Influence value determination method and product

Abstract

The embodiment of the invention provides an influence value determination method and product, which are used for solving the problem in the prior art that the prediction demand on the influence of meteorological disasters under the real condition is difficult to meet. In the embodiment of the invention, after the electronic equipment receives the prediction request, the electronic equipment can obtain the meteorological data of at least one preset meteorological attribute and input the obtained meteorological data into the pre-trained prediction model, and the prediction model is a model based on Stacking ensemble learning. According to the method, strong correlation factors influencing residential electricity load characteristics under abnormal weather such as violent snow and the like are fully considered, and the method has certain effectiveness and rationality.

Description

Technical Field

[0001] The invention relates to the field of power grid use dispatching assessment, and in particular to a method and product based on impact value determination. Background Art

[0002] In recent years, with the continuous intensification of the greenhouse effect, extreme weather disasters have occurred frequently around the world, causing serious impacts on the normal operation of the power system. Extreme blizzard weather will cause serious damage to the operation of the power grid and affect the electricity load of residents. Therefore, it is of great research significance to accurately and timely predict the distribution characteristics of urban residents' electricity load under blizzard weather, and then allocate and dispatch emergency resources according to the predicted situation, which will reduce economic losses and optimize the planning and management of power resources.

[0003] At present, domestic and foreign scholars mainly use probability statistics formulas and machine learning to predict the various impacts of disastrous weather, but there are relatively few studies on the evaluation of residential electricity load characteristics, which mainly use traditional supervised learning methods. However, traditional supervised learning methods are difficult to meet the needs of predicting the impact of meteorological disasters in real situations. Summary of the invention

[0004] The present application provides a method and product for determining impact values, which is used to solve the problem that the existing technology is difficult to meet the demand for predicting the impact of meteorological disasters under real conditions.

[0005] In a first aspect, an embodiment of the present application provides a method for determining an impact value, the method comprising:

[0006] If a forecast request is received, obtaining meteorological data of at least one preset meteorological attribute;

[0007] The acquired meteorological data is input into a pre-trained prediction model to obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

[0008] Furthermore, the method further comprises:

[0009] The type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge is obtained.

[0010] Furthermore, the prediction model includes a first preset number of base learners and a meta-learner for combining the base learners; wherein the first preset number of base learners respectively adopt decision tree, K-nearest neighbor, and lightweight gradient boosting machine LGBM algorithms.

[0011] Furthermore, the prediction model is trained in the following manner:

[0012] Obtain any meteorological data group pre-stored in a pre-stored training set, and obtain a value of the annotated impact on residential electricity charge stored for the meteorological data group;

[0013] Inputting the meteorological data group into the original prediction model before the prediction model is trained, and obtaining the output impact value of the original prediction model on the residential electricity charge;

[0014] The original prediction model is trained based on the labeled influence value and the output influence value.

[0015] Further, the obtaining of meteorological data of at least one preset meteorological attribute includes:

[0016] Obtaining each pre-set meteorological attribute that is previously saved and affects residential electricity charge;

[0017] Acquire meteorological data of each preset meteorological attribute.

[0018] Furthermore, each meteorological attribute includes: minimum wind speed, maximum wind speed, average wind speed, snowfall intensity, air temperature, coverage, geographical location, and duration of the snowstorm disaster.

[0019] Further, each preset meteorological attribute affecting the residential electricity charge is pre-saved in the following manner:

[0020] Get each weather attribute saved;

[0021] For each of the meteorological attributes, based on the data of the meteorological attribute and the residential electricity charge, calculate the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge;

[0022] The meteorological attributes whose determined Pearson correlation coefficients are within a preset value range are determined as preset meteorological attributes that affect residential electricity charges.

[0023] Furthermore, the Pearson correlation coefficient between the meteorological attribute data and the residential electricity charge is calculated by the following formula:

[0024]

[0025] Where n is the number of samples, X i is the i-th data of the meteorological attribute, Y is the i-th charge data of the residential electricity load, is the average value of each data of the meteorological attribute, is the average value of the electricity charge used by each resident, and r is the Pearson correlation coefficient.

[0026] Furthermore, the training set is constructed in the following manner:

[0027] Obtaining meteorological data of each preset meteorological attribute within a preset time period and each impact value on residential electricity charge;

[0028] For each preset meteorological attribute, determine sample meteorological data according to each meteorological data of the preset meteorological attribute; and determine a marked influence value according to each influence value within a preset time period;

[0029] A meteorological data group is determined according to the meteorological data of each preset meteorological attribute, and a corresponding relationship between the marked impact value and the meteorological data group is saved.

[0030] In a second aspect, an embodiment of the present application further provides an impact value determination product, the product comprising:

[0031] an acquisition module, for acquiring meteorological data of at least one preset meteorological attribute if a prediction request is received;

[0032] A processing module is used to input the acquired meteorological data into a pre-trained prediction model to obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

[0033] Furthermore, the processing module is also used to obtain the type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge.

[0034] Furthermore, the processing module is further used to train the prediction model in the following manner:

[0035] Obtain any meteorological data group pre-stored in a pre-stored training set, and obtain a value of the annotated impact on residential electricity charge stored for the meteorological data group;

[0036] Inputting the meteorological data group into the original prediction model before the prediction model is trained, and obtaining the output impact value of the original prediction model on the residential electricity charge;

[0037] The original prediction model is trained based on the labeled influence value and the output influence value.

[0038] Furthermore, the processing module is specifically used to obtain each pre-stored preset meteorological attribute that affects the residential electricity charge;

[0039] Acquire meteorological data of each preset meteorological attribute.

[0040] Furthermore, the processing module is further used to pre-save each preset meteorological attribute affecting the residential electricity charge in the following manner:

[0041] Get each weather attribute saved;

[0042] For each of the meteorological attributes, based on the data of the meteorological attribute and the residential electricity charge, calculate the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge;

[0043] The meteorological attributes whose determined Pearson correlation coefficients are within a preset value range are determined as preset meteorological attributes that affect residential electricity charges.

[0044] Furthermore, the processing module is specifically used to calculate the Pearson correlation coefficient between the meteorological attribute data and the residential electricity charge by the following formula:

[0045]

[0046] Where n is the number of samples, X i is the i-th data of the meteorological attribute, Y is the i-th charge data of the residential electricity load, is the average value of each data of the meteorological attribute, is the average value of the electricity charge used by each resident, and r is the Pearson correlation coefficient.

[0047] Furthermore, the processing module is further used to construct the training set in the following manner:

[0048] Obtaining meteorological data of each preset meteorological attribute within a preset time period and each impact value on residential electricity charge;

[0049] For each preset meteorological attribute, determine sample meteorological data according to each meteorological data of the preset meteorological attribute; and determine a marked influence value according to each influence value within a preset time period;

[0050] A meteorological data group is determined according to the meteorological data of each preset meteorological attribute, and a corresponding relationship between the marked impact value and the meteorological data group is saved.

[0051] Because in an embodiment of the present application, after receiving a prediction request, the electronic device can obtain meteorological data of at least one preset meteorological attribute, and input the obtained meteorological data into a pre-trained prediction model, wherein the prediction model is a model based on Stacking ensemble learning, and obtains the impact value of the meteorological data output by the prediction model on the residential electricity charge. This method fully considers the strong correlation factors that affect the residential electricity load characteristics under abnormal weather conditions such as blizzards, and has certain effectiveness and rationality. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0053] Figure 1 A schematic diagram of a process for determining an influencing value provided in an embodiment of the present application;

[0054] Figure 2 A schematic diagram of the principle of prediction model identification provided in an embodiment of the present application;

[0055] Figure 3 A schematic diagram of the principle of a 5-fold cross validation provided in an embodiment of the present application;

[0056] Figure 4 A flowchart of a Stacking-based ensemble learning method;

[0057] Figure 5 A schematic diagram of the structure of an impact value determination device provided in an embodiment of the present application;

[0058] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0060] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0061] In order to determine the residential electricity charge, the embodiments of the present application provide a method and product for determining an impact value.

[0062] The method for determining the impact value includes: if the electronic device receives a prediction request, it obtains meteorological data of at least one preset meteorological attribute; inputs the obtained meteorological data into a pre-trained prediction model, and obtains the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

[0063] Embodiment 1:

[0064] Figure 1A schematic diagram of a process for determining an impact value provided in an embodiment of the present application, the process includes the following steps:

[0065] S101: If a prediction request is received, obtaining meteorological data of at least one preset meteorological attribute.

[0066] The method for determining an impact value provided in the embodiment of the present application is applied to an electronic device, which may be an intelligent device such as a PC or a server.

[0067] In order to determine the impact on residents' electricity charges, the electronic device can first receive a prediction request, where the prediction request can be sent by the residents from the devices they use, or it can be triggered when abnormal weather such as blizzards is detected, and the prediction request is obtained by itself.

[0068] After obtaining the prediction request, the electronic device may obtain meteorological data of at least one preset meteorological attribute, wherein the preset meteorological attribute may be the minimum wind speed of a blizzard disaster.

[0069] S102: Input the acquired meteorological data into a pre-trained prediction model to obtain the impact value of the meteorological data on residential electricity charge and the type of load change output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

[0070] In order to determine the residential electricity charge of residents, the electronic device locally stores a pre-trained prediction model. After obtaining meteorological data of preset meteorological attributes, the electronic device can input the obtained meteorological data into the prediction model. It should be noted that if meteorological data of at least two preset meteorological attributes are obtained, the at least two meteorological attributes and the meteorological data corresponding to the meteorological attributes are input into the prediction model.

[0071] The prediction model is a model based on Stacking ensemble learning. The prediction model based on Stacking ensemble learning is an advanced and powerful machine learning technology that trains a meta-learner (also called a secondary learner or a strong learner) by taking the outputs of multiple different base learners (also called primary learners or weak learners) as input, thereby integrating the prediction results of these base learners in order to achieve better performance than any single base learner. Moreover, since Stacking allows the use of multiple different base learners, the differences between these learners help capture different aspects of information in the data, thereby improving the generalization ability of the model.

[0072] After the acquired meteorological data is input into a pre-trained prediction model, the electronic device can obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model.

[0073] It should be noted that the prediction model can be used to identify the impact of disasters such as snowstorms on residents' electricity charges in a certain area.

[0074] Because the situations in different regions may be different in actual scenarios, when inputting meteorological data into the prediction model, relevant geographic location information data may also be input into the prediction model, and power grid operation data may also be input into the prediction model, and the power grid operation data may be historical data.

[0075] In a possible implementation, a larger impact value indicates a greater impact on the residential electricity charge, and abnormal situations such as power outages may occur. If the impact value is greater than a preset value, an alarm message may be sent. Specifically, the alarm message may be sent to a preset device, and the staff may issue an early warning.

[0076] The embodiment of the present application proposes a method for predicting the impact of blizzard disasters on residential electricity load based on Stacking ensemble learning, combining the short-term and short-time blizzard forecast information of the meteorological department and the geographical distribution information of the power grid to determine the impact on residential electricity load.

[0077] Because in an embodiment of the present application, after receiving a prediction request, the electronic device can obtain meteorological data of at least one preset meteorological attribute, and input the obtained meteorological data into a pre-trained prediction model, wherein the prediction model is a model based on Stacking ensemble learning, and obtains the impact value of the meteorological data output by the prediction model on the residential electricity charge. This method fully considers the strong correlation factors that affect the residential electricity load characteristics under abnormal weather conditions such as blizzards, and has certain effectiveness and rationality.

[0078] Embodiment 2:

[0079] In order to accurately and effectively determine the degree of impact on the residential electricity charge, based on the above embodiment, in the embodiment of the present application, the method further includes:

[0080] The type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge is obtained.

[0081] In the embodiment of the present application, the electronic device can also obtain the type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge.

[0082] Among them, the type of load change can be normal load or abnormal load. Normal load means that under normal weather conditions, the residential electricity load shows a certain regularity and stability; abnormal load means that under extreme weather conditions such as blizzards, the residential electricity load may change significantly, such as an increase in load due to increased use of heating equipment, or a decrease in load due to power outages.

[0083] The types of load changes can also be divided into meteorological factors and social factors according to the influencing factors. Meteorological factors refer to the temperature drop directly caused by heavy snow, which increases residents' heating demand and thus causes an increase in electricity load; social factors refer to traffic interruptions and power supply facility damage that may be caused by heavy snow, which will indirectly affect residents' electricity consumption behavior, such as reducing going out and increasing indoor activity time, thereby further affecting the electricity load.

[0084] In order to accurately and effectively determine the degree of impact on residential electricity charges, based on the above embodiments, in an embodiment of the present application, the prediction model includes a first preset number of basis learners and a meta-learner for combining basis learners; wherein the first preset number of basis learners respectively adopt decision trees, K-nearest neighbors, and LGBM algorithms.

[0085] The prediction module of the embodiment of the present application includes a first preset number of base learners, wherein the first preset number may be 3, and the three base learners may be located in the first layer of the prediction model, and the second layer of the prediction model is a meta-learner for combining base learners. The three base learners respectively use decision trees, K-nearest neighbors, and LGBM algorithms to predict the impact of blizzard disasters on changes in electricity load in urban residential areas.

[0086] It should be noted that in practical applications, the decision tree algorithm calculates information gain, gain ratio and Gini index by analyzing the meteorological data set, and constructs a decision tree using the Classification and Regression Tree (CART) algorithm. The division of the characteristic threshold of residential power load is determined by calculating the Gini index using the following formula:

[0087]

[0088] Wherein, the Gini index Gini(D,A) represents the uncertainty of the blizzard disaster samples after segmentation by the feature threshold A, D1 and D2 represent the number of blizzard disaster samples in the two subsets after segmentation, and Gini(D1) and Gini(D2) represent the Gini impurity of the two subsets after segmentation.

[0089] In practical applications, the K-nearest neighbor algorithm is used to evaluate the change in residential electricity load in the predicted area under specific blizzard meteorological disaster characteristics. The distance function of the K-nearest neighbor algorithm is calculated using the following formula:

[0090]

[0091] Where d(x,y) is the Euclidean distance between blizzard disaster samples, and n is the dimension of blizzard disaster sample data.

[0092] In practical applications, the LGBM algorithm is based on the distributed gradient boosting framework. It uses the histogram algorithm to discretize the continuous floating-point features in the blizzard disaster samples into k discrete values, constructs a histogram with a width of k, traverses the blizzard disaster training data to count the cumulative statistics of each discrete value in the histogram, and uses the Leaf-Wise strategy to reduce the error and improve the prediction accuracy.

[0093] Furthermore, the logistic regression model of the second layer uses the following formula for prediction:

[0094]

[0095] In the formula, h θ (x) is the result of logistic regression, with a value range of (0,1), indicating the probability value of the impact of blizzard disaster on residential electricity load, and θ is the parameter that makes the prediction result most likely.

[0096] Figure 2 A schematic diagram of the principle of predictive model identification provided in an embodiment of the present application.

[0097] Depend on Figure 2 It can be seen that the blizzard and power grid data are included in the prediction model, wherein the blizzard and power grid data are the meteorological data and geographic location data described in the embodiment of the present application. The first layer of the prediction model determines the prediction results through the decision tree algorithm, K-nearest neighbor algorithm and LGBM algorithm respectively, and the second layer of the prediction model outputs the final prediction results through the logistic regression algorithm.

[0098] Embodiment 3:

[0099] In order to accurately and effectively determine the impact on the residential electricity charge, based on the above embodiments, in the embodiment of the present application, the prediction model is trained in the following manner:

[0100] Obtain any meteorological data group pre-stored in a pre-stored training set, and obtain a value of the annotated impact on residential electricity charge stored for the meteorological data group;

[0101] Inputting the meteorological data group into the original prediction model before the prediction model is trained, and obtaining the output impact value of the original prediction model on the residential electricity charge;

[0102] The original prediction model is trained based on the labeled influence value and the output influence value.

[0103] In order to realize the training of the prediction model, a meteorological data set for training is saved in the embodiment of the present application, and the meteorological data group in the training set is a data group obtained from different disaster weather.

[0104] In order to facilitate the training of predictions, a training set is pre-stored locally in the electronic device. The training set also stores, for each meteorological data group, the impact value of the meteorological data group on the residential electricity charge. For the sake of distinction, the impact value can be called a labeled impact value. It should be noted that the labeled impact value can be determined by the staff based on the residential electricity charge corresponding to the meteorological data group and the residential electricity charge when no disasters such as blizzards occur. In a possible implementation, the labeled impact value can be determined based on the ratio of the residential electricity charge when no disasters such as blizzards occur to the residential electricity charge corresponding to the meteorological data group.

[0105] In an embodiment of the present application, after obtaining any meteorological data group in the training set and the annotated impact value of the meteorological data group on the residential electricity charge, the meteorological data group is input into the original prediction model, and the original prediction model outputs the impact value of the meteorological data group on the residential electricity charge. For the sake of convenience in distinction, the impact value can be called the output impact value.

[0106] After the original prediction model determines the output impact value corresponding to the meteorological data group, the original prediction model is trained according to the labeled prediction value and the output prediction value corresponding to the meteorological data group.

[0107] The prediction model is trained in the above manner, and when the preset conditions are met, a trained prediction model is obtained. The preset conditions may be that the number of meteorological data groups corresponding to the meteorological data groups in the training set whose labeled influence values ​​and output influence values ​​are consistent is greater than a set number; or the number of iterations for training the original prediction model reaches a set maximum number of iterations, etc. Specifically, the embodiments of the present application do not limit this.

[0108] Embodiment 4:

[0109] In order to accurately and effectively determine the impact on the residential electricity charge, based on the above embodiments, in the embodiment of the present application, the acquisition of meteorological data of at least one preset meteorological attribute includes:

[0110] Obtaining each pre-set meteorological attribute that is previously saved and affects residential electricity charge;

[0111] Acquire meteorological data of each preset meteorological attribute.

[0112] Since meteorological data of certain meteorological attributes may have little impact on the residential electricity charge, in order to accurately determine the impact of disasters such as blizzards on the residential electricity charge, the electronic device may pre-save each preset meteorological attribute that affects the residential electricity charge, and the electronic device may obtain each pre-saved meteorological attribute that affects the residential electricity charge, and obtain the meteorological data of each meteorological attribute whose corresponding meteorological attribute is any preset meteorological attribute obtained.

[0113] It should be noted that each meteorological attribute includes the minimum wind speed, maximum wind speed, average wind speed, snowfall intensity, air temperature, coverage, geographical location, and duration of the blizzard disaster.

[0114] Embodiment 5:

[0115] In order to determine each preset meteorological attribute, based on the above embodiments, in the embodiment of the present application, each preset meteorological attribute affecting the residential electricity charge is pre-saved in the following manner:

[0116] Get each weather attribute saved;

[0117] For each of the meteorological attributes, calculating the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge;

[0118] The meteorological attributes whose determined Pearson correlation coefficients are within a preset value range are determined as preset meteorological attributes that affect residential electricity charges.

[0119] In order to determine each preset meteorological attribute, the electronic device can obtain each saved meteorological attribute, and for each meteorological attribute, calculate the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge. In order to determine the preset meteorological attribute, the electronic device locally stores a preset numerical range, which can be [0.5,1]. After obtaining the Pearson correlation coefficient between the data of each meteorological attribute and the residential electricity charge, the electronic device can determine the meteorological data with the determined Pearson correlation coefficient in the preset numerical range as the preset meteorological attribute that affects the residential electricity charge.

[0120] In order to determine each preset meteorological attribute, based on the above embodiments, in the embodiment of the present application, the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge is calculated by the following formula:

[0121]

[0122] Where n is the number of samples, X i is the i-th data of the meteorological attribute, Y is the i-th charge data of the residential electricity load, is the average value of each data of the meteorological attribute, is the average value of the electricity charge used by each resident, and r is the Pearson correlation coefficient.

[0123] Among them, the i-th data of the meteorological attribute and the charge data of the i-th resident electricity charge are obtained in the same scenario. For example, if the i-th data of the meteorological attribute is the data at a certain moment, then the charge data of the i-th resident electricity charge is the charge data at that moment.

[0124] Embodiment 6:

[0125] In order to accurately and effectively determine the degree of influence on the residential electricity charge, based on the above embodiments, in the embodiment of the present application, the training set is constructed in the following manner:

[0126] Obtaining meteorological data of each preset meteorological attribute within a preset time period and each impact value on residential electricity charge;

[0127] For each preset meteorological attribute, determine sample meteorological data according to each meteorological data of the preset meteorological attribute; and determine a marked influence value according to each influence value within a preset time period;

[0128] A meteorological data group is determined according to the meteorological data of each preset meteorological attribute, and a corresponding relationship between the marked impact value and the meteorological data group is saved.

[0129] In an embodiment of the present application, the electronic device can obtain meteorological data for each preset meteorological attribute within a preset time period and each impact value on residents' electricity charge, wherein the preset time period can be a time period when disasters such as blizzards occur, or it can be a preset time period before disasters such as blizzards end, for example, it can be 30 days before the end of disasters such as blizzards, and the meteorological data for each preset meteorological attribute and each impact value can be daily meteorological data and the corresponding impact value determined for each day.

[0130] The electronic device can determine the sample meteorological data for each preset meteorological attribute according to each meteorological data of the preset meteorological attribute. For example, the median, mode, mean, maximum value or minimum value of each meteorological data can be determined as the sample meteorological data. In a possible implementation, the median, mode, mean, maximum value and minimum value of each meteorological data can also be determined as the sample meteorological data respectively. The electronic device can determine the labeled influence value according to each influence value within a preset time period. For example, the median, mode, mean, maximum value or minimum value of each influence value can be determined as the labeled influence value. In a possible implementation, the median, mode, mean, maximum value and minimum value of each influence value can also be determined as the labeled influence value respectively.

[0131] It should be noted that the method for determining the sample meteorological data is the same as the method for determining the labeled influence value. If the median of each meteorological data is determined as the sample meteorological data, the median of each influence value is also determined as the labeled influence value. If the median, mode, mean, maximum value and minimum value of each meteorological data are determined as the sample meteorological data, the method for determining the labeled influence value corresponding to each sample meteorological data is the same as the method for determining the sample meteorological data.

[0132] The electronic device determines a meteorological data group according to the meteorological data of each preset meteorological attribute, and saves the corresponding relationship between the marked impact value and the meteorological data group.

[0133] It should be noted that the electronic device can use 80% of the data in the blizzard meteorological attribute feature data set consisting of each acquired meteorological data group and the labeled impact value as a training set, and can divide the training set into five subsets D1, D2, D3, D4, and D5 of the same size, and use 20% of the data as a test set T.

[0134] For each base learner, a 5-fold cross-validation method is used to train the data set. One subset is selected from the five subsets as the test set, and the remaining four subsets are used as training sets until all five subsets are used as test sets. The optimal prediction result is used as the new feature to form a new training set.

[0135] For the single meta-learner in the second layer, the new training set is input, and the logistic regression algorithm is used for training. The test set T is predicted to obtain the optimal prediction model.

[0136] Figure 3 A schematic diagram of the principle of a 5-fold cross-validation provided in an embodiment of the present application.

[0137] Depend on Figure 3 It can be seen that the electronic device divides the training set into five subsets. The first subset is used as a verification set for the first time, and the remaining four subsets are used as training sets; the second subset is used as a verification set for the second time, and the remaining four subsets are used as training sets; and so on, until all five subsets are used as verification sets, and new feature sets and are obtained based on the prediction results obtained when each subset is used as a verification set.

[0138] Figure 4 A flowchart of a Stacking-based ensemble learning method.

[0139] Depend on Figure 4It can be seen that electronic equipment can construct and select features based on meteorological data and historical load data of disasters such as blizzards to obtain a feature data set, which is divided into a training set D and a test set T. The training set is used to train the base learner through 5-fold cross-validation, and the meta-learner is trained based on the prediction results to obtain a prediction model. The prediction model is then predicted through the test set to obtain the final result.

[0140] Compared with the prior art, the present invention has the following beneficial effects: This application combines the short-term and short-term blizzard forecast information of the meteorological department and the historical load information of the power grid, and fully considers the strongly related factors that affect the distribution of residential power load under blizzard weather: the geographical location of the power grid, the intensity of blizzard, the air temperature and the wind speed, and has certain effectiveness and rationality. The residential power load characteristic prediction method proposed in this application has certain guiding significance for the power grid company to accurately predict the residential power consumption status and allocate emergency materials in advance.

[0141] The feature data set is actually constructed by collecting and organizing historical blizzard disaster meteorological data and historical power grid load data; it is constructed by extracting and screening key data through feature engineering.

[0142] Moreover, changes in residential electricity load characteristics are closely related to the meteorological attributes of blizzard disasters and the geographical location information of the power grid. The collection and organization of historical blizzard disaster data, power grid operation data and related geographical location information data are specifically as follows: based on the statistical data of the meteorological department, the historical load data of the power grid and the minimum wind speed, maximum wind speed, average wind speed, snowfall intensity, air temperature, coverage range, geographical location, and duration of blizzard disasters are used to construct a training set for prediction under blizzard disasters, and this is used as data for feature construction and feature selection.

[0143] Embodiment 7:

[0144] Figure 5 A schematic diagram of a structure of an impact value determination device provided in an embodiment of the present application, the device comprising:

[0145] The acquisition module 501 is used to acquire meteorological data of at least one preset meteorological attribute if a prediction request is received;

[0146] The processing module 502 is used to input the acquired meteorological data into a pre-trained prediction model to obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

[0147] Furthermore, the processing module 502 is also used to obtain the type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge.

[0148] Furthermore, the processing module 502 is further configured to train the prediction model in the following manner:

[0149] Obtain any meteorological data group pre-stored in a pre-stored training set, and obtain a value of the annotated impact on residential electricity charge stored for the meteorological data group;

[0150] Inputting the meteorological data group into the original prediction model before the prediction model is trained, and obtaining the output impact value of the original prediction model on the residential electricity charge;

[0151] The original prediction model is trained based on the labeled influence value and the output influence value.

[0152] Furthermore, the processing module 502 is specifically used to obtain each pre-stored preset meteorological attribute that affects the residential electricity charge;

[0153] Acquire meteorological data of each preset meteorological attribute.

[0154] Furthermore, the processing module 502 is further configured to pre-save each preset meteorological attribute affecting the residential electricity charge in the following manner:

[0155] Get each weather attribute saved;

[0156] For each of the meteorological attributes, based on the data of the meteorological attribute and the residential electricity charge, calculate the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge;

[0157] The meteorological attributes whose determined Pearson correlation coefficients are within a preset value range are determined as preset meteorological attributes that affect residential electricity charges.

[0158] Furthermore, the processing module 502 is specifically used to calculate the Pearson correlation coefficient between the meteorological attribute data and the residential electricity charge by the following formula:

[0159]

[0160] Where n is the number of samples, X i is the i-th data of the meteorological attribute, Y is the i-th charge data of the residential electricity load, is the average value of each data of the meteorological attribute, is the average value of the electricity charge used by each resident, and r is the Pearson correlation coefficient.

[0161] Furthermore, the processing module 502 is further configured to construct the training set in the following manner:

[0162] Obtaining meteorological data of each preset meteorological attribute within a preset time period and each impact value on residential electricity charge;

[0163] For each preset meteorological attribute, determine sample meteorological data according to each meteorological data of the preset meteorological attribute; and determine a marked influence value according to each influence value within a preset time period;

[0164] A meteorological data group is determined according to the meteorological data of each preset meteorological attribute, and a corresponding relationship between the marked impact value and the meteorological data group is saved.

[0165] Embodiment 8:

[0166] Figure 6 The present invention provides a schematic diagram of an electronic device structure according to an embodiment of the present invention. Based on the above embodiments, the present invention further provides an electronic device, such as Figure 4 As shown, it includes: a processor 601, a communication interface 602, a memory 603 and a communication bus 604, wherein the processor 601, the communication interface 602, and the memory 603 communicate with each other through the communication bus 604;

[0167] The memory 603 stores a computer program. When the program is executed by the processor 601, the processor 601 performs the following steps:

[0168] If a forecast request is received, obtaining meteorological data of at least one preset meteorological attribute;

[0169] The acquired meteorological data is input into a pre-trained prediction model to obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

[0170] In a possible implementation, the method further includes:

[0171] The type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge is obtained.

[0172] In a possible implementation, the prediction model includes a first preset number of base learners and a meta-learner for combining base learners; wherein the first preset number of base learners respectively adopt decision tree, K-nearest neighbor, and lightweight gradient boosting machine LGBM algorithms.

[0173] In a possible implementation, the prediction model is trained in the following manner:

[0174] Obtain any meteorological data group pre-stored in a pre-stored training set, and obtain a value of the annotated impact on residential electricity charge stored for the meteorological data group;

[0175] Inputting the meteorological data group into the original prediction model before the prediction model is trained, and obtaining the output impact value of the original prediction model on the residential electricity charge;

[0176] The original prediction model is trained based on the labeled influence value and the output influence value.

[0177] In a possible implementation manner, the obtaining meteorological data of at least one preset meteorological attribute includes:

[0178] Obtaining each pre-set meteorological attribute that is previously saved and affects residential electricity charge;

[0179] Acquire meteorological data of each preset meteorological attribute.

[0180] In a possible implementation manner, each meteorological attribute includes: minimum wind speed, maximum wind speed, average wind speed, snowfall intensity, air temperature, coverage, geographical location, and duration of the snowstorm disaster.

[0181] In one possible implementation, each preset meteorological attribute affecting residential electricity charge is pre-stored in the following manner:

[0182] Get each weather attribute saved;

[0183] For each of the meteorological attributes, based on the data of the meteorological attribute and the residential electricity charge, calculate the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge;

[0184] The meteorological attributes whose determined Pearson correlation coefficients are within a preset value range are determined as preset meteorological attributes that affect residential electricity charges.

[0185] In a possible implementation, the Pearson correlation coefficient between the meteorological attribute data and the residential electricity charge is calculated by the following formula:

[0186]

[0187] Where n is the number of samples, X i is the i-th data of the meteorological attribute, Y is the i-th charge data of the residential electricity load, is the average value of each data of the meteorological attribute, is the average value of the electricity charge used by each resident, and r is the Pearson correlation coefficient.

[0188] In a possible implementation, the training set is constructed in the following manner:

[0189] Obtaining meteorological data of each preset meteorological attribute within a preset time period and each impact value on residential electricity charge;

[0190] For each preset meteorological attribute, determine sample meteorological data according to each meteorological data of the preset meteorological attribute; and determine a marked influence value according to each influence value within a preset time period;

[0191] A meteorological data group is determined according to the meteorological data of each preset meteorological attribute, and a corresponding relationship between the marked impact value and the meteorological data group is saved.

[0192] The communication bus mentioned in the above server can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0193] The communication interface is used for communication between the above electronic device and other devices.

[0194] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0195] The above-mentioned processor can be a general-purpose processor, including a central processing unit, a network processor (Network Processor, NP), etc.; it can also be a digital signal processing processor (Digital Signal Processing, DSP), an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc.

[0196] Embodiment 9:

[0197] On the basis of the above embodiments, an embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program executable by an electronic device, and when the program is run on the electronic device, the electronic device implements the following steps when executing:

[0198] If a forecast request is received, obtaining meteorological data of at least one preset meteorological attribute;

[0199] The acquired meteorological data is input into a pre-trained prediction model to obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

[0200] In a possible implementation, the method further includes:

[0201] The type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge is obtained.

[0202] In a possible implementation, the prediction model includes a first preset number of base learners and a meta-learner for combining base learners; wherein the first preset number of base learners respectively adopt decision tree, K-nearest neighbor, and lightweight gradient boosting machine LGBM algorithms.

[0203] In a possible implementation, the prediction model is trained in the following manner:

[0204] Obtain any meteorological data group pre-stored in a pre-stored training set, and obtain a value of the annotated impact on residential electricity charge stored for the meteorological data group;

[0205] Inputting the meteorological data group into the original prediction model before the prediction model is trained, and obtaining the output impact value of the original prediction model on the residential electricity charge;

[0206] The original prediction model is trained based on the labeled influence value and the output influence value.

[0207] In a possible implementation manner, the obtaining meteorological data of at least one preset meteorological attribute includes:

[0208] Obtaining each pre-set meteorological attribute that is previously saved and affects residential electricity charge;

[0209] Acquire meteorological data of each preset meteorological attribute.

[0210] In a possible implementation manner, each meteorological attribute includes: minimum wind speed, maximum wind speed, average wind speed, snowfall intensity, air temperature, coverage, geographical location, and duration of the snowstorm disaster.

[0211] In one possible implementation, each preset meteorological attribute affecting residential electricity charge is pre-stored in the following manner:

[0212] Get each weather attribute saved;

[0213] For each of the meteorological attributes, based on the data of the meteorological attribute and the residential electricity charge, calculate the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge;

[0214] The meteorological attributes whose determined Pearson correlation coefficients are within a preset value range are determined as preset meteorological attributes that affect residential electricity charges.

[0215] In a possible implementation, the Pearson correlation coefficient between the meteorological attribute data and the residential electricity charge is calculated by the following formula:

[0216]

[0217] Where n is the number of samples, X i is the i-th data of the meteorological attribute, Y is the i-th charge data of the residential electricity load, is the average value of each data of the meteorological attribute, is the average value of the electricity charge used by each resident, and r is the Pearson correlation coefficient.

[0218] In a possible implementation, the training set is constructed in the following manner:

[0219] Obtaining meteorological data of each preset meteorological attribute within a preset time period and each impact value on residential electricity charge;

[0220] For each preset meteorological attribute, determine sample meteorological data according to each meteorological data of the preset meteorological attribute; and determine a marked influence value according to each influence value within a preset time period;

[0221] A meteorological data group is determined according to the meteorological data of each preset meteorological attribute, and a corresponding relationship between the marked impact value and the meteorological data group is saved.

[0222] It is known from common technical knowledge that the present application can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present application or within the scope equivalent to the present application are included in the present application.

[0223] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0224] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0225] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0226] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

Claims

1. A method for determining an impact value, characterized in that: The method comprises: If a forecast request is received, obtaining meteorological data of at least one preset meteorological attribute; The acquired meteorological data is input into a pre-trained prediction model to obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

2. The method according to claim 1, characterized in that The method further comprises: The type of load change of the impact of the meteorological data output by the prediction model on the residential electricity charge is obtained.

3. The method according to claim 1, characterized in that The prediction model includes a first preset number of base learners and a meta-learner for combining base learners; wherein the first preset number of base learners respectively adopt decision tree, K-nearest neighbor, and lightweight gradient boosting machine LGBM algorithms.

4. The method according to claim 1, characterized in that: The prediction model is trained in the following way: Obtain any meteorological data group in a pre-saved training set, and obtain a labeled impact value on residential electricity charge saved for the meteorological data group; Inputting the meteorological data group into the original prediction model before the prediction model is trained, and obtaining the output impact value of the original prediction model on the residential electricity charge; The original prediction model is trained based on the labeled influence value and the output influence value.

5. The method according to claim 1, characterized in that The step of obtaining meteorological data of at least one preset meteorological attribute includes: Obtaining each pre-set meteorological attribute that is previously saved and affects residential electricity charge; Acquire meteorological data of each preset meteorological attribute.

6. The method according to claim 5, characterized in that Each meteorological attribute includes: minimum wind speed, maximum wind speed, average wind speed, snowfall intensity, air temperature, coverage, geographical location, and duration of the snowstorm disaster.

7. The method according to claim 5, characterized in that Each preset meteorological attribute that affects residential electricity charges is pre-saved in the following ways: Get each weather attribute saved; For each of the meteorological attributes, based on the data of the meteorological attribute and the residential electricity charge, calculate the Pearson correlation coefficient between the data of the meteorological attribute and the residential electricity charge; The meteorological attributes whose determined Pearson correlation coefficients are within a preset value range are determined as preset meteorological attributes that affect residential electricity charges.

8. The method according to claim 6, characterized in that The Pearson correlation coefficient between the meteorological attribute data and the residential electricity charge is calculated using the following formula: Where n is the number of samples, X i is the i-th data of the meteorological attribute, Y is the i-th charge data of the residential electricity load, is the average value of each data of the meteorological attribute, is the average value of the electricity charge used by each resident, and r is the Pearson correlation coefficient.

9. The method according to claim 7, characterized in that: The training set is constructed in the following way: Obtaining meteorological data of each preset meteorological attribute within a preset time period and each impact value on residential electricity charge; For each preset meteorological attribute, determining sample meteorological data according to each meteorological data of the preset meteorological attribute; and determining a marked impact value according to each impact value within a preset time period; A meteorological data group is determined according to the meteorological data of each preset meteorological attribute, and a corresponding relationship between the marked impact value and the meteorological data group is saved.

10. A product for determining an impact value, characterized in that: The products include: an acquisition module, for acquiring meteorological data of at least one preset meteorological attribute if a prediction request is received; A processing module is used to input the acquired meteorological data into a pre-trained prediction model to obtain the impact value of the meteorological data on the residential electricity charge output by the prediction model; wherein the prediction model is a model based on Stacking ensemble learning.

Images