Low-voltage prediction model training method and device, storage medium and electronic equipment

Abstract

The application provides a low-voltage prediction model training method and device, a storage medium and an electronic device. The electronic device obtains a to-be-analyzed data set; the to-be-analyzed data set includes multiple groups of original records before a low voltage occurs in a power distribution network, and each group of original records includes observation values of multiple influence factors; according to the to-be-analyzed data set, multiple significant influence factors that have a significant influence on the low voltage are selected from the multiple influence factors; and a training data set is constructed according to the multiple significant influence factors; the training data set includes multiple training samples, and each training sample includes observation values of the multiple significant influence factors and a voltage state label corresponding to a prediction time point; and a to-be-trained model is trained through the training data set to obtain a low-voltage prediction model. In this way, by extracting significant influence factors, the interference of redundant information on model training can be effectively reduced, and then the prediction performance and stability of the trained model can be significantly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power grid, in particular to a low-voltage prediction model training method and device, a storage medium and an electronic equipment. BACKGROUND

[0002] With the continuous growth of electricity demand, there is a significant phenomenon of asynchronization between the speed of power distribution network transformation and the growth of electricity demand, which directly leads to frequent low-voltage problems in the power distribution network. In order to effectively cope with this challenge, it is particularly urgent and necessary to carry out prediction research on the changing situation of low-voltage in the power distribution network.

[0003] However, to train an efficient and accurate prediction model, it is inseparable from high-quality and suitable data support. The power distribution network data itself has high complexity, and is extremely susceptible to various internal and external factors, such as weather conditions, equipment operating state, and user electricity behavior. Therefore, in the process of building a prediction model, it is one of the key links to determine the effect of the model to clearly select which types of data as input features. SUMMARY

[0004] In order to overcome at least one of the deficiencies in the prior art, the present application provides a low-voltage prediction model training method, device, storage medium and electronic equipment, which specifically includes:

[0005] In a first aspect, the present application provides a low-voltage prediction model training method, which includes:

[0006] Obtaining a data set to be analyzed, wherein the data set to be analyzed includes a plurality of original records before the occurrence of low voltage in the power distribution network, and each original record includes observation values of a plurality of influence factors;

[0007] According to the data set to be analyzed, a plurality of significant influence factors having a significant influence on the low voltage are selected from the plurality of influence factors;

[0008] According to the plurality of significant influence factors, a training data set is constructed, wherein the training data set includes a plurality of training samples, and each training sample includes observation values of the plurality of significant influence factors and a voltage state label corresponding to a prediction time point;

[0009] The training data set is used to train a to-be-trained model to obtain a low-voltage prediction model.

[0010] In a second aspect, the present application provides a low-voltage prediction model training device, which includes:

[0011] The data preparation module is configured to obtain a data set to be analyzed, wherein the data set to be analyzed comprises a plurality of original records before low voltage occurs in the power distribution network, and each of the original records comprises observation values of a plurality of influence factors;

[0012] The data preparation module is further configured to select, from the plurality of influence factors, a plurality of significant influence factors that have a significant influence on the low voltage according to the data set to be analyzed.

[0013] The data preparation module is further configured to construct a training data set according to the plurality of significant influence factors, wherein the training data set comprises a plurality of training samples, and each of the training samples comprises observation values of the plurality of significant influence factors and a voltage state label corresponding to a prediction time point.

[0014] The model training module is configured to train a to-be-trained model by using the training data set to obtain a low voltage prediction model.

[0015] In a third aspect, the present application provides a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the low voltage prediction model training method.

[0016] In a fourth aspect, the present application provides an electronic device comprising a processor and a memory, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the low voltage prediction model training method.

[0017] Compared with the prior art, the present application has the following beneficial effects:

[0018] The present application provides a low voltage prediction model training method, device, storage medium and electronic device. The electronic device obtains a data set to be analyzed; wherein the data set to be analyzed comprises a plurality of original records before low voltage occurs in the power distribution network, and each of the original records comprises observation values of a plurality of influence factors; a plurality of significant influence factors that have a significant influence on the low voltage are selected from the plurality of influence factors according to the data set to be analyzed; a training data set is constructed according to the plurality of significant influence factors; wherein the training data set comprises a plurality of training samples, and each of the training samples comprises observation values of the plurality of significant influence factors and a voltage state label corresponding to a prediction time point; and a to-be-trained model is trained by using the training data set to obtain a low voltage prediction model. In this way, by obtaining the data set to be analyzed comprising observation values of a plurality of influence factors, and extracting significant influence factors based on statistical analysis or feature selection algorithm, the interference of redundant information on model training can be effectively reduced; therefore, when a to-be-trained model is trained based on the optimized training data set, the model can more accurately capture the core law of low voltage occurrence, thereby significantly improving the prediction performance and stability. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those of ordinary skill in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0020] Figure 1 A flowchart of a low-voltage prediction model training method provided by the embodiments of the present application is shown in the figure.

[0021] Figure 2 One of the implementation details of the low-voltage prediction model training method provided by the embodiments of the present application is shown in the figure.

[0022] Figure 3 Another of the implementation details of the low-voltage prediction model training method provided by the embodiments of the present application is shown in the figure.

[0023] Figure 4 A structure diagram of a low-voltage prediction model training device provided by the embodiments of the present application is shown in the figure.

[0024] Figure 5 A structure diagram of an electronic device provided by the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present application (hereinafter referred to as the present embodiments) clearer, the technical solutions of the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0026] Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art on the basis of the embodiments in the present application without creative labor are within the scope of the present application.

[0027] It should be noted that: similar reference numbers and letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0028] In the description of the present application, it should be noted that the terms "first", "second", "third" and the like are only used to distinguish descriptions and cannot be understood as indicating or implying relative importance. In addition, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or equipment including the element.

[0029] Based on the above statement, as introduced in the background art, the power distribution network data itself has high complexity, and is also easily affected by various internal and external factors. Therefore, in the process of building the prediction model, it is one of the key links to determine the model effect to clearly select which types of data as input features.

[0030] Exemplarily, these external factors can include the low voltage condition of the day, voltage, line radius, number of low voltage occurrences per day, low voltage duration, load rate, whether it is an urban area or a rural area, three-phase voltage value, whether it is a holiday, whether the equipment fails, equipment running time, daily maximum temperature, minimum temperature, average temperature of the transformer area, whether there is lightning, whether there is icing, rainfall, wind speed, etc.

[0031] The above-mentioned influence factors cover multiple dimensions such as power system operation, equipment state and external environment, and some of the factors will be explained next.

[0032] The low voltage condition of the day, since the occurrence of low voltage often has certain inertia characteristics, that is, if low voltage has occurred in a certain transformer area on a certain day, the probability of low voltage occurring again the next day will increase significantly. Therefore, this factor can be an important reference for predicting future low voltage.

[0033] Load rate directly measures the proportional relationship between current power demand and power supply capacity. High load rate usually means greater current flow, which will cause the resistance voltage drop in the line to increase, thereby causing low voltage problems. Therefore, this factor can be an important reference for predicting future low voltage.

[0034] Geographical area attributes are also important for low voltage prediction. There are significant differences in power consumption patterns between urban and rural areas, for example, rural areas may rely more on agricultural loads, while urban areas are dominated by residential and commercial electricity. This difference directly affects the shape of the load curve and its corresponding voltage variation law, so this factor can be an important reference for predicting future low voltage.

[0035] Weather factors (daily maximum temperature, minimum temperature, average temperature, whether there is lightning, whether there is icing, rainfall, and wind speed, etc.) can also directly or indirectly affect the voltage level. For example, under low temperature conditions and high temperature conditions, the user's power consumption behavior will change significantly, resulting in changes in power demand. Therefore, these factors can be important references for predicting future low voltage.

[0036] Based on the discovery of the above technical problems, the following technical solutions are proposed after creative labor to solve or improve the above problems. It should be noted that the defects in the above prior art solutions are the result of careful research and practice, so the discovery process of the above problems and the solutions proposed by the embodiments of the present application to solve the above problems should be considered as contributions to the present application in the process of invention and creation, and should not be understood as technical content known to those skilled in the art.

[0037] In view of the above technical problems, the embodiment provides a low voltage prediction model training method. As Figure 1 The method comprises:

[0038] S1, obtaining a to-be-analyzed data set.

[0039] The to-be-analyzed data set includes a plurality of original records before the occurrence of low voltage in the power distribution network, and each original record includes observation values of a plurality of influence factors.

[0040] S2, according to the to-be-analyzed data set, a plurality of significant influence factors having a significant influence on low voltage are selected from a plurality of influence factors.

[0041] S3, constructing a training data set according to the plurality of significant influence factors.

[0042] The training data set includes a plurality of training samples, and each training sample includes observation values of a plurality of significant influence factors and a voltage state label corresponding to a prediction time point.

[0043] S4, training the to-be-trained model through the training data set to obtain a low voltage prediction model.

[0044] In this way, by obtaining the to-be-analyzed data set containing observation values of a plurality of influence factors, and extracting significant influence factors based on statistical analysis or feature selection algorithm, the interference of redundant information on model training can be effectively reduced. Therefore, when the to-be-trained model is trained based on the optimized training data set, the model can more accurately capture the core law of low voltage occurrence, thereby significantly improving the prediction performance and stability.

[0045] In this embodiment, the electronic device implementing the low-voltage prediction model training method can be, but is not limited to, a mobile terminal, a tablet computer, a laptop computer, a desktop computer, a server, etc. As long as it can provide sufficient computing power for model training. When it is a server, the server can be a single server or a server group. The server group can be centralized or distributed (for example, the server can be a distributed system). In some embodiments, the server can be local or remote relative to the user terminal. In some embodiments, the server can be implemented on a cloud platform; for example, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an inter-cloud, a multi-cloud, etc., or any combination thereof. In some embodiments, the server can be implemented on an electronic device with one or more components.

[0046] To make the scheme provided by the present embodiment clearer, the following will take the server as the electronic device implementing the low-voltage prediction model training method, and elaborate on each step of the method shown in Figure 1 It should be understood that the operations of the flowchart can not be implemented in sequence, and the steps without logical context relationship can be reversed in order or implemented simultaneously. In addition, one or more other operations can be added to the flowchart or one or more operations can be removed from the flowchart under the guidance of the content of the present application. Continue to refer to Figure 1 The method comprises:

[0047] S1, obtaining a to-be-analyzed data set.

[0048] The to-be-analyzed data set comprises a plurality of original records before the occurrence of low voltage in the power distribution network, and each original record comprises observation values of a plurality of influence factors.

[0049] Specifically, these data are derived from historical operation data of a plurality of transformer areas and transformer area meteorological data. The historical operation data includes basic information such as line type, line length, equipment operation time, and equipment fault information. In addition, the historical operation data also includes historical load and voltage data, such as daily load rate, transformer area load, heavy overload data, voltage data, and low voltage conditions from September 2023 to August 2024. The transformer area meteorological data includes meteorological data of each transformer area every day throughout the year, such as air temperature, rainfall, wind speed, icing, lightning, and other weather conditions, and information about whether it is a holiday.

[0050] To ensure the validity of the data, after obtaining the data, preliminary data processing work is needed. Specifically, the server filters the transformer area with missing data, and excludes the transformer area that does not meet the conditions. For example, if some transformer areas lack key information, such as line information, device operation information or voltage data, and these missing data cannot be filled by correlation, these transformer areas will be excluded. In this way, it is ensured that the data used in subsequent analysis and modeling has high quality and integrity.

[0051] For those data that can be completed by calculation or other methods, appropriate measures are taken to fill in, for example, the number of times and the duration of low voltage occurrence of the transformer area per day can be calculated from the three-phase voltage data per day. The expression for calculating the voltage unbalance degree is:

[0052]

[0053] In the formula, represents the voltage unbalance degree, represents the maximum negative sequence voltage, represents the average voltage.

[0054] The server compares the voltage unbalance degree with the lowest value in the three-phase voltage and the standard voltage value. When the lowest value is lower than 80% of the standard voltage and the voltage unbalance degree exceeds 40%, it is determined that low voltage has occurred.

[0055] Based on the above description of the data set to be analyzed, the following step S2 in the method is described: Figure 1

[0056] S2, according to the data set to be analyzed, screening a plurality of significant influence factors from a plurality of influence factors which have significant influence on low voltage.

[0057] In order to screen the plurality of influence factors, the embodiment calculates the influence score of each influence factor on low voltage. As an optional implementation manner, as shown in Figure 2 , step S2 can include:

[0058] S2-1, according to the data set to be analyzed, obtaining the influence score of the plurality of influence factors on low voltage.

[0059] The influence score of each influence factor on low voltage can be evaluated in various ways. In this embodiment, the server analyzes the data set to be analyzed by using the PCA algorithm to obtain the influence score of the plurality of influence factors on low voltage.

[0060] S2-2, according to the influence score of the plurality of influence factors on low voltage, selecting a plurality of significant influence factors from the plurality of influence factors. ​

[0061] wherein the plurality of significant influencing factors are the top N influencing factors with the largest influence scores.

[0062] It should be understood that, in the process of constructing the low-voltage prediction model, screening the significant influencing factors is of great significance to the subsequent training of the model. That is, it is necessary to screen a few key factors that have the most significant influence on the low-voltage phenomenon from a large number of possible influencing factors.

[0063] Exemplarily, the data set to be analyzed contains various information accumulated during the operation of the power distribution network, such as the low-voltage condition of the day, the voltage value, the line radius, the number of low-voltage occurrences of the day, the low-voltage duration, the load rate, the urban or rural area attribute, the three-phase voltage value, the holiday status, the equipment failure condition, the equipment operation time, and the meteorological condition (such as the maximum air temperature, the minimum air temperature, the average air temperature, lightning, icing, rainfall, and wind speed).

[0064] In order to quantify the importance of each influencing factor, the present embodiment introduces the principal component analysis (PCA). Through PCA analysis, the server can evaluate the specific contribution of each influencing factor to the low-voltage condition, and screen the top N factors with the largest influence scores as the significant influencing factors. After the above analysis, 10 influencing factors that have the greatest influence on the low-voltage condition are finally screened, including the low-voltage condition of the day, the voltage value, the line condition, the urban or rural area attribute, whether it is a holiday, whether the equipment has failed, the equipment operation time, the daily maximum air temperature, the daily rainfall, and whether there is icing.

[0065] S3, constructing a training data set according to the plurality of significant influencing factors.

[0066] wherein the training data set includes a plurality of training samples, and each training sample includes observation values of the plurality of significant influencing factors and a voltage state label corresponding to the prediction time point.

[0067] Further research has found that, in the process of constructing the low-voltage prediction model, the traditional non-time sequence model often cannot capture the dynamic changes of the influencing factors. However, in the actual operation of the power system, many significant influencing factors (such as air temperature, equipment load rate, or meteorological condition) are not static but change constantly over time. These dynamic changes may be important indications of the occurrence of low-voltage phenomena.

[0068] Specifically, the low voltage phenomenon in the power distribution network is usually the result of the joint action of multiple factors. For example, when the temperature suddenly rises at a certain moment, it may cause a sharp increase in electricity demand, leading to voltage drop in the local area; or if a device fails after a long time of high load operation, it may also quickly lead to low voltage problems in the region. These dynamic changes are crucial for accurate prediction of low voltage, but traditional non-temporal models lack attention to the time dimension.

[0069] For this problem, each training sample in the embodiment also includes the change information of the part of the significant influencing factors from the sampling time point to the prediction time point.

[0070] It should be understood that when constructing the training data set, for part of the significant influencing factors, the actual observation value at the prediction time point and the observation value at the sampling time point can be explicitly known. Therefore, by comparing the actual observation value at the prediction time point with the observation value at the previous sampling time point, the change information of the significant influencing factor can be calculated.

[0071] But in the actual reasoning process, for the same significant influencing factor, considering that the actual observation value at a future time point cannot be known in advance, but in order to still be able to obtain the change information of the significant influencing factor, the prediction value of the significant influencing factor at a future time point can be obtained by means of a time-based prediction model.

[0072] Exemplarily, suppose it is necessary to predict whether a certain district will have a low voltage phenomenon tomorrow, and the temperature, load rate and wind speed are screened out as significant influencing factors. In the training phase, it can be obtained from the historical data that the temperature at 8 o'clock this morning is 25°C, the load rate is 70%, and the wind speed is 5m / s; at the same time, it can also be obtained from the historical data that the actual temperature at 4 o'clock tomorrow afternoon is 30°C, the load rate is 80%, and the wind speed is 6m / s. By comparing the data at these two time points, we can calculate the change information of the temperature as "+5°C", the change information of the load rate as "+10%", and the change information of the wind speed as "+1m / s". Carrying these change information in the training sample is used to train the low voltage prediction model.

[0073] In the actual inference phase, assuming the temperature at 8 AM today was 26°C, the load factor was 80%, and the wind speed was 7 m / s, since the temperature, load factor, and wind speed at 4 PM tomorrow cannot be known in advance, a time-series prediction model (e.g., LSTM) can be used to predict that the temperature at 4 PM tomorrow will be 31°C, the load factor 82%, and the wind speed 7 m / s. These predicted values ​​are then compared with the actual observations at 8 AM today, calculating the temperature change as "+5°C", the load factor change as "+2%", and the wind speed change as "+0 m / s". This change information, along with the observed values ​​of other significant influencing factors, is input into the pre-trained low-voltage prediction model to obtain a clear prediction of whether low voltage will occur.

[0074] In this way, the model can more comprehensively capture the dynamic changes of influencing factors, thereby enhancing the training effect of the model.

[0075] Based on the explanation of the training dataset in step S3 in the above embodiments, the following will continue to discuss... Figure 1 Step S4 will be explained below:

[0076] S4. Train the model to be trained using the training dataset to obtain the low voltage prediction model.

[0077] Research has found that different hyperparameter settings significantly impact model performance during the training of low-voltage prediction models. For example, when training a random forest model, the number of trees, maximum depth, and minimum number of samples per leaf node are all hyperparameters. The choice of these hyperparameters directly affects the model's structure and behavior. If the number of trees is too small, the model may be too simple and unable to fully capture the complex patterns in the data; while if the number of trees is too large, the model may become too complex, leading to overfitting. Similarly, the maximum depth setting also affects the model's fit to the data: too little depth may lead to underfitting, while too much depth may introduce excessive noise.

[0078] In view of the above findings, such as Figure 3 As shown, in Figure 2 Based on this, this embodiment provides the following optional implementation methods for step S4:

[0079] S4-1 initializes the model to be trained using multiple sets of hyperparameters, resulting in multiple initialized models to be trained.

[0080] Exemplarily, continue to take the random forest as an example. The random forest relies on the setting of several key hyperparameters, such as the number of trees, the maximum depth of the tree, the minimum sample number of the leaf node, and the like. The selection of these hyperparameters directly affects the complexity, generalization ability, and prediction accuracy of the model. However, in actual applications, due to the complex interaction relationship between the hyperparameters, it is difficult to directly determine the optimal combination of hyperparameters by intuition or experience. In this regard, the embodiment can traverse all possible combinations of parameters in a preset parameter range by the grid method for initializing the random forest, so as to obtain the initialized random forest. The parameter range of each hyperparameter of the random forest tree model can refer to the following ranges:

[0081] The number of trees ranges from 1 to 300, the maximum depth of the tree is 31, the minimum sample number required is from 2 to 11, the minimum sample number of the leaf node is from 1 to 4, and whether to return the sample (yes). The maximum number of iterations is set to 1000, and the target error is 0.05%.

[0082] S4-2, training the plurality of initialized to-be-trained models by using the training data set to obtain a plurality of candidate models.

[0083] S4-3, selecting a plurality of preferred models from the plurality of candidate models.

[0084] In order to select a plurality of preferred models from the plurality of candidate models, the training effect of the plurality of candidate models needs to be quantitatively evaluated. In comparison, the embodiment provides the following optional implementation manner of step S4-3:

[0085] S4-3-1, obtaining the training effect of the plurality of candidate models after being trained by using the training data set.

[0086] Optionally, the embodiment evaluates the training effect of the candidate model by comprehensively considering multiple dimensions. For each candidate model, the server determines, from the plurality of training samples, to-be-counted samples predicted as low voltage by the candidate model; determines, from all the to-be-counted samples, the number of positive samples, wherein the positive sample represents a training sample with a voltage state label of low voltage; obtains a first proportion between the number of samples and the total amount of to-be-counted samples, and a second proportion between the number of samples and the total amount of positive samples in the plurality of training samples; calculates a proportion product between the first proportion and the second proportion, and a proportion sum between the first proportion and the second proportion; and obtains the training effect of the candidate model according to the ratio between the proportion product and the proportion sum.

[0087] The expression for obtaining the training effect of the candidate model according to the ratio between the proportion product and the proportion sum is:

[0088]

[0089] In the formula, a training score for characterizing the training effect, represents the first proportion, represents the second proportion.

[0090] Exemplarily, it is assumed that there are 1000 training samples in the training data set, of which 800 samples have a label of low voltage (i.e. positive samples), and the remaining 200 samples are non-low voltage (i.e. negative samples). The candidate model identifies 1000 training samples, and predicts 750 training samples as positive samples and 250 training samples as negative samples. Of the 750 training samples predicted as positive samples, only 600 training samples actually have a label of low voltage, and therefore the first proportion is 600 / 750 = 0.8. Similarly, the second proportion is 600 / 800 = 0.75. The first proportion and the second proportion are brought into the above expression, and the training score is obtained as:

[0091]

[0092] S4-3-2, selecting a plurality of preferred models from the plurality of candidate models according to the training effect of the plurality of candidate models on the training data set.

[0093] Optionally, the plurality of candidate models are filtered by a preset score threshold. The score threshold can be 0.9, and the preferred model is selected as the training score greater than 0.9.

[0094] The selection method of the preferred model is explained in the above embodiment, and the step S4 further includes: Figure 3

[0095] S4-4, selecting a low voltage prediction model from the plurality of preferred models according to the test results of the plurality of preferred models on the test data set.

[0096] It should be understood that if a certain preferred model performs too well on the training data (i.e. the training score is very high), it may over-learn the specific noise or details in the training data, resulting in a decline in its performance on unseen data (such as test data). In order to further evaluate the generalization ability of the preferred model, a test data set is introduced. The test data set is composed of data that does not participate in the training, and therefore can effectively reflect the performance of the model on unknown data. Each preferred model is verified in order from high to low according to the training score, and the test results of the plurality of preferred models are obtained; and the test result best is selected as the low voltage prediction model.

[0097] ​Based on the same inventive concept as the low-voltage prediction model training method provided in the embodiment, the embodiment further provides a low-voltage prediction model training method device, which comprises at least one software function module stored in the form of software in the memory or solidified in the electronic device. The processor in the electronic device is used to execute the executable modules stored in the memory. For example, the software function modules and computer programs included in the device. Please refer to Figure 4 Functionally, the device can include:

[0098] The data preparation module 11 is configured to obtain a to-be-analyzed data set, wherein the to-be-analyzed data set comprises a plurality of original records before the low-voltage of the power distribution network, and each original record comprises observation values of a plurality of influence factors.

[0099] The data preparation module 11 is further configured to select a plurality of significant influence factors from the plurality of influence factors according to the to-be-analyzed data set, wherein the plurality of significant influence factors have a significant influence on the low-voltage.

[0100] The data preparation module 11 is further configured to construct a training data set according to the plurality of significant influence factors, wherein the training data set comprises a plurality of training samples, and each training sample comprises observation values of the plurality of significant influence factors and a voltage state label corresponding to a prediction time point.

[0101] The model training module 12 is configured to train the to-be-trained model by using the training data set to obtain a low-voltage prediction model.

[0102] In the embodiment, the data preparation module 11 is configured to implement steps S1, S2 and S3 in Figure 1 The model training module 12 is configured to implement step S4 in Figure 1 Therefore, the detailed description of each module can be found in the specific embodiments of the corresponding steps.

[0103] Optionally, the data preparation module 11 is further configured to:

[0104] According to the to-be-analyzed data set, obtain an influence score of the plurality of influence factors on the low-voltage;

[0105] According to the influence score of the plurality of influence factors on the low-voltage, select a plurality of significant influence factors from the plurality of influence factors, wherein the plurality of significant influence factors are the top N influence factors with the largest influence score.

[0106] Optionally, the data preparation module 11 is further configured to:

[0107] Analyze the to-be-analyzed data set by using the PCA algorithm to obtain the influence score of the plurality of influence factors on the low-voltage.

[0108] Optionally, the model training module 12 is further configured to:

[0109] The to-be-trained model is initialized by multiple groups of hyperparameters to obtain multiple initialized to-be-trained models;

[0110] The multiple initialized to-be-trained models are respectively trained by a training data set to obtain multiple candidate models;

[0111] The multiple preferred models are selected from the multiple candidate models;

[0112] According to the test results of the multiple preferred models on the test data set, a low-voltage prediction model is selected from the multiple preferred models.

[0113] Optionally, the model training module 12 is further configured to:

[0114] Obtain the training effect of the multiple candidate models after being trained by the training data set;

[0115] According to the training effect of the multiple candidate models after being trained by the training data set, the multiple preferred models are selected from the multiple candidate models.

[0116] Optionally, the model training module 12 is further configured to:

[0117] For each candidate model, the to-be-counted samples predicted as low voltage by the candidate model are determined from the multiple training samples;

[0118] The number of positive samples is determined from all the to-be-counted samples, wherein the positive sample represents the training sample with the voltage state label as low voltage;

[0119] Obtain the first proportion between the number of samples and the total amount of to-be-counted samples, and the second proportion between the number of samples and the total amount of positive samples in the multiple training samples;

[0120] Calculate the proportion product between the first proportion and the second proportion, and the proportion sum between the first proportion and the second proportion;

[0121] According to the ratio between the proportion product and the proportion sum, the training effect of the candidate model is obtained.

[0122] In addition, each functional module in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0123] It should also be understood that the above implementation, if implemented in the form of a software functional module and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art that makes a contribution or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0124] Therefore, the embodiment also provides a storage medium, which is a computer readable storage medium. The storage medium stores a computer program. When the computer program is executed by a processor, the low-voltage prediction model training method provided by the embodiment is implemented. The storage medium can be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0125] The embodiment provides an electronic device for implementing the low-voltage prediction model training method. As shown in the Figure 5 The electronic device can include a processor 22 and a memory 21. The memory 21 stores a computer program. The processor reads and executes the computer program corresponding to the above implementation in the memory 21, and implements the low-voltage prediction model training method provided by the embodiment.

[0126] Continuing to refer to Figure 5 The electronic device further includes a communication unit 23. The memory 21, the processor 22 and the communication unit 23 are directly or indirectly electrically connected to each other through a system bus 24 to realize data transmission or interaction.

[0127] The memory 21 can be an information recording device based on any electronic, magnetic, optical or other physical principles for recording execution instructions, data, etc. In some embodiments, the memory 21 can be, but is not limited to, a volatile memory, a non-volatile memory, a storage drive, etc.

[0128] In some embodiments, the volatile memory can be a Random Access Memory (RAM); in some embodiments, the non-volatile memory can be a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electric Erasable Programmable Read-Only Memory (EEPROM), a flash memory, or the like; in some embodiments, the storage drive can be a magnetic disk drive, a solid state drive, any type of storage disk (such as an optical disk, a DVD, and the like), or a similar storage medium, or a combination thereof, and the like.

[0129] The communication unit 23 is configured to transceive data over a network. In some embodiments, the network can include a wired network, a wireless network, a fiber optic network, a telecommunications network, an intranet, the Internet, a Local Area Network (LAN), a Wide Area Network (WAN), a Wireless Local Area Network (WLAN), a Metropolitan Area Network (MAN), a Wide Area Network (WAN), a Public Switched Telephone Network (PSTN), a Bluetooth network, a ZigBee network, or a Near Field Communication (NFC) network, or the like, or any combination thereof. In some embodiments, the network can include one or more network access points. For example, the network can include wired or wireless network access points, such as base stations and / or network switching nodes, through which one or more components of the service request processing system can connect to the network to exchange data and / or information.

[0130] The processor 22 may be an integrated circuit chip with signal processing capabilities, and may include one or more processing cores (e.g., a single-core processor or a multi-core processor). By way of example only, the processor described above may include a Central Processing Unit (CPU), an Application Specific Integrated Circuit (ASIC), an Application Specific Instruction-set Processor (ASIP), a Graphics Processing Unit (GPU), a Physics Processing Unit (PPU), a Digital Signal Processor (DSP), a Field Programmable Gate Array (FPGA), a Programmable Logic Device (PLD), a controller, a microcontroller unit, a Reduced Instruction Set Computing (RISC) computer, or a microprocessor, or any combination thereof.

[0131] Understandable. Figure 5 The structure shown is for illustrative purposes only. The electronic device 100 may also have... Figure 5 Showing more or fewer components, or having with Figure 5 The different configurations shown. Figure 5 The components shown can be implemented using hardware, software, or a combination thereof.

[0132] It should be understood that all the devices and methods disclosed in the above embodiments can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the drawings show the possible implementation architectures, functions and operations of the devices, methods and computer program products according to the embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different orders from those noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and the combination of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0133] The above describes only various embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A low-voltage predictive model training method, characterized in that, The method comprises: acquiring a to-be-analyzed data set, wherein the to-be-analyzed data set comprises a plurality of groups of original records before low voltage occurs in a power distribution network, and each group of original records comprises observation values of a plurality of influence factors; screening a plurality of significant influence factors having a significant influence on the low voltage from the plurality of influence factors according to the to-be-analyzed data set; constructing a training data set according to the plurality of significant influence factors, wherein the training data set comprises a plurality of training samples, each training sample comprises observation values of the plurality of significant influence factors and a voltage state label corresponding to a prediction time point, and each training sample further comprises change information of part of the significant influence factors from a collection time point of the observation values to the prediction time point; in an actual reasoning process, for the same significant influence factor, a prediction value of the significant influence factor at a future time point is obtained by means of a time-series-based prediction model, and the prediction value is compared with an actual observation value to obtain the change information; training a to-be-trained model by using the training data set to obtain a low-voltage prediction model. 2.The low-voltage predictive model training method of claim 1, wherein, The method comprises: obtaining influence scores of the plurality of influence factors on low voltage according to the to-be-analyzed data set; selecting the plurality of significant influence factors from the plurality of influence factors according to the influence scores of the plurality of influence factors on low voltage, wherein the plurality of significant influence factors are the top N influence factors with the largest influence scores. 3.The low-voltage predictive model training method of claim 2, wherein, The method comprises: analyzing the to-be-analyzed data set by using a PCA algorithm to obtain the influence scores of the plurality of influence factors on low voltage. 4.The low-voltage predictive model training method of claim 1, wherein, The method comprises: initializing the to-be-trained model by using a plurality of groups of hyperparameters to obtain a plurality of initialized to-be-trained models; training the plurality of initialized to-be-trained models by using the training data set to obtain a plurality of candidate models; selecting a plurality of preferred models from the plurality of candidate models; selecting a low-voltage prediction model from the plurality of preferred models according to test results of the plurality of preferred models on a test data set.

5. The low-voltage predictive model training method of claim 4, wherein, The to-be-trained model is a random forest model, and the method of selecting the plurality of preferred models from the plurality of candidate models comprises: obtaining training effects of the plurality of candidate models after being trained by using the training data set; selecting the plurality of preferred models from the plurality of candidate models according to the training effects of the plurality of candidate models after being trained by using the training data set.

6. The low-voltage predictive model training method of claim 5, wherein, The method of obtaining the training effects of the plurality of candidate models after being trained by using the training data set comprises: for each candidate model, determining to-be-counted samples predicted as low voltage by the candidate model from the plurality of training samples; determining a sample number of positive samples from all the to-be-counted samples, wherein the positive samples represent training samples with a voltage state label of low voltage; obtain a first proportion between the sample quantity and a total quantity of the samples to be counted, and a second proportion between the sample quantity and a total quantity of positive samples in the plurality of training samples; calculate a product of proportions between the first proportion and the second proportion, and a sum of proportions between the first proportion and the second proportion; obtain a training effect of the candidate model according to a ratio between the product of proportions and the sum of proportions. 7.A low-voltage predictive model training apparatus, characterized in that, The apparatus comprises: a data preparation module configured to obtain a data set to be analyzed, wherein the data set to be analyzed comprises a plurality of original records before low voltage occurs in a power distribution network, and each of the original records comprises observation values of a plurality of influence factors; the data preparation module is further configured to select a plurality of significant influence factors having a significant influence on the low voltage from the plurality of influence factors according to the data set to be analyzed; the data preparation module is further configured to construct a training data set according to the plurality of significant influence factors, wherein the training data set comprises a plurality of training samples, each of the training samples comprises observation values of the plurality of significant influence factors and a voltage state label corresponding to a prediction time point, and each of the training samples further comprises change information of part of the significant influence factors from a collection time point of the observation values to the prediction time point; in an actual reasoning process, for the same significant influence factor, a prediction value of the significant influence factor at a future time point is obtained by means of a time series-based prediction model, and the prediction value is compared with an actual observation value to obtain the change information; a model training module configured to train a to-be-trained model by using the training data set to obtain a low voltage prediction model.

8. A storage medium, characterized by The storage medium stores a computer program, and the computer program, when executed by a processor, implements the low voltage prediction model training method in any one of claims 1-6.

9. An electronic device, comprising: The electronic device comprises a processor and a memory, and the memory stores a computer program, and the computer program, when executed by the processor, implements the low voltage prediction model training method in any one of claims 1-6.

Images