss;

[0029] The temporal consistency loss function weight x3 is:

[0030] where x 30 is the temporal consistency loss benchmark weight; Δts_loss is the difference between the previous round of temporal consistency loss and the current loss; ts_loss(t - 1) is the previous round of temporal consistency loss.

[0031] In a second aspect, the present invention provides a power system load data super - resolution reconstruction system, including:

[0032] A correlation analysis module, configured to obtain load data and weather influence factor data, calculate the correlation between the load data and the weather influence factor data after weighting each weather influence factor under different weather conditions, thereby obtaining key influence factor data;

[0033] A normalization processing module, configured to perform dynamic range normalization processing on the load data and the key influence factor data by using a sliding window;

[0034] A reconstruction module, configured to perform super - resolution reconstruction on the normalized load data and key influence factor data by using a trained time - series generation model to obtain high - resolution load data, and perform denormalization on the high - resolution load data to obtain high - resolution load data that conforms to the actual dimension;

[0035] When training, the time - series generation model adjusts the loss function weight according to the change of the current training loss, and thus performs super - resolution reconstruction by extracting the combined features of the normalized load data and key influence factor data after training.

[0036] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the method described in the first aspect is completed.

[0037] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the method described in the first aspect.

[0038] Fifthly, the present invention provides a computer program product comprising a computer program which, when executed by a processor, implements the method described in the first aspect.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] The present invention proposes a super-resolution reconstruction method for low-voltage substation area power load data considering the influence of weather factors. The method sequentially performs outlier processing on the original data to obtain preprocessed data, and conducts correlation analysis on the preprocessed data to obtain weather factor data most relevant to load changes. Based on an improved time series generation model, data super-resolution reconstruction is carried out, and the model parameters are adjusted according to the training process, realizing the super-resolution reconstruction of low-resolution load data in low-voltage substation areas.

[0041] The present invention uses an improved Pearson correlation analysis method. Compared with traditional correlation analysis methods, weights related to weather influence factors are innovatively added, different weights are assigned to each weather influence factor under different weather conditions, and then the correlation between load data and weather influence factor data is analyzed separately to reflect the correlation between load data and weather influence factor data under different weather conditions.

[0042] The present invention uses an improved normalization method. Compared with traditional correlation analysis methods, considering that the variation range of weather factors is relatively large, the present invention innovatively proposes a dynamic normalization method based on a sliding window to dynamically adjust the normalization range, better adapt to the dynamic changes of time series data, eliminate the dimensional difference between data features, and avoid the problem that the normalized data becomes invalid due to a fixed global normalization range.

[0043] The present invention uses an improved time series generation model for super-resolution reconstruction, improves the model architecture so that the generated data can consider the influence of weather factors, and at the same time introduces a loss function considering weights. Compared with traditional methods, the improved method can automatically update the weights according to the change trend of the training loss each time, enabling the generated data to better fit the real data.

[0044] The present invention provides a method and system for super-resolution reconstruction of power system load data. Specifically, it performs super-resolution reconstruction on the low-resolution load data of low-voltage areas in the power system. On the basis of considering the weather impact brought by the high proportion of new energy access, it overcomes the defect of insufficient resolution of low-quality sampling data, improves the quality of the reconstructed data through meteorological factor constraints, and at the same time solves the problem that the low resolution of sampling data cannot meet the training or usage requirements. It can be applied to the field of improving the quality of distribution network data and enhance the development and utilization value of the massive data resources in the power system.

[0045] Advantages of additional aspects of the present invention will be partly given in the following description, partly will become obvious from the following description, or will be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0047] Figure 1 It is a flowchart of the method for super-resolution reconstruction of power system load data provided in Embodiment 1 of the present invention;

[0048] Figure 2 It is a schematic diagram of the working process of the time series generation model provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The following will further illustrate the present invention in conjunction with the drawings and embodiments.

[0050] It should be noted that the following detailed description is exemplary and is intended to provide further illustration of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0051] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "comprise" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0052] Without conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0053] Embodiment 1

[0054] As Figure 1 shown, this embodiment provides a super-resolution reconstruction method for power system load data considering the influence of weather factors, including:

[0055] Obtain load data and weather influence factor data. After weighting each weather influence factor under different weather conditions, calculate the correlation between the load data and the weather influence factor data, and thus obtain the key influence factor data;

[0056] Use a sliding window to perform normalization processing on the dynamic range of the load data and the key influence factor data;

[0057] For the normalized load data and key influence factor data, use the trained time series generation model for super-resolution reconstruction to obtain high-resolution load data, and after denormalizing the high-resolution load data, obtain high-resolution load data that conforms to the actual dimension;

[0058] When training, the time series generation model adjusts the weight of the loss function according to the change of the current training loss, and thus after training, super-resolution reconstruction is performed by extracting the combined features of the normalized load data and the key influence factor data.

[0059] The method of this embodiment will be elaborated in detail below.

[0060] S1: Obtain load data and weather influence factor data; First, use a random forest model to preprocess the outliers of the load data and the weather influence factor data to identify and replace the outliers in the low-resolution load data and the weather influence factor data.

[0061] Among them, the weather influence factors include temperature, light, relative humidity, and wind speed; the load data uses the low-resolution load data of the low-voltage power supply area (the low-voltage power supply area refers to the area powered by the low-voltage (380V) of the transformer).

[0062] Then, use the improved Pearson correlation analysis method to analyze the correlation between the load data and the weather influence factor data, determine the key influence factor data, and obtain the preprocessed load data and the key influence factor data.

[0063] Specifically:

[0064] During the correlation analysis process, the improved Pearson correlation analysis method is used. The improved Pearson correlation analysis method is: innovatively adding weights related to weather influence factors, and assigning different weights w to each weather influence factor under different weather conditions. k Then, analyze the correlation between the load data and the weather influence factor data respectively, reflecting the correlation between the load data and the weather influence factor data under different weather conditions.

[0065] The correlation coefficient is:

[0066]

[0067] In the formula, r is the correlation coefficient, w k is the weight of the k-th weather influence factor, x i and y i are the i-th data values of the load data sample and the i-th data values of the weather influence factor data sample under the k-th weather influence factor, n is the total number of data in the sample, and are the average values.

[0068] The setting of w k is as follows:

[0069] (1) Temperature weight: The influence of temperature on load data is usually large, especially in extreme weather (such as high temperature in summer or cold in winter), and its influence on loads such as air conditioning and heating is particularly significant.

[0070] w1 = α TE k T k S k E,TE ;

[0071] Among them: w1 is the temperature weight; α TE is the temperature reference weight; k T is the time coefficient, which is divided into night and day, higher during the day and lower at night, and its value is automatically generated by the model. The model divides the data interval into day and night by identifying the season corresponding to the input temperature data. For example, when the model identifies that the temperature data is sampled in spring, the day is set as 6:00 - 18:00; k S is the season coefficient, which is divided into spring and autumn seasons and winter and summer seasons, higher in winter and summer and lower in spring and autumn, and its value is automatically generated by the model. The model generates the corresponding value by identifying the season corresponding to the input temperature data; k E，TE is the extreme temperature correction coefficient, which is divided into high temperature and low temperature, higher in high temperature and lower in low temperature, and its value is automatically generated by the model. The model judges low temperature and high temperature by identifying whether the input temperature data exceeds the set threshold, and thus generates the corresponding value.

[0072] (2) Illumination weight: Illumination mainly affects the load during the day, especially the photovoltaic power generation load. When the temperature is relatively high during the day, illumination has a greater impact on the load.

[0073] w2 = α L k T k W ;

[0074] where: w2 is the illumination weight; α L is the reference illumination weight; k T is the time coefficient, which is divided into night and day, and its value is automatically generated by the model. The model divides the data interval into day and night by identifying the season corresponding to the input illumination data. For example, when the model identifies that the illumination data is sampled in spring, the day is set as 6:00 - 18:00, 1 during the day, and 0 at night; k W is the sunny and cloudy coefficient, which is divided into sunny and cloudy days. It is higher on sunny days and lower on cloudy days, and its value is automatically generated by the model. The model divides the data into sunny and cloudy days by identifying the input illumination data. When the illumination is strong, it is a sunny day, otherwise it is a cloudy day, and corresponding values are generated accordingly.

[0075] (3) Relative humidity weight: Relative humidity has a certain impact on the load in summer and winter, especially in the performance of air-conditioning load (the air-conditioning load is large when the humidity is high), but the change range of humidity is relatively small, and usually has a relatively small direct impact on the load.

[0076] w3 = α H k T k S ;

[0077] where: w3 is the relative humidity weight; α H is the reference humidity weight; k T is the time coefficient, which is divided into night and day, higher during the day and lower at night, and its value is automatically generated by the model. The model divides the data interval into day and night by identifying the season corresponding to the input humidity data. For example, when the model identifies that the humidity data is sampled in spring, the day is set as 6:00 - 18:00; k S is the season coefficient, which is divided into summer, winter, and spring-autumn seasons, higher in summer, second in winter, and lower in spring and autumn, and its value is automatically generated by the model by identifying the season corresponding to the input humidity data, and corresponding values are generated accordingly.

[0078] (4) Wind speed weight:

[0079] w4 = α W k WS k E,WS ;

[0080] Where: w4 is the wind speed weight; α W is the wind speed reference weight; k WS is the wind speed strength coefficient, which is divided into strong wind, normal wind speed, and no wind / light wind. The strong wind value is relatively high, the normal wind speed is 1, and the no wind / light wind value is relatively low. Its value is automatically generated by the model. The model determines which interval among strong wind, normal wind speed, and no wind / light wind the input wind speed data belongs to, and judges the wind speed accordingly, thereby generating the corresponding value; k E，WS is the extreme weather correction coefficient, which is divided into strong wind and normal wind. The value is relatively high during strong wind and 1 during normal wind. Its value is automatically generated by the model. The model determines which interval among strong wind and normal wind the input wind speed data belongs to, and judges the wind speed accordingly, thereby generating the corresponding value.

[0081] S2: For time series data, the range of the data may change over time. At the same time, considering the large variation range of weather factors, in order to avoid the deviation and data invalidation caused by fixed-range normalization, this embodiment proposes an improved normalization method to normalize the load data and key influencing factor data. By using a sliding window of a fixed size, the minimum and maximum values within the current time window are calculated, and the low-resolution load data and key influencing factor data are normalized with a dynamic range;

[0082] The size of the sliding window can be manually adjusted according to the number of data points. Traverse the input load data and key influencing factor data, and the window slides forward accordingly, thereby dynamically updating the normalization parameters; by dynamically adjusting the normalization range, it better adapts to the dynamic changes of time series data, eliminates the dimensional differences between data features, and avoids the invalidation of the normalized data due to the fixed global normalization range.

[0083] S3: Combine the normalized low-resolution load data and key influencing factor data at the dimension level. First, input the normalized load data and key influencing factor data into the generator. The generator learns the combined time series features of the normalized load data and key influencing factor data to obtain high-resolution load data.

[0084] Specifically: In the generator, comprehensively consider the influence of weather factors and the time series characteristics of the load, improve the generator architecture, add a self-attention mechanism layer, and use the multi-dimensionally combined data to learn the multi-dimensional fusion features of time series data; that is, use a two-layer long short-term memory network (LSTM, Long Short-Term Memory) to extract the combined features of the low-resolution load data and key influencing factor data, and add a self-attention mechanism to each LSTM layer. Then, use a deconvolution layer to expand the low-resolution load data into high-resolution load data, and then transmit the generated high-resolution load data to the discriminator.

[0085] S4: Secondly, in the discriminator, it receives the high-resolution load data generated by the generator and the real high-resolution load data, learns to distinguish real samples and generated samples, and provides feedback parameters for the generated samples to the generator to update the generator parameters. By continuously updating the parameters until a well-trained super-resolution reconstruction model based on the time series generation model (TimeGAN) is obtained.

[0086] Figure 2 It is a schematic diagram of the working processes of the discriminator and the generator in the TimeGAN model. Specifically, the generator learns the features of the low-resolution data, thereby generating high-resolution load data. The discriminator distinguishes real samples and generated samples, updates the generator parameters, and the loss function reaches the minimum value. Analysis Figure 2 It can be seen that the generator and the discriminator continuously modify their respective parameters through interactive training until the loss function reaches the minimum value.

[0087] In the discriminator, the discriminator architecture is improved. The double-layer LSTM is used to process the time series data, and the self-attention mechanism layer is added to learn the long-distance dependence relationship of the time series data. The sigmoid activation function is used to implement the output of the binary judgment result. Through providing the feedback parameters for the generated samples to the generator, the interactive training between the generator and the discriminator is carried out.

[0088] In the TimeGAN model, the parameters transmitted interactively between the generator and the discriminator are called loss functions. In this embodiment, the total loss function is composed of the generator loss function, the discriminator loss function, and the time series consistency loss function ts_loss. Among them, the generator loss function gen_loss uses the mean square error loss; the discriminator loss function disc_loss uses the binary cross-entropy loss;

[0089] The time series consistency loss function ts_loss is as follows:

[0090]

[0091] In the formula, T is the time series length, y_true t and y_pred t are the real data and the generated data at time t, y_true t+1 and y_pred t+1 are the real data and the generated data at time t + 1.

[0092] According to the model training bias and the characteristics of the generated data, three loss function weights are set as x1, x2, and x3 respectively, and the weights are automatically updated according to the change trend of the loss in each training to avoid a certain loss term from being saturated or ignored prematurely.

[0093] The total loss function total_loss is as follows:

[0094] total_loss = x1 × gen_loss + x2 × disc_loss + x3 × ts_loss.

[0095] The weight x1 of the generator loss function is:

[0096]

[0097] where x 10 is the baseline weight of the generator loss; Δgen_loss is the difference between the previous generator loss and the current loss; λ is the adjustment coefficient that controls the speed of weight change; gen_loss(t - 1) is the previous generator loss.

[0098] The weight x2 of the discriminator loss function is:

[0099]

[0100] where x 20 is the baseline weight of the discriminator loss; Δdisc_loss is the difference between the previous discriminator loss and the current loss; disc_loss(t - 1) is the previous discriminator loss.

[0101] The weight x3 of the temporal consistency loss function is:

[0102]

[0103] where x 30 is the baseline weight of the temporal consistency loss; Δts_loss is the difference between the previous temporal consistency loss and the current loss; ts_loss(t - 1) is the previous temporal consistency loss.

[0104] S5: Generate the generator and discriminator through interactive training, continuously update the loss function, and finally obtain the trained TimeGAN model; use the trained model to perform super-resolution reconstruction on the actual low-resolution low-voltage substation area load data, obtain the high-resolution low-voltage substation area load data, and perform anti-normalization on the generated data to obtain high-resolution data that conforms to the actual dimension.

[0105] This embodiment proposes a method for super-resolution reconstruction of low-resolution load data in the low-voltage substation area of the power system. On the basis of considering the weather impact brought by the high proportion of new energy access, it overcomes the defect of insufficient resolution of low-quality sampled data, improves the quality of the reconstructed data through meteorological factor constraints, and at the same time solves the problem that the resolution of the sampled data is too low to meet the training or use requirements. It can be applied to the field of improving the data quality of the distribution network and improve the development and utilization value of the massive data resources of the power system.

[0106] Example 2

[0107] This embodiment provides a power system load data super-resolution reconstruction system, including:

[0108] A correlation analysis module, configured to obtain load data and weather influence factor data, calculate the correlation between the load data and the weather influence factor data after weighting each weather influence factor under different weather conditions, thereby obtaining key influence factor data;

[0109] A normalization processing module, configured to perform dynamic range normalization processing on the load data and the key influence factor data by using a sliding window;

[0110] A reconstruction module, configured to perform super-resolution reconstruction on the normalized load data and key influence factor data by using a trained time series generation model to obtain high-resolution load data, and perform denormalization on the high-resolution load data to obtain high-resolution load data that conforms to the actual dimension;

[0111] When the time series generation model is trained, it adjusts the loss function weight according to the change of the current training loss, and thus performs super-resolution reconstruction by extracting the combined features of the normalized load data and key influence factor data after training.

[0112] It should be noted here that the above modules correspond to the steps described in Embodiment 1. The examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in the above Embodiment 1. It should be noted that the above modules can be executed in a computer system such as a set of computer executable instructions as part of the system.

[0113] In more embodiments, there is also provided:

[0114] An electronic device, including a memory and a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the method described in Embodiment 1 is completed. For the sake of brevity, it will not be repeated here.

[0115] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, field-programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0116] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0117] A computer-readable storage medium for storing computer instructions, which when executed by a processor, implement the method described in Embodiment 1.

[0118] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0119] A computer program product including a computer program, which when executed by a processor, implements the method described in Embodiment 1.

[0120] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which are executed in a device on a target real or virtual processor to perform the process / method as described above. Generally, program modules include routines, programs, libraries, objects, classes, components, data structures, etc. that perform specific tasks or implement specific abstract data types. In various embodiments, the functions of program modules can be combined or divided as needed. The machine-executable instructions for program modules can be executed locally or within a distributed device. In a distributed device, program modules can be located in local and remote storage media.

[0121] The computer program code for implementing the method of the present invention can be written in one or more programming languages. These computer program codes can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, such that when the program code is executed by the computer or other programmable data processing devices, the functions / operations specified in the flowchart and / or block diagram are implemented. The program code can be executed entirely on the computer, partially on the computer, as an independent software package, partially on the computer and partially on a remote computer, or entirely on a remote computer or server.

[0122] In the context of the present invention, the computer program code or related data can be carried by any suitable carrier so that the device, apparatus or processor can perform the various processes and operations described above. Examples of the carrier include signals, computer-readable media, and the like. Examples of signals can include electrical, optical, radio, acoustic or other forms of propagated signals, such as carrier waves, infrared signals, etc.

[0123] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in conjunction with this embodiment can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0124] Although the specific implementation manners of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that, based on the technical solution of the present invention, various modifications or deformations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present invention.

Claims

1. A method for super-resolution reconstruction of power system load data, characterized in that, Including: Obtain load data and weather influence factor data. After weighting each weather influence factor under different weather conditions, calculate the correlation between the load data and the weather influence factor data, thereby obtaining key influence factor data; Use a sliding window to perform dynamic range normalization processing on the load data and the key influence factor data; For the normalized load data and key influence factor data, use the trained time series generation model for super-resolution reconstruction to obtain high-resolution load data, and after denormalizing the high-resolution load data, obtain high-resolution load data that conforms to the actual dimension; When training, the time series generation model adjusts the loss function weight according to the change of the current training loss. After training is completed, super-resolution reconstruction is performed by extracting the combined features of the normalized load data and key influence factor data.

2. The method for super-resolution reconstruction of power system load data according to claim 1, wherein The correlation between the load data and the weather influence factor data is: where r is the correlation coefficient, w k is the weight of the k-th weather influence factor, x i and y i are the i-th data value of the load data sample and the i-th data value of the weather influence factor data sample under the k-th weather influence factor, n is the total number of data in the sample, and are the averages.

3. The method for super-resolution reconstruction of power system load data according to claim 2, characterized in that, The weather influence factors include temperature, light, relative humidity, and wind speed. The process of weighting each weather influence factor under different weather conditions includes: The temperature weight w1 is: w1 = α TE k T k S k E,TE ; The illumination weight w2 is: w2 = α L k T k W ; The relative humidity weight w3 is: w3 = α H k T k S ; The wind speed weight w4 is: w4 = α W k WS k E,WS ; Among them, α TE is the temperature reference weight; k E，TE is the extreme temperature correction coefficient; α L is the light reference weight; k W is the sunny and cloudy coefficient; α H is the humidity reference weight; α W is the wind speed reference weight; k WS is the wind speed strength coefficient; k E，WS is the extreme weather correction coefficient; k T is the time coefficient, k S is the season coefficient.

4. A super-resolution reconstruction method for power system load data according to claim 1, characterized in that The time series generation model includes a generator and a discriminator; the generator uses a double-layer long short-term memory network to extract the combined features of the normalized load data and key influence factor data, and adds a self-attention mechanism to each long short-term memory network layer. Then, an anti-convolution layer is used to expand the normalized load data into high-resolution load data and transmit the generated high-resolution load data to the discriminator.

5. A super-resolution reconstruction method for power system load data according to claim 1, characterized in that The total loss function of the time series generation model is composed of a generator loss function, a discriminator loss function, and a temporal consistency loss function through weighting, and the weight of each loss function is adjusted according to the change of the current training loss.

6. A method for super-resolution reconstruction of power system load data as described in claim 5, characterized in that, The weight x1 of the generator loss function is: where x 10 is the generator loss benchmark weight; Δgen_loss is the difference between the previous generator loss and the current loss; λ is the adjustment coefficient; gen_loss(t - 1) is the previous generator loss; The discriminator loss function weight x2 is as follows: where x 20 is the discriminator loss benchmark weight; Δdisc_loss is the difference between the discriminator loss in the previous round and the current loss; disc_loss(t-1) is the discriminator loss in the previous round; The weight x3 of the temporal consistency loss function is as follows: where x 30 is the weight of the temporal consistency loss benchmark; Δts_loss is the difference between the previous temporal consistency loss and the current loss; ts_loss(t - 1) is the previous temporal consistency loss.

7. A power system load data super-resolution reconstruction system, characterized in that Including: A correlation analysis module configured to obtain load data and weather influence factor data. After weighting each weather influence factor under different weather conditions, calculate the correlation between the load data and the weather influence factor data, thereby obtaining key influence factor data; A normalization processing module configured to use a sliding window to perform dynamic range normalization processing on the load data and the key influence factor data; A reconstruction module configured to perform super-resolution reconstruction on the normalized load data and key influence factor data using the trained time series generation model to obtain high-resolution load data, and after denormalizing the high-resolution load data, obtain high-resolution load data that conforms to the actual dimension; When training, the time series generation model adjusts the loss function weight according to the change of the current training loss. After training is completed, super-resolution reconstruction is performed by extracting the combined features of the normalized load data and key influence factor data.

8. An electronic device, characterized in that, Including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the method according to any one of claims 1-6 is completed.

9. A computer-readable storage medium, characterized in that, For storing computer instructions, when the computer instructions are executed by the processor, the method according to any one of claims 1-6 is completed.

10. A computer program product, characterized in that, Including a computer program, which, when executed by a processor, implements the method according to any one of claims 1-6.

Images