Crowdsourced sensor data prediction-based dynamic construction method and system for electricity consumption map

Abstract

The application relates to a dynamic construction method and system of an electricity consumption map based on crowd-sourced sensing data prediction, which comprises the following steps: S1, a real-time crowd-sourced sensing data cleaning step, which is used for collecting historical electricity consumption data and real-time sensor data, eliminating redundant data, error data, and simulating and supplementing missing data; S2, a regional electricity consumption prediction step, which is used for constructing a model; S3, a regional electricity consumption prediction dynamic updating step, which is used for periodically optimizing an electricity consumption prediction model based on newly collected user electricity consumption information, so that the prediction accuracy is improved; S4, an electricity consumption prediction data map grading association step, which is used for associating real-time electricity consumption prediction data with a map; and S5, an electricity consumption map dynamic construction step, which is used for real-time rendering and display of a newly-built ArcGIS map in step four and associated prediction electricity consumption. The application can realize small-granularity prediction of regional electricity consumption.

Description

TECHNICAL FIELD

[0001] The present application relates to the research of the power consumption map construction method, in particular to a power consumption map dynamic construction method and system based on crowd source sensing data prediction. BACKGROUND

[0002] With the advent of the intelligent and digital era, global information resources have entered a stage of rapid growth, and big data technology, which is born from huge information resources, has caused industrial effects in various industries and become a new production factor. In recent years, with the progress of society and the improvement of people's material living standards, industrial and personal power consumption continues to increase, and power consumption information and environmental and geographical location also have high correlation. Therefore, how to accurately predict regional and user power consumption at the macro and micro levels can provide effective data support for power dispatching and power marketing of the power department.

[0003] On the other hand, power consumption information has high correlation with spatial location, and today when map visualization is gradually mature, associating power consumption information with electronic maps can provide intuitive and accurate auxiliary decision-making services for power site dispatching, disaster emergency response, and real-time command of decision-making centers. SUMMARY

[0004] The present application mainly solves the above technical problems existing in the prior art, and provides a power consumption map dynamic construction method and system based on crowd source sensing data prediction. The method uses real-time collected sensor data, combines historical power consumption data, establishes a regional power consumption prediction model, supports real-time dynamic updating of prediction data, and finally realizes dynamic construction of power consumption prediction map by associating power consumption prediction data with map location information.

[0005] The above technical problems of the present application are mainly solved by the following technical scheme:

[0006] A power consumption map dynamic construction method based on crowd source sensing data prediction comprises the following steps:

[0007] S1, a real-time crowd source sensing data cleaning step, is used to collect historical power consumption data and real-time sensor data, eliminate redundant data, error data, and simulate to supplement missing data, so as to provide complete data support for subsequent power consumption prediction;

[0008] S2, a regional power consumption prediction step, is used to construct a model to realize real-time prediction of regional power consumption;

[0009] S3, a regional power consumption prediction dynamic updating step, is used to periodically optimize the power consumption prediction model based on newly collected user power consumption information, so as to improve the prediction accuracy;

[0010] S4 use electricity consumption prediction data map classification association step, this step user will real-time electricity consumption prediction data and map classification association;

[0011] S5 electricity map dynamic construction step, for the step four new ArcGIS map and associated prediction electricity consumption, real-time rendering display.

[0012] Preferably, S1 specific implementation steps are:

[0013] S11 history data collection part, need to collect the user's electricity information in the past years, including the daily electricity consumption, electricity consumption period and the temperature, humidity and rainfall information of the day in the past three years;

[0014] S12 real-time sensor data part, need to collect real-time or daily average temperature, average humidity and rainfall information;

[0015] S13 in real-time sensor data part, need to collect the humidity and rainfall information corresponding to the user's location, because the humidity and rainfall information cannot be directly obtained from the user terminal device, so the virtual sensor is used to collect;

[0016] S14 WD is considered as the current location real-time humidity and rainfall data obtained by the user virtual sensor.

[0017] Preferably, the S12 collection method is:

[0018] S121 for real-time daily average temperature, from the temperature sensor module of the user's smart meter to obtain 24 hours of time period temperature data, collecting once an hour;

[0019] S122 automatic analysis of 24 data collected every day, when the temperature difference of consecutive hours is greater than 5℃, it is considered as abnormal data and is eliminated;

[0020] S123 the temperature data of this hour after elimination, is replaced by the average interpolation method of the previous and next hour;

[0021] S124 take the average of the 24 hours of time period temperature data after sorting, as the real-time daily average temperature data of the smart meter user.

[0022] Preferably, the S13 specific collection method is:

[0023] S131 use crawler software to obtain real-time weather information on the Internet, collect the minimum administrative area weather data that can be obtained, such as Wuhan area, can collect the weather information of each administrative region of Wuchang district, Jiang'an district and Hanyang district;

[0024] S132in the planar map, the geometric center position of each administrative region in the region is calculated, and the collected administrative region weather information is marked as the weather information of the corresponding geometric center;

[0025] S133for the power user, according to its location, the straight line distance from each administrative region geometric center in the region is calculated;

[0026] S134synchronize the weather information of each administrative region geometric center to the user according to the length of the straight line distance between the user and each administrative region; A B C A B C A B C A B C

[0027]

[0028] Preferably, S2 is specifically implemented as follows:

[0029] S21firstly, a deep learning training framework is constructed, Tensorflow 2.0 is selected as the deep learning platform, and the long short-term memory network in the recurrent neural network is used as the training framework for the subsequent training of the power consumption prediction model;

[0030] S22secondly, samples and sample labels need to be divided, the samples are the historical data center each power user daily multi-source sensor data, including real-time collected and recorded temperature data, and the humidity and rainfall data obtained by constructing the virtual sensor in step S13; the sample label is labeled as the daily power consumption corresponding to the daily multi-source sensor data of a user in the historical data;

[0031] S23then, the training of the prediction model needs to be realized, the LSTM training framework constructed in step S21 and the sample set sorted in step S22 are used for model training; all users in the region have an independent power consumption prediction model;

[0032] S24finally, the regional power consumption is predicted in real time and the numerical output is realized.

[0033] Preferably, S24 is specifically implemented as follows:

[0034] ​​​​​​​​​​​​S241 apply the series of models trained in step S23 to each independent model of each power user to realize the power consumption prediction of each independent user on a certain day in the future;

[0035] S242 statistically classify the location information and the region information of all power users;

[0036] S243 predict the power consumption of all users in different regions respectively, for example, in two levels: the first level predicts the power consumption of a city, and the second level predicts the power consumption of each administrative region;

[0037] S244 digitally output the power consumption prediction results of different levels of regions.

[0038] Preferably, the step S3 is specifically implemented as follows:

[0039] S31 periodically add the latest power consumption data of each user, including the cleaning and arrangement according to the method in step S1, as a unit of week or month;

[0040] S32 add the cleaned and arranged new power consumption data of the user to the sample library, including the sample and the label;

[0041] S33 under the training environment of S2, retrain the trained power consumption prediction model of the user, and adjust the training sample set to the sample set after the addition in step S32;

[0042] S34 use the model after the retraining for real-time power consumption prediction, and output the data according to the method in step S24.

[0043] Preferably, the step S4 is specifically implemented as follows:

[0044] S41 in the ArcGIS software, based on the base map, newly create a zoning layer, and the number of layers is according to the number of levels in step S243; if the number of levels is two, i.e. Wuhan city and each district, then two layers are newly created, one layer displays the whole Wuhan city, and the other layer displays each district;

[0045] S42 in the newly created layer of ArcGIS, each region is constructed by a surface feature, and each surface feature is newly added with an attribute of power consumption;

[0046] S43 newly create a layer for storing all power users, and the power users are stored by point features, and the position of the point feature is the actual position of the power user;

[0047] S44 for the region where the power users are concentrated, the point features are merged and stored as one point feature, and the power consumption attribute of the point feature is recorded as the sum of the power consumptions of all users at the position;

[0048] S45 to this is achieved hierarchical regional surface element and the data associated with the region predicted power consumption, and the predicted power consumption of the user and point element associated.

[0049] Preferably, S5 specifically implements the following steps:

[0050] S51 adds an interactive control in the map interface, which is used to realize human-computer interaction and respond to the region selection command of the operator.

[0051] S52 renders the region in the layer with a single color, and the single color is selected from a single red color, a single green color or a single blue color in the RGB color space.

[0052] S53 normalizes the power consumption prediction data in each layer to the interval of 0-255.

[0053] S54 renders the corresponding region with the color number processed in step S53.

[0054] S55 normalizes the single-user power consumption prediction data to the interval of 0-100 for the user layer built in step S43.

[0055] S56 displays the size of the point element of the single user with the value processed in step S55.

[0056] S57 to this is achieved dynamic construction of the power consumption map.

[0057] A system for dynamically constructing a power consumption map based on crowd-sourced sensing data prediction, comprising a real-time crowd-sourced sensing data cleaning module, a regional power consumption prediction module, a regional power consumption prediction dynamic updating module, a power consumption prediction data map hierarchical correlation module and a power consumption map dynamic construction module connected in sequence.

[0058] The present application has the following beneficial effects:

[0059] A power consumption prediction model is established for all users in the region, and small-grained prediction of regional power consumption can be achieved.

[0060] The real temperature sensor and the virtual sensor are combined to realize more accurate perception of the power consumption environment.

[0061] The model is updated regularly, and the generalization ability of the prediction model to new samples is realized, which means that the power consumption can be analyzed and predicted according to the recent power consumption behavior and power consumption environment characteristics of the user.

[0062] Real-time dynamic implementation of power consumption prediction map hierarchical display, including regional power consumption and individual power consumption, is more intuitive.BRIEF DESCRIPTION OF DRAWINGS BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 is a flow diagram of a method for dynamically constructing an electricity consumption map based on crowd-sourced sensor data according to the present application;

[0064] Figure 2 is a block diagram of a system for dynamically constructing an electricity consumption map based on crowd-sourced sensor data according to the present application. DETAILED DESCRIPTION

[0065] In order to make the objects, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in connection with the drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0066] In the embodiments of the present application, taking Wuhan City in Hubei Province as an example, a method and system for dynamically constructing an electricity consumption map based on crowd-sourced sensor data are constructed, and the entire system is realized in the form of software.

[0067] As shown in Figure 1 , a method for dynamically constructing an electricity consumption map based on crowd-sourced sensor data comprises the following steps:

[0068] S1, a real-time crowd-sourced sensor data cleaning step, is configured to collect and clean historical electricity consumption data and real-time sensor data, eliminate redundant data and error data, and simulate to supplement missing data, thereby providing complete data support for subsequent electricity consumption prediction;

[0069] The specific implementation steps of S1 are as follows:

[0070] S11, a historical data collection part, is configured to collect the electricity consumption information of users in the region in previous years, specifically including the daily electricity consumption in the previous three years, the electricity consumption period, and the temperature, humidity, and rainfall information on the day;

[0071] S12, a real-time sensor data part, is configured to collect real-time or daily average temperature, average humidity, and rainfall information; the collection method of S12 is as follows:

[0072] S121, for real-time daily average temperature, the temperature data of each time period in 24 hours is obtained from the temperature sensor module at the user's smart meter end, and the data is collected every hour;

[0073] S122, the collected 24 pieces of data per day are automatically analyzed, and when the temperature difference of consecutive hours is greater than 5℃, the abnormal data is identified and eliminated;

[0074] S123 The hourly temperature data after the rejection is replaced by the average interpolation of the previous and next hour segments;

[0075] S124 The average of the sorted 24-hour time period temperature data is taken as the real-time daily average temperature data of the smart meter user.

[0076] S13 In the real-time sensing data part, humidity and rainfall information corresponding to the user's location needs to be collected. Since humidity and rainfall information cannot be directly obtained from the user terminal device, a virtual sensor is used to collect it. The specific collection method of S13 is:

[0077] S131 Use crawler software to obtain real-time weather information on the Internet, collect the minimum administrative area weather data that can be obtained, such as Wuhan area, and collect the weather information of each administrative area in Wuchang District, Jiang'an District, and Hanyang District.

[0078] S132 In the plane map, calculate the geometric center position of each administrative area in the region, and mark the collected administrative area weather information as the weather information of the corresponding geometric center.

[0079] S133 For electricity users, calculate the straight-line distance from their location to the geometric center of each administrative area in the region.

[0080] S134 Use the straight-line distance between the electricity user and each administrative area as the weight to synchronize the weather information of the geometric center of each administrative area to the user. The calculation method of S134 is as follows: A region has three administrative areas A, B, and C, and their geometric centers are C A 、C B 、C C , the weather information corresponding to the geometric center is W A 、W B 、W C , the straight-line distance between user X and C A 、C B 、C C is D A 、D B 、D C , then the weather data of the user's location includes humidity and rainfall, and the calculation is:

[0081]

[0082] S14 W D is considered the current location real-time humidity and rainfall data obtained by the user's virtual sensor.

[0083] S2 Regional electricity consumption prediction step, used to build a model to realize real-time prediction of regional electricity consumption;

[0084] S2 is specifically implemented as follows:

[0085] S21 first constructs a deep learning training framework, selects Tensorflow 2.0 as a deep learning platform, and adopts a long short-term memory network in a recurrent neural network as a training framework for training of a subsequent power consumption prediction model;

[0086] S22 secondly needs to divide samples and sample labels, the sample is the historical data center each power user daily multi-source sensor data, including real-time collection and recording of temperature data, and humidity and rainfall data obtained by constructing a virtual sensor in step S13; the sample label is labeled as the daily power consumption corresponding to the multi-source sensor data of a user on a day in the historical data;

[0087] S23 then needs to realize training of the prediction model, and the model is trained by using the LSTM training framework constructed in step S21 and the sample set sorted in step S22; all users in the region have an independent power consumption prediction model;

[0088] S24 finally predicts the regional power consumption in real time and outputs the numerical value.

[0089] S24 is specifically implemented as follows:

[0090] S241 uses a series of models trained in step S23, and realizes prediction of the power consumption of each independent user on a future day for the independent model of each power user;

[0091] S242 classifies the location information and the region information of all power users;

[0092] S243 predicts the power consumption of all users in different regions respectively, such as two levels: the first level predicts the municipal power consumption, and the second level predicts the power consumption of each administrative region;

[0093] S244 digitizes and outputs the power consumption prediction results of different levels of regions.

[0094] S3 is a dynamic updating step of the regional power consumption prediction, which is used for periodically optimizing the power consumption prediction model based on newly collected user power consumption information, and realizing improvement of prediction accuracy;

[0095] S3 is specifically implemented as follows:

[0096] S31 periodically adds the latest power consumption data of each user to the sample library in a weekly or monthly unit, including cleaning and sorting according to the method in step S1;

[0097] S32 adds the cleaned and sorted user new power consumption data to the sample library, including samples and labels;

[0098] S33 In the training environment of S2, the trained user electricity consumption prediction model is retrained, and the training sample set is adjusted to the sample set newly added in step S32;

[0099] S34 The retrained model is used for real-time electricity consumption prediction, and data is output in the manner of step S24.

[0100] S4 Electricity consumption prediction data map classification association step, in which the user associates real-time electricity consumption prediction data with map classification;

[0101] S4 The specific implementation steps are:

[0102] S41 In the ArcGIS software, a new zoning layer is created based on the base map, and the number of layers is determined according to the number of layers in step S243. If the number of layers is 2, two new layers are created, one for displaying the entire city of Wuhan and the other for displaying each district.

[0103] S42 In the newly added layer of ArcGIS, each region is constructed as a face element, and a new attribute of electricity consumption is added to each face element.

[0104] S43 A new layer is created again to store all electricity users, and the electricity users are stored as point elements with the actual location of the electricity users as the position of the point elements.

[0105] S44 For areas with dense electricity users, they are merged into a point element, and the electricity consumption attribute of the point element is recorded as the sum of the electricity consumption of all users at that location.

[0106] S45 The hierarchical regional face element and the predicted electricity consumption of the region are associated, and the predicted electricity consumption of the electricity user and the point element are associated.

[0107] S5 Electricity map dynamic construction step, used to real-time render and display the ArcGIS map newly created in step four and the associated predicted electricity consumption.

[0108] S5 The specific implementation steps are:

[0109] S51 Add an interactive control to the map interface to realize human-computer interaction and respond to the operator's area selection command.

[0110] S52 Render the area in the layer with a single color, and choose a single red, single green, or single blue color in the RGB color space.

[0111] S53 Normalize the electricity consumption prediction data in each layer to the range of 0-255.

[0112] S54 Render the corresponding area with the color number processed in step S53.

[0113] S55 For the power user layer built in step S43, the single user power consumption prediction data is normalized to the 0-100 interval;

[0114] S56, the single user point element size is displayed with the value processed in step S55;

[0115] S57 Thus, the dynamic construction of the power consumption map is realized.

[0116] As Figure 2 shown: in the power consumption map dynamic construction system based on crowd source sensing data prediction of the embodiment of the application, the following functional modules are included:

[0117] Real-time crowd source sensing data cleaning module 10. The main function of this module is to collect and clean the historical power consumption data and real-time sensor data, eliminate redundant data, error data, and simulate missing data, to provide complete data support for subsequent power consumption prediction. Among them, the historical data acquisition function part collects the user's power consumption information in Wuhan in previous years, specifically including the daily power consumption, power consumption period, and temperature, humidity and rainfall information of the day in the previous three years; the real sensor data acquisition function part obtains 24-hour time period temperature data from the temperature sensor module at the user's smart meter end, collects once an hour, and then calculates to obtain the daily average temperature data of the associated user; the virtual sensor data acquisition function part uses a crawler software to obtain the real-time weather information of Wuhan city on the Internet, including the daily average temperature of Wuhan city and the daily average temperature of 13 administrative districts in Wuhan city.

[0118] Further, for the humidity and rainfall data collected by the virtual sensor, linear interpolation is used to achieve, specifically, the straight-line distance between the power consumption user and each administrative district is used as the weight, and the weather information of the geometric center of each administrative district is synchronized to the user.

[0119] In particular, automatic analysis is performed on the 24 collected data each day, and when the temperature difference of consecutive hours is greater than 5℃, it is determined as abnormal data and is eliminated, and after the elimination of the temperature data of the hour, the average interpolation method of the previous and next hour is used to replace it.

[0120] The regional power consumption prediction module 20. The main function of this module is to build a model to realize real-time prediction of regional power consumption. First, build a deep learning training framework, this embodiment selects Tensorflow 2.0 as the deep learning platform, and uses LSTM as the training framework; second, divide the samples and sample labels, the samples are the historical data of each power user on a certain day, including the temperature data collected and recorded in real time, and the humidity and rainfall data obtained by the virtual sensor, and the sample labels are labeled as the daily power consumption corresponding to the daily multi-source sensor data of a certain user in the historical data; then the training of the prediction model needs to be realized, the LSTM training framework and the sample set constructed in step 2.1 are used for model training, and all users in the region have an independent power consumption prediction model; finally, the regional power consumption is predicted in real time and the numerical output is obtained.

[0121] The regional power consumption prediction dynamic updating module 30. The main function of this step is to periodically optimize the power consumption prediction model based on the newly collected user power consumption information, and to realize the improvement of prediction accuracy. First, the latest power consumption data of each user needs to be regularly cleaned and arranged according to the method in step one, including weekly or monthly units (this embodiment takes weeks as the unit); then add the cleaned and arranged user new power consumption data to the sample library, including samples and labels; then, under the training environment of step two, the trained user power consumption prediction model is trained again, and the training sample set is adjusted to the sample set after the addition of real-time samples; finally, the model after the secondary training is used for real-time power consumption prediction, and the new prediction data can be output by the regional power consumption prediction module 20.

[0122] The power consumption prediction data map classification association module 40. The main function of this module is to associate the real-time power consumption prediction data with the map classification. First, in the ArcGIS software, a new zoning layer is created based on the base map, and the number of layers is divided into two levels of Wuhan City and administrative districts, corresponding to two new layers, one layer displays the whole of Wuhan City, and the other layer displays each district; then in the newly added layer of ArcGIS, each region is constructed with a surface feature, and each surface feature adds a new attribute - power consumption; then a new layer is created again to store all power users, and the power users are stored as point features, and the location of the point feature is the actual location of the power user; thus, the data association between the classified regional surface feature and the predicted power consumption of the region, and the association between the predicted power consumption of the power user and the point feature are realized.

[0123] In particular, for the densely populated area of power users, it is stored as a point feature, and the power consumption attribute of the point feature is recorded as the sum of the power consumption of all users at this location.

[0124] The power map dynamic construction module 50. The main function of this module is to render and display the ArcGIS map and the associated predicted power consumption newly built by the module 40 in real time. First, add an interactive control to the map interface to realize human-computer interaction and respond to the area selection command of the operator; second, render the area in the layer with a single color, and choose a single red, single green, or single blue color in the RGB color space; then normalize the power consumption prediction data in each layer to the interval of 0-255, and render the corresponding area with the processed color number; then, for the power user layer built, normalize the single user power consumption prediction data to the interval of 0-100, and then display the single user point feature size with the value.

[0125] Those skilled in the art can further realize that the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been described in the above description in a functional manner. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not exceed the scope of the present application.

[0126] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein can be directly implemented in hardware, software modules executed by a processor, or a combination of both. The software modules can be placed in a random storage, a memory, a read-only memory, an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0127] It can be understood that, for those skilled in the art, other various corresponding changes and modifications can be made according to the technical concept of the present application, and all such changes and modifications shall fall within the protection scope of the claims of the present application.

[0128] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present application can be easily thought of by those skilled in the art, and shall be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for dynamically constructing a power consumption map based on crowdsourced sensor data prediction, characterized in that, Comprise the following steps: S1 real-time crowd source sensor data cleaning step, for collecting and historical electricity data and real-time sensor data cleaning, eliminate redundant data, error data, simulation to supplement missing data, for subsequent electricity consumption prediction provides complete data support; S2 regional electricity consumption prediction step, for building model, realize the real-time prediction of regional electricity consumption; S3 regional electricity consumption prediction dynamic update step, for periodically based on the new collected user electricity information, optimization electricity consumption prediction model, realize the improvement of prediction accuracy; S4 electricity consumption prediction data map classification association step, this step user real-time electricity consumption prediction data and map classification association; S5 electricity map dynamic construction step, for the step four new ArcGIS map and associated prediction electricity, real-time rendering display; S4 specific implementation steps are: S41 in ArcGIS software, based on the base map, new zoning layer, the number of layers is accurate to the number of sub-layers; Divided into Wuhan and each district 2 level, then add two layers, one layer shows the whole of Wuhan, one layer shows each district; S42 in ArcGIS new layer, each region is built with face elements, each face element adds a property of electricity consumption; S43 again new layer, for storing all electricity users, electricity users are stored as point elements, the location of point element is the actual location of electricity user; S44 for the area where electricity users are concentrated, then merge and store as a point element, the electricity consumption attribute of point element is recorded as the sum of electricity consumption of all users at this location; S45 thus realized the data association of hierarchical regional face elements and the predicted electricity consumption of the region, as well as the association of electricity user predicted electricity consumption and point elements.

2. The method of claim 1, wherein the method further comprises: S1 specific implementation steps are: S11 historical data collection part, need to collect the electricity information of users in the region in the past years, including daily electricity consumption, electricity time period and temperature, humidity and rainfall information of the day in the past three years; S12 real-time sensor data part, need to collect real-time or daily average temperature, average humidity and rainfall information; S13 in real-time sensor data part, need to collect the humidity and rainfall information corresponding to the location of the user, since the humidity and rainfall information cannot be directly obtained from the user terminal device, therefore, virtual sensor is adopted to collect; S14W D The current location real-time humidity and rainfall data obtained by the user virtual sensor are considered. 3.The method of claim 2, wherein, S12 collection method is: S121 for real-time daily average temperature, obtain 24 hours of time period temperature data from temperature sensor module of user smart meter, collect once an hour; S122 automatically analyze 24 data collected every day, if the temperature difference of consecutive hours is greater than 5℃, it is determined as abnormal data and eliminated; S123 the temperature data of this hour after elimination, is replaced by the average interpolation method of previous and next hour; S124 take the average of 24 hours of time period temperature data after sorting, as the real-time daily average temperature data of the smart meter user. 4.The method of claim 2, wherein, S13 specific collection method is: S131 obtain real-time weather information on the Internet by using crawler software, collect the minimum administrative area weather data that can be obtained; S132in the planar map, the geometric center position of each administrative region in the region is calculated, and the collected administrative region weather information is marked as the weather information of the corresponding geometric center; S133for the power user, according to its location, the straight line distance with the geometric center of each administrative region in the region is calculated; S134 synchronizes the weather information of the geometric center of each administrative region to the user with the length of the straight line distance between the user and each administrative region as the weight; the calculation method of S134 is as follows: a region has three administrative regions A, B and C, the geometric centers of which are recorded as C A , B , C , the weather information corresponding to the geometric centers is recorded as W A , W B , W C , the straight line distance between user X and C A , C B , C C is recorded as D A , D B , D C , then the weather data of the location where the user is located includes humidity and rainfall and is calculated as:

5. The method of claim 2, wherein the method further comprises: S2the specific implementation steps are: S21first, a deep learning training framework is constructed, tensorflow 2.0 is selected as the deep learning platform, and long short-term memory network in recurrent neural network is used as the training framework for subsequent power consumption prediction model training; S22secondly, samples and sample labels need to be divided, samples are historical data center each power user's daily multi-source sensor data on a certain day, including real-time collected and recorded temperature data, and humidity and rainfall data obtained by constructing virtual sensors in step S13; The sample label is labeled as the daily power consumption corresponding to the multi-source sensor data of a certain user on a certain day in the historical data; S23then the training of the prediction model needs to be realized, the LSTM training framework constructed in step S21 and the sample set sorted in step S22 are used for model training; All users in the region have an independent power consumption prediction model; S24finally, the regional power consumption is predicted in real time and the numerical output is realized.

6. The method of claim 5, wherein the method further comprises: S24the specific method is: S241a series of models trained in step S23 are used to realize the power consumption prediction of each independent user on a certain future day for each power user's independent model; S242the location information and region information of all power users are statistically classified; S243the power consumption of all users in different regions is predicted respectively, which is divided into two levels: the first level is to predict the city-level power consumption, and the second level is to predict the power consumption of each administrative region; S244the power consumption prediction results of different levels of regions are digitized and output.

7. The method of claim 5, wherein the method further comprises: S3the specific implementation steps are: S31periodically, the latest power consumption data of each user is cleaned and sorted according to the method in step S1; S32the cleaned and sorted user's new power consumption data is added to the sample library, including samples and labels; S33under the training environment of S2, the trained user power consumption prediction model is trained again, and the training sample set is adjusted to the sample set after step S32 is added; S34the model after the second training is used for real-time power consumption prediction, and the data is output according to the method of step S24. 8.The method of claim 1, wherein, S5the specific implementation steps are: S51add an interactive control in the map interface to realize human-computer interaction and respond to the region selection command of the operator; S52for the region in the layer, render it in a single color, and choose a single red, green or blue color in the RGB color space; S53normalize the power consumption prediction data in each layer to the interval of 0-255; S54render the corresponding region with the color number processed in step S53; S55for the power user layer built in step S43, normalize the single user power consumption prediction data to the interval of 0-100; S56display the point feature size of a single user with the value processed in step S55; S57thus, the dynamic construction of the power consumption map is realized.

9. A system for dynamic construction of a power consumption map based on crowd-sourced sensor data prediction, for the method of dynamic construction of a power consumption map based on crowd-sourced sensor data prediction according to any one of claims 1-8, characterized in that, The application comprises a real-time crowd-sourced sensor data cleaning module (10), a regional power consumption prediction module (20), a regional power consumption prediction dynamic updating module (30), a power consumption prediction data map hierarchical correlation module (40) and a power consumption map dynamic construction module (50) connected in sequence.

Images