CN111612275A - Method and device for predicting load of regional user - Google Patents

Method and device for predicting load of regional user Download PDF

Info

Publication number
CN111612275A
CN111612275A CN202010471759.3A CN202010471759A CN111612275A CN 111612275 A CN111612275 A CN 111612275A CN 202010471759 A CN202010471759 A CN 202010471759A CN 111612275 A CN111612275 A CN 111612275A
Authority
CN
China
Prior art keywords
user
predicted
data
load
electricity utilization
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202010471759.3A
Other languages
Chinese (zh)
Other versions
CN111612275B (en
Inventor
王凌谊
王志敏
顾洁
陈宇
李静涛
张秀钊
刘明伟
刘娟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Yunnan Power Grid Co Ltd
Original Assignee
Yunnan Power Grid Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Yunnan Power Grid Co Ltd filed Critical Yunnan Power Grid Co Ltd
Priority to CN202010471759.3A priority Critical patent/CN111612275B/en
Publication of CN111612275A publication Critical patent/CN111612275A/en
Application granted granted Critical
Publication of CN111612275B publication Critical patent/CN111612275B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply

Abstract

The application provides a method and a device for predicting regional user load. The method of the present application comprises: acquiring user data in an area to be predicted; determining the electricity utilization type of a user to be predicted from a plurality of electricity utilization types according to the user data in the area to be predicted; determining a target load prediction probability model according to the power utilization type of a user to be predicted and load prediction probability models corresponding to the power utilization types respectively; inputting weather data of a user to be predicted and date data of the user to be predicted into a target load prediction probability model to obtain an output result of the target load prediction probability model; determining the predicted electricity utilization load of the user to be predicted according to the output result; and determining the total predicted electricity load of the area to be predicted according to the predicted electricity load of all users to be predicted in the area to be predicted. According to the method and the device, different load prediction probability models are matched according to different power utilization types, and the accuracy of power utilization load prediction can be improved.

Description

Method and device for predicting load of regional user
Technical Field
The present application relates to the field of power technologies, and in particular, to a method and an apparatus for predicting a load of a regional user.
Background
With the continuous development of the power industry, power companies gradually change to the direction of comprehensive energy service, and a refined user power utilization management mode gradually replaces a rough user power utilization management mode.
At present, a refined user power consumption management mode is mainly realized by a load forecasting probability modeling method, and a power company forecasts the total amount of user power consumption loads in the whole area through the load forecasting probability modeling, and adjusts the power supply amount in the next time slot, so that the refined user power consumption management mode is realized. However, in the prior art, when the load prediction probability modeling is performed, all users in an area are built under the same model, when the data volume is large, the types of covered users are large, and the same model is used for predicting the power consumption of different types of users, so that the prediction accuracy of individual users is low easily, and further, the total power load of the whole area obtained through prediction has large fluctuation. For example, the total amount of users in a certain area is 10 thousands, a model is uniformly established in the prior art for the 10 thousands of users, however, the 10 thousands of users may relate to industrial users and residential users, the difference between the electricity consumption of the industrial users and the electricity consumption of the residential users is obviously large, if the same model is used for load prediction, the model is difficult to be well matched with the electricity consumption conditions of various types of users in the whole area, and further, the total amount of electricity consumption in the corresponding area predicted by an electric power company has a large deviation.
Based on this, there is a need for a method for predicting the user load in an area, which is used to solve the problem of low accuracy in predicting the total user load in a certain area in the prior art.
Disclosure of Invention
The application provides a method for predicting the user load in an area, which can be used for solving the problem that the accuracy of predicting the total user load in a certain area in the prior art is low.
In a first aspect, the present application provides a method for predicting an electric load capacity for a region, the method comprising:
acquiring user data in an area to be predicted; the user data comprises a typical electricity consumption mode of the user to be predicted, an electricity consumption behavior information entropy of the user to be predicted, an image characteristic of the user to be predicted, weather data of the user to be predicted and date data of the user to be predicted;
determining the electricity utilization type of the user to be predicted from a plurality of electricity utilization types according to the typical electricity utilization mode of the user to be predicted, the electricity utilization behavior information entropy of the user to be predicted and the portrait characteristics of the user to be predicted; the plurality of electricity utilization types are determined according to typical electricity utilization modes of sample users, the electricity utilization behavior information entropies of the sample users and the portrait characteristics of the sample users;
determining a target load prediction probability model according to the power utilization type of the user to be predicted and the load prediction probability models corresponding to the power utilization types respectively; the first load prediction probability model is established according to the weather data of the first sample user, the date data of the first sample user and the actual electricity utilization load data of the first sample user; the first load prediction probability model is a load prediction probability model corresponding to a first power utilization type; the first electricity utilization type is any one of the electricity utilization types; the first sample user is a sample user that conforms to the first power usage type;
inputting weather data of a user to be predicted and date data of the user to be predicted into the target load prediction probability model to obtain an output result of the target load prediction probability model;
determining the predicted electricity utilization load quantity of the user to be predicted according to the output result;
and determining the total predicted electricity load of the area to be predicted according to the predicted electricity load of all users to be predicted in the area to be predicted.
With reference to the first aspect, in an implementation manner of the first aspect, the load prediction probability model corresponding to the first power consumption type is specifically established in the following manner:
dividing the user data of the first sample user into a first training data set, a second training data set and a test data set; the first training data set comprises weather data of a first training sample user, date data of the first training sample user and actual electricity load data of the first training sample user; the second training data set comprises weather data of a second training sample user, date data of the second training sample user and actual electricity load data of the second training sample user; the test data set comprises weather data of a test sample user, date data of the test sample user and actual electricity utilization load data of the test sample user;
training a preset point prediction model according to the weather data of the first training sample user, the date data of the first training sample user and the actual power consumption load data of the first training sample user to obtain a trained point prediction model;
inputting the weather data of the second training sample user and the date data of the second training sample user into the trained point prediction model to obtain predicted electricity utilization load data of the second training sample user;
taking the difference value of the actual electricity utilization load data of the second training sample user and the predicted electricity utilization load data of the second training sample user as the residual error of the second training sample user;
training a preset condition residual error prediction model according to the weather data of the second training sample user, the date data of the second training sample user, the predicted power consumption load data of the second training sample user and the residual error of the second training sample user to obtain a trained condition residual error prediction model;
inputting the weather data of the test sample user and the date data of the test sample user into the trained point prediction model to obtain the predicted electricity utilization load data of the test sample user;
inputting the weather data of the test sample user, the date data of the test sample user and the predicted power utilization load data of the test sample user into the trained conditional residual prediction model to obtain a residual of the test sample user;
taking a difference value corresponding to the predicted electricity utilization load quantity data of the test sample user and the residual error of the test sample user as the predicted electricity utilization load quantity of the test sample user;
and if the difference value between the predicted electricity utilization load of the test sample user and the actual electricity utilization load data of the test sample user is smaller than a preset threshold value, determining the trained point prediction model and the trained conditional residual prediction model as a load prediction probability model corresponding to the first electricity utilization type.
With reference to the first aspect, in an implementation manner of the first aspect, the plurality of power usage types are specifically determined by:
randomly dividing the sample user data into a plurality of initial data sets;
according to each initial data set, local clustering is carried out on sample users by using a local clustering model according to typical electricity utilization modes of the sample users in the initial data sets, electricity utilization behavior information entropies of the sample users and portrait characteristics of the sample users, and a plurality of sub-electricity utilization types are obtained;
and utilizing a global clustering model to perform global clustering on the sub-electricity utilization types respectively corresponding to the plurality of initial data sets to obtain the plurality of electricity utilization types.
With reference to the first aspect, in an implementation manner of the first aspect, the typical power usage pattern of the user to be predicted is specifically determined by:
acquiring various historical electricity utilization modes of a user to be predicted and the occurrence times of each historical electricity utilization mode;
determining the proportion of each historical electricity utilization mode according to the occurrence frequency of each historical electricity utilization mode and the total occurrence frequency of all historical electricity utilization modes;
and determining the historical power utilization pattern with the highest percentage as the typical power utilization pattern of the user to be predicted.
With reference to the first aspect, in an implementation manner of the first aspect, the image feature of the user to be predicted is specifically determined by:
collecting questionnaire information of a user to be predicted; the questionnaire information comprises at least one of a user basic attribute, a user electricity utilization habit attribute, a user house attribute and a user household equipment configuration attribute;
discretizing, normalizing and coding the questionnaire information of the user to be predicted to obtain the image characteristics of the user to be predicted.
In a second aspect, the present application provides an apparatus for predicting an electricity load amount for a region, the apparatus comprising:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring user data in an area to be predicted; the user data comprises a typical electricity consumption mode of the user to be predicted, an electricity consumption behavior information entropy of the user to be predicted, an image characteristic of the user to be predicted, weather data of the user to be predicted and date data of the user to be predicted;
the processing module is used for determining the electricity utilization type of the user to be predicted from a plurality of electricity utilization types according to the typical electricity utilization mode of the user to be predicted, the electricity utilization behavior information entropy of the user to be predicted and the portrait characteristics of the user to be predicted; the plurality of electricity utilization types are determined according to typical electricity utilization modes of sample users, the electricity utilization behavior information entropies of the sample users and the portrait characteristics of the sample users;
the processing module is further used for determining a target load prediction probability model according to the power utilization type of the user to be predicted and the load prediction probability models corresponding to the power utilization types respectively; the first load prediction probability model is established according to the weather data of the first sample user, the date data of the first sample user and the actual electricity utilization load data of the first sample user; the first load prediction probability model is a load prediction probability model corresponding to a first power utilization type; the first electricity utilization type is any one of the electricity utilization types; the first sample user is a sample user that conforms to the first power usage type;
the forecasting module is used for inputting the weather data of the user to be forecasted and the date data of the user to be forecasted into the target load forecasting probability model to obtain an output result of the target load forecasting probability model; the power consumption prediction method comprises the steps of outputting an output result to a user to be predicted; and the total predicted electricity load quantity of the area to be predicted is determined according to the predicted electricity load quantities of all users to be predicted in the area to be predicted.
With reference to the second aspect, in an implementation manner of the second aspect, the load prediction probability model corresponding to the first power usage type is specifically established in the following manner:
dividing the user data of the first sample user into a first training data set, a second training data set and a test data set; the first training data set comprises weather data of a first training sample user, date data of the first training sample user and actual electricity load data of the first training sample user; the second training data set comprises weather data of a second training sample user, date data of the second training sample user and actual electricity load data of the second training sample user; the test data set comprises weather data of a test sample user, date data of the test sample user and actual electricity utilization load data of the test sample user;
training a preset point prediction model according to the weather data of the first training sample user, the date data of the first training sample user and the actual power consumption load data of the first training sample user to obtain a trained point prediction model;
inputting the weather data of the second training sample user and the date data of the second training sample user into the trained point prediction model to obtain predicted electricity utilization load data of the second training sample user;
taking the difference value of the actual electricity utilization load data of the second training sample user and the predicted electricity utilization load data of the second training sample user as the residual error of the second training sample user;
training a preset condition residual error prediction model according to the weather data of the second training sample user, the date data of the second training sample user, the predicted power consumption load data of the second training sample user and the residual error of the second training sample user to obtain a trained condition residual error prediction model;
inputting the weather data of the test sample user and the date data of the test sample user into the trained point prediction model to obtain the predicted electricity utilization load data of the test sample user;
inputting the weather data of the test sample user, the date data of the test sample user and the predicted power utilization load data of the test sample user into the trained conditional residual prediction model to obtain a residual of the test sample user;
taking a difference value corresponding to the predicted electricity utilization load quantity data of the test sample user and the residual error of the test sample user as the predicted electricity utilization load quantity of the test sample user;
and if the difference value between the predicted electricity utilization load of the test sample user and the actual electricity utilization load data of the test sample user is smaller than a preset threshold value, determining the trained point prediction model and the trained conditional residual prediction model as a load prediction probability model corresponding to the first electricity utilization type.
With reference to the second aspect, in an implementable manner of the second aspect, the plurality of power usage types is specifically determined by:
randomly dividing the sample user data into a plurality of initial data sets;
according to each initial data set, local clustering is carried out on sample users by using a local clustering model according to typical electricity utilization modes of the sample users in the initial data sets, electricity utilization behavior information entropies of the sample users and portrait characteristics of the sample users, and a plurality of sub-electricity utilization types are obtained;
and utilizing a global clustering model to perform global clustering on the sub-electricity utilization types respectively corresponding to the plurality of initial data sets to obtain the plurality of electricity utilization types.
With reference to the second aspect, in an implementation manner of the second aspect, the typical power usage pattern of the user to be predicted is specifically determined by:
acquiring various historical electricity utilization modes of a user to be predicted and the occurrence times of each historical electricity utilization mode;
determining the proportion of each historical electricity utilization mode according to the occurrence frequency of each historical electricity utilization mode and the total occurrence frequency of all historical electricity utilization modes;
and determining the historical power utilization pattern with the highest percentage as the typical power utilization pattern of the user to be predicted.
With reference to the second aspect, in an implementation manner of the second aspect, the image feature of the user to be predicted is specifically determined by:
collecting questionnaire information of a user to be predicted; the questionnaire information comprises at least one of a user basic attribute, a user electricity utilization habit attribute, a user house attribute and a user household equipment configuration attribute;
discretizing, normalizing and coding the questionnaire information of the user to be predicted to obtain the image characteristics of the user to be predicted.
The application groups different user data. And modeling is carried out aiming at the user data of the same type in each group, so that the established model can be well matched with each user data, and the precision of predicting the user load in the region is improved. The grouping approach also avoids the huge workload that may be incurred to model each user. The method provided by the application can improve the precision of predicting the user load in the region on the premise of avoiding a large amount of workload.
Drawings
Fig. 1 is a schematic flowchart corresponding to a method for predicting a load of a regional user according to an embodiment of the present disclosure;
fig. 2a is a schematic diagram of an exemplary power consumption mode of a user to be predicted according to an embodiment of the present application;
fig. 2b is a second schematic diagram of a typical power consumption mode of a user to be predicted according to an embodiment of the present application;
fig. 2c is a third schematic diagram of a typical power consumption mode of a user to be predicted according to an embodiment of the present application;
fig. 3 is a schematic flowchart of determining a plurality of power usage types according to an embodiment of the present application;
fig. 4 is a schematic diagram of a result after global clustering according to an embodiment of the present application;
fig. 5 is a schematic flowchart illustrating a process of establishing a load prediction probability model corresponding to a first power consumption type according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of an apparatus for predicting a load of a regional user according to an embodiment of the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
In order to realize fine management of power consumption of users, an electric power company usually estimates the total load in an area in the next time period by using an established load prediction probability model, so that the generated energy is adjusted, and the waste of energy is avoided. At present, in the prior art, all users in an area are generally established under the same model, but even in the same area, the electricity utilization characteristics of the users are different, and a single model is difficult to match with each user.
Based on the above problems, the embodiments of the present application provide a method for predicting a load amount of a regional user. Referring to fig. 1, a schematic flow chart corresponding to a method for predicting a load of a regional user according to an embodiment of the present application is exemplarily shown, where the specific flow chart is as follows:
step 101, obtaining user data in an area to be predicted.
And 102, determining the electricity utilization type of the user to be predicted from the plurality of electricity utilization types according to the typical electricity utilization mode of the user to be predicted, the electricity utilization behavior information entropy of the user to be predicted and the portrait characteristics of the user to be predicted.
And 103, determining a target load prediction probability model according to the power utilization types of the users to be predicted and the load prediction probability models corresponding to the power utilization types respectively.
And 104, inputting the weather data of the user to be predicted and the date data of the user to be predicted into the target load prediction probability model to obtain an output result of the target load prediction probability model.
And 105, determining the predicted electricity utilization load of the user to be predicted according to the output result.
And step 106, determining the total predicted electricity load of the area to be predicted according to the predicted electricity load of all users to be predicted in the area to be predicted.
The embodiment of the application groups different user data. And modeling is carried out aiming at the user data of the same type in each group, so that the established model can be well matched with each user data, and the precision of predicting the user load in the region is improved. The grouping approach also avoids the huge workload that may be incurred to model each user. The method provided by the embodiment of the application can improve the precision of the load of the user in the prediction region on the premise of avoiding a large amount of workload.
Specifically, in step 101, the user data includes a typical electricity consumption mode of the user to be predicted, an electricity consumption behavior information entropy of the user to be predicted, an image characteristic of the user to be predicted, weather data of the user to be predicted, and date data of the user to be predicted.
The typical power consumption mode of the user to be predicted, the power consumption behavior information entropy of the user to be predicted and the portrait characteristics of the user to be predicted are combined to form a feature vector of the user to be predicted.
The determination of the portions of the feature vector is described separately below.
(1) Typical power usage patterns of a user to be predicted
The typical power consumption pattern of the user to be predicted is determined by the following specific means:
firstly, a plurality of historical electricity utilization patterns of a user to be predicted and the occurrence times of each historical electricity utilization pattern are obtained.
Then, the proportion of each historical electricity usage pattern is determined based on the number of occurrences of each historical electricity usage pattern and the total number of occurrences of all historical electricity usage patterns.
And finally, determining the historical power utilization pattern with the highest percentage as a typical power utilization pattern of the user to be predicted.
Fig. 2a is a schematic diagram of a typical power consumption mode of a user to be predicted according to an embodiment of the present application.
Fig. 2b is a schematic diagram of a second exemplary power consumption mode of the user to be predicted according to the embodiment of the present application.
Fig. 2c is a schematic diagram of a third exemplary power consumption mode of the user to be predicted according to the embodiment of the present application.
It should be noted that, in the implementation process of the present application, the typical power consumption patterns of the user to be predicted are various, and the three figures are only an example of three of them. From these three graphs, it can be known that the user's load generally exhibits a single peak pattern. Generally, electricity consumption is in a low-valley stage in the morning. These rules are only general rules of typical electricity usage patterns of the users to be predicted, and electricity usage characteristics of the users are different, for example, peak times in the typical electricity usage patterns of different users to be predicted are different.
(2) Image characteristics of a user to be predicted
The image characteristics of the user to be predicted are determined in the following way:
collecting questionnaire information of a user to be predicted, and carrying out discretization, normalization and coding on the questionnaire information of the user to be predicted to obtain image characteristics of the user to be predicted.
Specifically, the questionnaire information may include a variety of information, and for example, may include at least one of a user basic attribute, a user electricity usage habit attribute, a user house attribute, and a user home device configuration attribute. Wherein the user basic attribute may include at least one of: the age of the householder, the employment situation of the householder, the social class of the householder, etc.
It should be noted that the questionnaire information may not be limited to the four pieces of information provided in the embodiment of the present application, and information related to the power consumption condition of the user may be entered into the questionnaire information.
(3) Power consumption behavior information entropy of user to be predicted
To determine the power consumption behavior information entropy of a user to be predicted, the probability that the user to be predicted is in a specific power consumption mode type is determined by the following method:
Figure BDA0002514500560000061
in the formula (1), the first and second groups,
Figure BDA0002514500560000062
the probability that the power utilization mode type with the number of i appears in the user to be predicted with the number of n, namely the probability that the user to be predicted is in a specific power utilization mode type; n is the number of the user to be predicted; i is the number of the power consumption mode type of the user to be predicted; m represents the number of the historical electricity utilization modes of the user to be predicted; m represents the number of the historical electricity utilization mode of the user to be predicted; i is an indicative function, when a predetermined condition is satisfied (sm),nI), namely the mth historical electricity utilization mode of the user to be predicted, which is numbered n, is just the electricity utilization mode corresponding to the electricity utilization mode type, which is numbered i, the value is 1, otherwise, the value is 0;
with the formula (1), the power consumption behavior information entropy of the user to be predicted is determined specifically by the following method:
Figure BDA0002514500560000063
in the formula (2), E(n)The power consumption behavior information entropy of the user to be predicted is numbered n;
Figure BDA0002514500560000064
the probability that the power utilization mode type with the number of i appears in the user to be predicted with the number of n; s is the number of power mode types. If when i is an arbitrary value, then,
Figure BDA0002514500560000065
the values are all 1/S, the electricity utilization behaviors of the user to be predicted are variable at the moment, and the corresponding E(n)Maximum; if when i is a certain value,
Figure BDA0002514500560000066
if the value is 1, the electricity utilization behavior of the user to be predicted is most stable at the moment, and the corresponding E(n)And minimum, the value is 0.
After determining a typical electricity consumption mode of a user to be predicted, an electricity consumption behavior information entropy of the user to be predicted and an portrait feature of the user to be predicted, combining the three to obtain a feature vector of the user to be predicted.
In step 102, a plurality of electricity utilization types are determined according to typical electricity utilization patterns of sample users, the electricity utilization behavior information entropies of the sample users and the portrait characteristics of the sample users.
Specifically, as shown in fig. 3, a schematic flow chart for determining multiple power consumption types provided in the embodiment of the present application specifically includes the following steps:
in step 301, sample user data is randomly divided into a plurality of initial data sets.
In particular, each initial data set may contain sample users of one or more electricity usage types.
And 302, carrying out local clustering on the sample users by using a local clustering model according to the typical electricity utilization mode of the sample users in the initial data set, the electricity utilization behavior information entropy of the sample users and the portrait characteristics of the sample users to obtain a plurality of sub-electricity utilization types for each initial data set.
Various local clustering methods can be used here, and one possible method is the K-means clustering algorithm.
And 303, carrying out global clustering on the sub-electricity utilization types respectively corresponding to the plurality of initial data sets by using a global clustering model to obtain a plurality of electricity utilization types.
Fig. 4 is a schematic diagram of a result after global clustering according to an embodiment of the present application. As can be seen from the figure, the sample users of the same electricity consumption type are basically gathered together, and the sample users of different electricity consumption types belong to different electricity consumption types.
As shown in table 1, is an example of partial data of a sample user under different electricity usage types. The number of sample users with electricity type number 1 is 1752, the corresponding entropy mean value of user behavior information is 2.821, the corresponding typical electricity consumption mode mean value is 1.031, the corresponding average value of the owner age group (one portrait characteristic) is 4.531, and the corresponding average value of the owner employment situation (the second portrait characteristic) is 3.513. Specifically, reference may be made to the contents shown in table 1, which are not described in detail here.
Table 1: example of partial data of sample users under different electricity usage types
Figure BDA0002514500560000071
As can be seen from table 1, the number of users of different electricity usage types is greatly different, and the data of corresponding sample users is also greatly different. If all data are predicted by using the same model, the accuracy of the predicted data is necessarily low.
In step 103, the load prediction probability models corresponding to the electricity usage types are obtained by training before prediction.
Taking the first load prediction probability model as an example, the first load prediction probability model is established according to the weather data of the first sample user, the date data of the first sample user and the actual electricity load data of the first sample user. The first load prediction probability model is a load prediction probability model corresponding to the first power usage type. The first electricity usage type is any one of the electricity usage types; the first sample user is a sample user that conforms to the first power usage type.
It should be noted that "first" is merely numbered herein to distinguish types, and similar contents may be denoted by "second", "third", and the like.
Specifically, referring to fig. 5, a schematic flow chart illustrating establishment of a load prediction probability model corresponding to a first power consumption type provided in the embodiment of the present application is exemplarily shown, and the method specifically includes the following steps:
step 501, divide the user data of the first sample user into a first training data set, a second training data set, and a test data set.
Wherein the number of first sample users in the test data set is approximately twenty percent of the total number of first sample users.
Specifically, the first training data set comprises weather data of a first training sample user, date data of the first training sample user and actual electricity load data of the first training sample user; the second training data set comprises weather data of a second training sample user, date data of the second training sample user and actual electricity load data of the second training sample user; the test data set includes weather data for the test specimen user, date data for the test specimen user, and actual electricity load data for the test specimen user.
Step 502, training a preset point prediction model according to weather data of the first training sample user, date data of the first training sample user and actual power consumption load data of the first training sample user to obtain a trained point prediction model.
Specifically, a preset point prediction model is trained to obtain parameters of the point prediction model, and the parameters of the point prediction model can be optimally fitted with actual electricity utilization load data of a first training sample user. The trained point prediction model can be determined according to the parameters of the point prediction model.
Step 503, inputting the weather data of the second training sample user and the date data of the second training sample user into the trained point prediction model to obtain the predicted electricity consumption load data of the second training sample user.
Specifically, the predicted electricity load data of the second training sample user is determined by:
Figure BDA0002514500560000081
in the formula (3), the first and second groups,
Figure BDA0002514500560000082
in step 503, the predicted electricity load data is represented as predicted electricity load data of the second training sample user; f is a point prediction model; w is a parameter of the point prediction model; xtThe input vectors corresponding to the weather data and the date data are embodied as the weather data of the second training sample user and the date data of the second training sample user in step 503.
Step 504, a difference value between the actual electricity load data of the second training sample user and the predicted electricity load data of the second training sample user is used as a residual error of the second training sample user.
And 505, training a preset condition residual error prediction model according to the weather data of the second training sample user, the date data of the second training sample user, the predicted electricity load data of the second training sample user and the residual error of the second training sample user to obtain a trained condition residual error prediction model.
Specifically, when the preset conditional residual prediction model is trained, the quantile regression model is used instead of the normal distribution model and other models commonly used in the prior art. The quantile regression model can more accurately represent the expression capacity of the residual errors.
Step 506, inputting the weather data of the test sample user and the date data of the test sample user into the trained point prediction model to obtain the predicted electricity utilization load data of the test sample user.
Specifically, the predicted electricity load data of the test sample user can be determined according to equation (3).
And step 507, inputting the weather data of the test sample user, the date data of the test sample user and the predicted power utilization load data of the test sample user into the trained conditional residual error prediction model to obtain the residual error of the test sample user.
Specifically, the residual error of the test sample user is determined by:
Figure BDA0002514500560000083
in the formula (4), the first and second groups,t,qin order to predict the residual error, the residual error is embodied as the residual error of the user of the test sample in step 507; gqA quantile regression model (namely a conditional residual prediction model); wqParameters of a quantile regression model; xtThe input vectors corresponding to the weather data and the date data are embodied as the weather data of the test sample user and the date data of the test sample user in step 507;
Figure BDA0002514500560000084
to predict the electricity load amount data, the predicted electricity load amount data of the test sample user is embodied in step 507.
And step 508, taking the difference value corresponding to the residual error of the test sample user and the predicted electricity load data of the test sample user as the predicted electricity load of the test sample user.
Specifically, the predicted electricity load of the test sample user is determined by the following method:
Figure BDA0002514500560000091
in the formula (5), the first and second groups,
Figure BDA0002514500560000092
for final prediction of the electrical load, stepsEmbodied in step 508 as testing the predicted electricity load capacity of the sample user;
Figure BDA0002514500560000093
in order to predict the electricity usage load amount data, step 508 is embodied as the predicted electricity usage load amount data of the test sample user;t,qto predict the residual, step 508 is embodied as testing the residual of the sample user.
In step 509, if the difference between the predicted electricity load of the test sample user and the actual electricity load data of the test sample user is smaller than the preset threshold, the trained point prediction model and the trained conditional residual prediction model are determined as the load prediction probability model corresponding to the first electricity type.
If the difference value between the predicted electricity load of the test sample user and the actual electricity load data of the test sample user is larger than or equal to the preset threshold value, the trained point prediction model and the trained condition residual prediction model need to be adjusted until the difference value between the predicted electricity load of the test sample user and the actual electricity load data of the test sample user is smaller than the preset threshold value.
Through steps 501 to 509, a plurality of load prediction probability models may be obtained, where each load prediction probability model corresponds to one electricity usage type. And inputting the power utilization type of the user to be predicted, matching to obtain a target load prediction probability model, wherein the power utilization type corresponding to the target load prediction probability model is closest to the power utilization type of the user to be predicted.
In steps 104 to 106, as described above, in the embodiment of the present application, the target load prediction probability model may include a trained point prediction model and a trained conditional residual prediction model.
Based on the method, when the electricity load is predicted, the weather data and the date data of the user to be predicted can be input into the trained point prediction model to obtain the output result of the point prediction model; then inputting the weather data of the user to be predicted, the date data of the user to be predicted and the output result of the point prediction model into the trained conditional residual prediction model to obtain the output result of the conditional residual prediction model; and finally, determining the output result of the target load prediction probability model according to the difference value between the output result of the point prediction model and the output result of the conditional residual prediction model.
Further, the output result of the target load prediction probability model may be used as the predicted electricity load amount of the user to be predicted.
After the electricity utilization load quantities of all the users to be predicted are predicted, the predicted electricity utilization load quantities of all the users to be predicted in the same type can be added to obtain the aggregated total load quantity formed by the users to be predicted in the same type; and then adding the total aggregated load quantities of different types to obtain the total predicted electricity load quantity of the area to be predicted.
It should be noted that, in step 505, when the preset conditional residual prediction model is trained, a quantile regression model is used, so the output result of the conditional residual prediction model is also an expression form of q quantiles, which results in that the output result of the target load prediction probability model is also an expression form of q quantiles.
It should be noted that, since the output result of the target load prediction probability model is an expression form of q quantiles, an appropriate prediction electric load amount may be selected as needed, that is, the load amount at the position with the highest probability may be used as the final prediction electric load amount, the load amount at the position with the probability of 80% may be used as the final prediction electric load amount, and a load section may be selected as needed as the final prediction electric load amount. Different options may provide different adjustment schemes for the utility company. For example, the upper limit of the required power generation amount of the electric power company can be determined by selecting the load amount at the position with the maximum probability as the last predicted power consumption load amount. And for example, selecting a section of load interval to estimate the interval of the power generation amount required by the power company.
According to the steps 101 to 106, the total predicted electricity load of the area to be predicted can be obtained according to the method embodiment of the application. In order to verify the effect of the embodiment of the method of the present application, two models are used in the following for comparison with the embodiment of the present application.
The comparison model 1 uses the same model to predict all users to be predicted.
The comparison model 2 uses a method of predicting a model that is trained differently for each user.
The Mean Absolute Percent Error (MAPE) and the Quantile Score (QS) are common indicators of the prediction accuracy, and lower values of MAPE and QS represent higher prediction accuracy.
As shown in table 2, the accuracy is predicted for the individual user load for the three models. After the model used in the embodiment of the present application is used to predict the user 1, the obtained MAPE value is 4.91%, and the obtained QS value is 0.146; after prediction of user 1 using comparative model 1, the obtained MAPE value was 5.65% and the obtained QS value was 0.167; after prediction of user 1 using comparative model 2, the obtained MAPE value was 5.74% and the obtained QS value was 0.179. Specifically, reference may be made to the content shown in table 2, which is not described in detail here.
Table 2: single user load prediction accuracy of three models
User 1 User 2 User 3
MAPE values for the models used in the examples of this application 4.91％ 3.95％ 4.23％
QS value of the model used in the examples of the present application 0.146 0.104 0.124
MAPE values for comparative model 1 5.65％ 6.15％ 5.30％
QS value of comparative model 1 0.167 0.189 0.156
MAPE values for comparative model 2 5.74％ 4.92％ 5.28％
QS value of comparative model 2 0.179 0.145 0.161
As can be seen from table 2, when the model provided in the embodiment of the present application is used to predict a single user, the prediction accuracy is higher than the prediction accuracy of the comparison model 1 or the comparison model 2, so that it can be known that the target load prediction probability model provided in the embodiment of the present application can ensure the prediction accuracy.
The following are embodiments of the apparatus of the present application that may be used to perform embodiments of the method of the present application. For details which are not disclosed in the embodiments of the apparatus of the present application, reference is made to the embodiments of the method of the present application.
Fig. 6 is a schematic structural diagram illustrating an apparatus for predicting a load of a regional user according to an embodiment of the present application. As shown in fig. 6, the apparatus has a function of implementing the method for predicting the load of the regional user, and the function may be implemented by hardware, or may be implemented by hardware executing corresponding software. The apparatus may include: an acquisition module 601, a processing module 602, and a prediction module 603.
An obtaining module 601, configured to obtain user data in an area to be predicted; the user data comprises typical electricity utilization modes of the users to be predicted, electricity utilization behavior information entropies of the users to be predicted, portrait characteristics of the users to be predicted, weather data of the users to be predicted and date data of the users to be predicted.
The processing module 602 is configured to determine the electricity utilization type of the user to be predicted from the multiple electricity utilization types according to the typical electricity utilization mode of the user to be predicted, the electricity utilization behavior information entropy of the user to be predicted, and the portrait characteristics of the user to be predicted; the plurality of electricity utilization types are determined according to typical electricity utilization patterns of sample users, the electricity utilization behavior information entropies of the sample users and the portrait characteristics of the sample users.
The processing module 602 is further configured to determine a target load prediction probability model according to the power utilization types of the users to be predicted and load prediction probability models corresponding to the power utilization types respectively; the first load prediction probability model is established according to the weather data of the first sample user, the date data of the first sample user and the actual electricity utilization load data of the first sample user; the first load prediction probability model is a load prediction probability model corresponding to the first power utilization type; the first electricity usage type is any one of the electricity usage types; the first sample user is a sample user that conforms to the first power usage type.
The prediction module 603 is configured to input the weather data of the user to be predicted and the date data of the user to be predicted into the target load prediction probability model, so as to obtain an output result of the target load prediction probability model; the power consumption prediction method comprises the steps of outputting an output result to a user to be predicted; and the total predicted electricity load amount of the area to be predicted is determined according to the predicted electricity load amounts of all users to be predicted in the area to be predicted.
Optionally, the load prediction probability model corresponding to the first power consumption type is specifically established in the following manner:
dividing user data of a first sample user into a first training data set, a second training data set and a test data set; the first training data set comprises weather data of a first training sample user, date data of the first training sample user and actual electricity load data of the first training sample user; the second training data set comprises weather data of a second training sample user, date data of the second training sample user and actual electricity load data of the second training sample user; the test data set includes weather data for the test specimen user, date data for the test specimen user, and actual electricity load data for the test specimen user.
And training a preset point prediction model according to the weather data of the first training sample user, the date data of the first training sample user and the actual power consumption load data of the first training sample user to obtain the trained point prediction model.
And inputting the weather data of the second training sample user and the date data of the second training sample user into the trained point prediction model to obtain the predicted electricity utilization load data of the second training sample user.
And taking the difference value of the actual electricity utilization load data of the second training sample user and the predicted electricity utilization load data of the second training sample user as the residual error of the second training sample user.
And training a preset condition residual error prediction model according to the weather data of the second training sample user, the date data of the second training sample user, the predicted electricity load data of the second training sample user and the residual error of the second training sample user to obtain a trained condition residual error prediction model.
And inputting the weather data of the test sample user and the date data of the test sample user into the trained point prediction model to obtain the predicted electricity utilization load data of the test sample user.
Inputting the weather data of the test sample user, the date data of the test sample user and the predicted power utilization load data of the test sample user into the trained condition residual error prediction model to obtain the residual error of the test sample user.
And taking the difference value corresponding to the residual error of the test sample user and the predicted electricity load data of the test sample user as the predicted electricity load of the test sample user.
And if the difference value between the predicted electricity utilization load of the test sample user and the actual electricity utilization load data of the test sample user is smaller than a preset threshold value, determining the trained point prediction model and the trained conditional residual prediction model as a load prediction probability model corresponding to the first electricity utilization type.
Optionally, the plurality of power usage types are specifically determined by:
randomly dividing sample user data into a plurality of initial data sets;
and for each initial data set, carrying out local clustering on the sample users by using a local clustering model according to the typical electricity utilization modes of the sample users in the initial data sets, the electricity utilization behavior information entropies of the sample users and the portrait characteristics of the sample users to obtain a plurality of sub-electricity utilization types.
And carrying out global clustering on the sub-electricity utilization types respectively corresponding to the plurality of initial data sets by using a global clustering model to obtain a plurality of electricity utilization types.
Optionally, the typical power consumption mode of the user to be predicted is specifically determined by the following method:
the method comprises the steps of obtaining multiple historical electricity utilization modes of a user to be predicted and the occurrence times of each historical electricity utilization mode.
And determining the proportion of each historical electricity utilization mode according to the occurrence number of each historical electricity utilization mode and the total occurrence number of all historical electricity utilization modes.
And determining the historical power utilization pattern with the highest percentage as the typical power utilization pattern of the user to be predicted.
Optionally, the image feature of the user to be predicted is specifically determined by the following method:
collecting questionnaire information of a user to be predicted; the questionnaire information comprises at least one of a user basic attribute, a user electricity usage habit attribute, a user house attribute and a user home device configuration attribute.
Discretizing, normalizing and coding the questionnaire information of the user to be predicted to obtain the image characteristics of the user to be predicted.
The embodiment of the application groups different user data. And modeling is carried out aiming at the user data of the same type in each group, so that the established model can be well matched with each user data, and the precision of predicting the user load in the region is improved. The grouping approach also avoids the huge workload that may be incurred to model each user. The method provided by the embodiment of the application can improve the precision of the load of the user in the prediction region on the premise of avoiding a large amount of workload.
The invention is operational with numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
The invention may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.
It will be understood that the invention is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the invention is limited only by the appended claims.

Claims (10)

1. A method of predicting regional user load capacity, the method comprising:
acquiring user data in an area to be predicted; the user data comprises a typical electricity consumption mode of the user to be predicted, an electricity consumption behavior information entropy of the user to be predicted, an image characteristic of the user to be predicted, weather data of the user to be predicted and date data of the user to be predicted;
determining the electricity utilization type of the user to be predicted from a plurality of electricity utilization types according to the typical electricity utilization mode of the user to be predicted, the electricity utilization behavior information entropy of the user to be predicted and the portrait characteristics of the user to be predicted; the plurality of electricity utilization types are determined according to typical electricity utilization modes of sample users, the electricity utilization behavior information entropies of the sample users and the portrait characteristics of the sample users;
determining a target load prediction probability model according to the power utilization type of the user to be predicted and the load prediction probability models corresponding to the power utilization types respectively; the first load prediction probability model is established according to the weather data of the first sample user, the date data of the first sample user and the actual electricity utilization load data of the first sample user; the first load prediction probability model is a load prediction probability model corresponding to a first power utilization type; the first electricity utilization type is any one of the electricity utilization types; the first sample user is a sample user that conforms to the first power usage type;
inputting weather data of a user to be predicted and date data of the user to be predicted into the target load prediction probability model to obtain an output result of the target load prediction probability model;
determining the predicted electricity utilization load quantity of the user to be predicted according to the output result;
and determining the total predicted electricity load of the area to be predicted according to the predicted electricity load of all users to be predicted in the area to be predicted.
2. The method according to claim 1, wherein the load prediction probability model corresponding to the first power usage type is specifically established in the following manner:
dividing the user data of the first sample user into a first training data set, a second training data set and a test data set; the first training data set comprises weather data of a first training sample user, date data of the first training sample user and actual electricity load data of the first training sample user; the second training data set comprises weather data of a second training sample user, date data of the second training sample user and actual electricity load data of the second training sample user; the test data set comprises weather data of a test sample user, date data of the test sample user and actual electricity utilization load data of the test sample user;
training a preset point prediction model according to the weather data of the first training sample user, the date data of the first training sample user and the actual power consumption load data of the first training sample user to obtain a trained point prediction model;
inputting the weather data of the second training sample user and the date data of the second training sample user into the trained point prediction model to obtain predicted electricity utilization load data of the second training sample user;
taking the difference value of the actual electricity utilization load data of the second training sample user and the predicted electricity utilization load data of the second training sample user as the residual error of the second training sample user;
training a preset condition residual error prediction model according to the weather data of the second training sample user, the date data of the second training sample user, the predicted power consumption load data of the second training sample user and the residual error of the second training sample user to obtain a trained condition residual error prediction model;
inputting the weather data of the test sample user and the date data of the test sample user into the trained point prediction model to obtain the predicted electricity utilization load data of the test sample user;
inputting the weather data of the test sample user, the date data of the test sample user and the predicted power utilization load data of the test sample user into the trained conditional residual prediction model to obtain a residual of the test sample user;
taking a difference value corresponding to the predicted electricity utilization load quantity data of the test sample user and the residual error of the test sample user as the predicted electricity utilization load quantity of the test sample user;
and if the difference value between the predicted electricity utilization load of the test sample user and the actual electricity utilization load data of the test sample user is smaller than a preset threshold value, determining the trained point prediction model and the trained conditional residual prediction model as a load prediction probability model corresponding to the first electricity utilization type.
3. The method according to claim 1, characterized in that said plurality of electricity usage types is determined in particular by:
randomly dividing the sample user data into a plurality of initial data sets;
according to each initial data set, local clustering is carried out on sample users by using a local clustering model according to typical electricity utilization modes of the sample users in the initial data sets, electricity utilization behavior information entropies of the sample users and portrait characteristics of the sample users, and a plurality of sub-electricity utilization types are obtained;
and utilizing a global clustering model to perform global clustering on the sub-electricity utilization types respectively corresponding to the plurality of initial data sets to obtain the plurality of electricity utilization types.
4. The method according to claim 1, characterized in that the typical electricity usage pattern of the user to be predicted is determined in particular by:
acquiring various historical electricity utilization modes of a user to be predicted and the occurrence times of each historical electricity utilization mode;
determining the proportion of each historical electricity utilization mode according to the occurrence frequency of each historical electricity utilization mode and the total occurrence frequency of all historical electricity utilization modes;
and determining the historical power utilization pattern with the highest percentage as the typical power utilization pattern of the user to be predicted.
5. The method according to claim 1, wherein the image characteristics of the user to be predicted are determined by:
collecting questionnaire information of a user to be predicted; the questionnaire information comprises at least one of a user basic attribute, a user electricity utilization habit attribute, a user house attribute and a user household equipment configuration attribute;
discretizing, normalizing and coding the questionnaire information of the user to be predicted to obtain the image characteristics of the user to be predicted.
6. An apparatus for predicting a load capacity of a regional user, the apparatus comprising:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring user data in an area to be predicted; the user data comprises a typical electricity consumption mode of the user to be predicted, an electricity consumption behavior information entropy of the user to be predicted, an image characteristic of the user to be predicted, weather data of the user to be predicted and date data of the user to be predicted;
the processing module is used for determining the electricity utilization type of the user to be predicted from a plurality of electricity utilization types according to the typical electricity utilization mode of the user to be predicted, the electricity utilization behavior information entropy of the user to be predicted and the portrait characteristics of the user to be predicted; the plurality of electricity utilization types are determined according to typical electricity utilization modes of sample users, the electricity utilization behavior information entropies of the sample users and the portrait characteristics of the sample users;
the processing module is further used for determining a target load prediction probability model according to the power utilization type of the user to be predicted and the load prediction probability models corresponding to the power utilization types respectively; the first load prediction probability model is established according to the weather data of the first sample user, the date data of the first sample user and the actual electricity utilization load data of the first sample user; the first load prediction probability model is a load prediction probability model corresponding to a first power utilization type; the first electricity utilization type is any one of the electricity utilization types; the first sample user is a sample user that conforms to the first power usage type;
the forecasting module is used for inputting the weather data of the user to be forecasted and the date data of the user to be forecasted into the target load forecasting probability model to obtain an output result of the target load forecasting probability model; the power consumption prediction method comprises the steps of outputting an output result to a user to be predicted; and the total predicted electricity load quantity of the area to be predicted is determined according to the predicted electricity load quantities of all users to be predicted in the area to be predicted.
7. The apparatus of claim 6, wherein the load prediction probability model corresponding to the first power usage type is specifically established by:
dividing the user data of the first sample user into a first training data set, a second training data set and a test data set; the first training data set comprises weather data of a first training sample user, date data of the first training sample user and actual electricity load data of the first training sample user; the second training data set comprises weather data of a second training sample user, date data of the second training sample user and actual electricity load data of the second training sample user; the test data set comprises weather data of a test sample user, date data of the test sample user and actual electricity utilization load data of the test sample user;
training a preset point prediction model according to the weather data of the first training sample user, the date data of the first training sample user and the actual power consumption load data of the first training sample user to obtain a trained point prediction model;
inputting the weather data of the second training sample user and the date data of the second training sample user into the trained point prediction model to obtain predicted electricity utilization load data of the second training sample user;
taking the difference value of the actual electricity utilization load data of the second training sample user and the predicted electricity utilization load data of the second training sample user as the residual error of the second training sample user;
training a preset condition residual error prediction model according to the weather data of the second training sample user, the date data of the second training sample user, the predicted power consumption load data of the second training sample user and the residual error of the second training sample user to obtain a trained condition residual error prediction model;
inputting the weather data of the test sample user and the date data of the test sample user into the trained point prediction model to obtain the predicted electricity utilization load data of the test sample user;
inputting the weather data of the test sample user, the date data of the test sample user and the predicted power utilization load data of the test sample user into the trained conditional residual prediction model to obtain a residual of the test sample user;
taking a difference value corresponding to the predicted electricity utilization load quantity data of the test sample user and the residual error of the test sample user as the predicted electricity utilization load quantity of the test sample user;
and if the difference value between the predicted electricity utilization load of the test sample user and the actual electricity utilization load data of the test sample user is smaller than a preset threshold value, determining the trained point prediction model and the trained conditional residual prediction model as a load prediction probability model corresponding to the first electricity utilization type.
8. The apparatus of claim 6, wherein the plurality of power usage types are determined by:
randomly dividing the sample user data into a plurality of initial data sets;
according to each initial data set, local clustering is carried out on sample users by using a local clustering model according to typical electricity utilization modes of the sample users in the initial data sets, electricity utilization behavior information entropies of the sample users and portrait characteristics of the sample users, and a plurality of sub-electricity utilization types are obtained;
and utilizing a global clustering model to perform global clustering on the sub-electricity utilization types respectively corresponding to the plurality of initial data sets to obtain the plurality of electricity utilization types.
9. The apparatus according to claim 6, wherein the typical power usage pattern of the user to be predicted is determined by:
acquiring various historical electricity utilization modes of a user to be predicted and the occurrence times of each historical electricity utilization mode;
determining the proportion of each historical electricity utilization mode according to the occurrence frequency of each historical electricity utilization mode and the total occurrence frequency of all historical electricity utilization modes;
and determining the historical power utilization pattern with the highest percentage as the typical power utilization pattern of the user to be predicted.
10. The apparatus according to claim 6, wherein the image characteristic of the user to be predicted is determined by:
collecting questionnaire information of a user to be predicted; the questionnaire information comprises at least one of a user basic attribute, a user electricity utilization habit attribute, a user house attribute and a user household equipment configuration attribute;
discretizing, normalizing and coding the questionnaire information of the user to be predicted to obtain the image characteristics of the user to be predicted.
CN202010471759.3A 2020-05-29 2020-05-29 Method and device for predicting load of regional user Active CN111612275B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010471759.3A CN111612275B (en) 2020-05-29 2020-05-29 Method and device for predicting load of regional user

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010471759.3A CN111612275B (en) 2020-05-29 2020-05-29 Method and device for predicting load of regional user

Publications (2)

Publication Number Publication Date
CN111612275A true CN111612275A (en) 2020-09-01
CN111612275B CN111612275B (en) 2022-04-01

Family

ID=72201600

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010471759.3A Active CN111612275B (en) 2020-05-29 2020-05-29 Method and device for predicting load of regional user

Country Status (1)

Country Link
CN (1) CN111612275B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112070307A (en) * 2020-09-10 2020-12-11 云南电网有限责任公司电力科学研究院 Method and device for predicting energy source load in region
CN112330017A (en) * 2020-11-04 2021-02-05 上海博辕信息技术服务有限公司 Power load prediction method, power load prediction device, electronic device, and storage medium
CN112769135A (en) * 2021-03-26 2021-05-07 中国电力科学研究院有限公司 Method and system for power regulation of power consumption unit
CN113193556A (en) * 2021-05-12 2021-07-30 燕山大学 Short-term wind power prediction method based on probability prediction model
CN113762601A (en) * 2021-08-12 2021-12-07 北京市燃气集团有限责任公司 Gas user gas consumption prediction method and device
CN114841832A (en) * 2022-07-04 2022-08-02 国网湖北省电力有限公司营销服务中心(计量中心) Power consumer portrait label establishing method based on secondary clustering of power loads

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9432901B1 (en) * 2015-07-24 2016-08-30 Cisco Technology, Inc. System and method to facilitate radio access point load prediction in a network environment
CN105989420A (en) * 2015-02-12 2016-10-05 西门子公司 Method of determining user electricity consumption behavior features, method of predicting user electricity consumption load and device
CN106096805A (en) * 2016-05-10 2016-11-09 华北电力大学 A kind of residential electricity consumption load classification method based on entropy assessment feature selection
CN108062560A (en) * 2017-12-04 2018-05-22 贵州电网有限责任公司电力科学研究院 A kind of power consumer feature recognition sorting technique based on random forest
CN109815994A (en) * 2019-01-04 2019-05-28 国网山西省电力公司长治供电公司 A kind of user power utilization load characteristic analysis method
CN109886449A (en) * 2018-12-30 2019-06-14 国网浙江省电力有限公司衢州供电公司 A kind of area power grid load forecasting method based on user power utilization behavior
CN110674993A (en) * 2019-09-26 2020-01-10 广东电网有限责任公司 User load short-term prediction method and device

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105989420A (en) * 2015-02-12 2016-10-05 西门子公司 Method of determining user electricity consumption behavior features, method of predicting user electricity consumption load and device
US9432901B1 (en) * 2015-07-24 2016-08-30 Cisco Technology, Inc. System and method to facilitate radio access point load prediction in a network environment
CN106096805A (en) * 2016-05-10 2016-11-09 华北电力大学 A kind of residential electricity consumption load classification method based on entropy assessment feature selection
CN108062560A (en) * 2017-12-04 2018-05-22 贵州电网有限责任公司电力科学研究院 A kind of power consumer feature recognition sorting technique based on random forest
CN109886449A (en) * 2018-12-30 2019-06-14 国网浙江省电力有限公司衢州供电公司 A kind of area power grid load forecasting method based on user power utilization behavior
CN109815994A (en) * 2019-01-04 2019-05-28 国网山西省电力公司长治供电公司 A kind of user power utilization load characteristic analysis method
CN110674993A (en) * 2019-09-26 2020-01-10 广东电网有限责任公司 User load short-term prediction method and device

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
ALEX RODRIGUEZ等: "clustering by fast search and find of density peaks", 《SCIENCE》 *
RUIJIU JIN等: "an orderly power utilization scheme based on an imtelligent multi-agent apanage management system", 《ENERGIES》 *
彭晨宇等: "结合多重聚类和分层聚类的超短期风电功率预测方法", 《电力系统自动化》 *
武星辰: "地市级地区短期负荷分类预测方法的研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 *
龚钢军等: "智能用电用户行为分析的聚类优选策略", 《电力系统自动化》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112070307A (en) * 2020-09-10 2020-12-11 云南电网有限责任公司电力科学研究院 Method and device for predicting energy source load in region
CN112070307B (en) * 2020-09-10 2023-01-20 云南电网有限责任公司电力科学研究院 Method and device for predicting energy source load in region
CN112330017A (en) * 2020-11-04 2021-02-05 上海博辕信息技术服务有限公司 Power load prediction method, power load prediction device, electronic device, and storage medium
CN112769135A (en) * 2021-03-26 2021-05-07 中国电力科学研究院有限公司 Method and system for power regulation of power consumption unit
CN113193556A (en) * 2021-05-12 2021-07-30 燕山大学 Short-term wind power prediction method based on probability prediction model
CN113762601A (en) * 2021-08-12 2021-12-07 北京市燃气集团有限责任公司 Gas user gas consumption prediction method and device
CN114841832A (en) * 2022-07-04 2022-08-02 国网湖北省电力有限公司营销服务中心(计量中心) Power consumer portrait label establishing method based on secondary clustering of power loads
CN114841832B (en) * 2022-07-04 2023-01-24 国网湖北省电力有限公司营销服务中心(计量中心) Power consumer portrait label establishing method based on secondary clustering of power loads

Also Published As

Publication number Publication date
CN111612275B (en) 2022-04-01

Similar Documents

Publication Publication Date Title
CN111612275B (en) Method and device for predicting load of regional user
Ghadimi et al. PSO based fuzzy stochastic long-term model for deployment of distributed energy resources in distribution systems with several objectives
CN111199016A (en) DTW-based improved K-means daily load curve clustering method
EP3104323A1 (en) Electric power business profit and loss calculation system and electric power business profit and loss calculation method
CN106022614A (en) Data mining method of neural network based on nearest neighbor clustering
Xie et al. An integrated Gaussian process modeling framework for residential load prediction
CN114372360A (en) Method, terminal and storage medium for power load prediction
Gao et al. Spectral clustering based demand-oriented representative days selection method for power system expansion planning
CN109583086A (en) Distribution transformer heavy-overload prediction technique and terminal device
Hou et al. Reliability assessment of power systems with high renewable energy penetration using shadow price and impact increment methods
Khuntia et al. Risk‐based security assessment of transmission line overloading considering spatio‐temporal dependence of load and wind power using vine copula
Sari et al. Statistical reliability of wind power scenarios and stochastic unit commitment cost
Ilkhani et al. Extraction test cases by using data mining; reducing the cost of testing
CN101206727B (en) Data processing apparatus, data processing method
CN107274025B (en) System and method for realizing intelligent identification and management of power consumption mode
CN112508254B (en) Method for determining investment prediction data of transformer substation engineering project
CN113887809A (en) Power distribution network supply and demand balance method, system, medium and computing equipment under double-carbon target
CN111145045B (en) VaR-considered power large-user flexible load assessment method and system
Niu et al. Knowledge mining based on environmental simulation applied to wind farm power forecasting
CN111080489A (en) Load trend prediction method, system and equipment
CN114372835A (en) Comprehensive energy service potential customer identification method, system and computer equipment
Ji et al. Online refinement of day‐ahead forecasting using intraday data for campus‐level load
CN114970928A (en) Electric power data energy consumption analysis and prediction method
Gajowniczek et al. Seasonal peak demand classification with machine learning techniques
CN111915105A (en) Method and device for predicting electricity consumption in area

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant