CN112614011B

CN112614011B - Power distribution network material demand prediction method and device, storage medium and electronic equipment

Info

Publication number: CN112614011B
Application number: CN202011419119.4A
Authority: CN
Inventors: 陆斯悦; 徐蕙; 陈平; 王艳松; 李香龙; 张禄; 王培祎; 盛慧慧; 严嘉慧; 马龙飞
Original assignee: State Grid Corp of China SGCC; State Grid Beijing Electric Power Co Ltd
Current assignee: State Grid Corp of China SGCC; State Grid Beijing Electric Power Co Ltd
Priority date: 2020-12-07
Filing date: 2020-12-07
Publication date: 2024-03-15
Anticipated expiration: 2040-12-07
Also published as: CN112614011A

Abstract

The invention discloses a method and a device for predicting material demands of a power distribution network, a storage medium and electronic equipment. The method comprises the following steps: preprocessing the use record parameters, the distribution network history planning data and the distribution network regional economic development data of the historical distribution network materials of the target region to obtain a preprocessing data set of the distribution network materials of the target region; dividing the preprocessing data set to obtain preprocessing sub-data sets; clustering the preprocessed sub-data sets to obtain a data cluster; inputting the data cluster into a material prediction model to output the predicted material usage amount required by the power distribution network construction material project of the target area operation; the material prediction model is a decision model for predicting the material usage amount, which is obtained by training sample material data for multiple times. The invention solves the technical problem of larger subjective judgment error caused by the traditional distribution network material manual prediction.

Description

Power distribution network material demand prediction method and device, storage medium and electronic equipment

Technical Field

The invention relates to the technical field of electric power material prediction, in particular to a method and a device for predicting demand of electric power distribution network materials, a storage medium and electronic equipment.

Background

With the development of the Internet of things, the construction and maintenance of the distribution network are one of important works of the power system, the related engineering of the distribution network is complex, and a large amount of materials are consumed in the implementation. By predicting the demand of the distribution network materials, the total demand of various materials in the electric power construction project in the future time period is determined, reasonable material purchasing and mobilization are carried out, the influence of insufficient materials on the distribution network engineering can be avoided, and meanwhile, the material backlog risk and the fund waste caused by excessive purchasing are reduced.

The traditional distribution network material demand method adopts a manual mode and combines project report and field investigation to further perfect, but the distribution network material demand prediction result is subjectively influenced by a decision maker, the actual error between the prediction result and the time is often larger, and the prediction efficiency is lower.

In view of the above problems, no effective solution has been proposed at present.

Disclosure of Invention

The embodiment of the invention provides a method and a device for predicting the demand of power distribution network materials, a storage medium and electronic equipment, which are used for at least solving the technical problem of larger subjective judgment error caused by manual prediction of the traditional distribution network materials.

According to an aspect of the embodiment of the invention, there is provided a method for predicting a demand for materials of a power distribution network, including: preprocessing the use record parameters, the distribution network history planning data and the distribution network regional economic development data of the historical distribution network materials of the target region to obtain a preprocessing data set of the distribution network materials of the target region; dividing the preprocessing data set to obtain preprocessing sub-data sets; wherein the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected; clustering the preprocessed sub-data sets to obtain a data cluster; inputting the data cluster into a material prediction model to output the predicted material usage amount required by the target area to run the power distribution network construction material project; the material prediction model is a decision model for predicting the material usage amount, which is obtained by training sample material data for multiple times.

According to another aspect of the embodiment of the present invention, there is also provided a power distribution network material demand prediction apparatus, including: the preprocessing unit is used for preprocessing the use record parameters, the distribution network history planning data and the distribution network regional economic development data of the historical distribution network materials of the target region to obtain a preprocessing data set of the distribution network materials of the target region; the dividing unit is used for dividing the preprocessing data set to obtain a preprocessing sub-data set; wherein the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected; the clustering unit is used for clustering the preprocessed sub-data sets to obtain a data cluster; the estimating unit is used for inputting the data cluster into a material estimating model so as to output the estimated material usage amount required by the target area to run the power distribution network construction material project; the material prediction model is a decision model for predicting the material usage amount, which is obtained by training sample material data for multiple times.

According to yet another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium having a computer program stored therein, wherein the computer program is configured to execute the above-described power distribution network material demand prediction method when running.

According to still another aspect of the embodiment of the present invention, there is also provided an electronic device including a memory and a processor, where the memory stores a computer program, and the processor is configured to execute the power distribution network material demand prediction method described above through the computer program.

In the embodiment of the invention, the pretreatment data set of the distribution network materials of the target area is obtained by preprocessing the use recording parameters, the distribution network history planning data and the distribution network area economic development data of the historical distribution network materials of the target area; dividing the preprocessing data set to obtain preprocessing sub-data sets; wherein the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected; clustering the preprocessed sub-data sets to obtain a data cluster; the data cluster is input into a material prediction model to output the predicted material usage amount mode required by the target area to run the power distribution network construction material project, so that the purpose of reducing the actual error between the distribution network material prediction result and the time is achieved, the accuracy of the distribution network material prediction result is improved, the prediction time is shortened, the work efficiency of a material management department is improved, and the technical problem of large subjective judgment error caused by the traditional distribution network material manual prediction is solved.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:

FIG. 1 is a schematic illustration of an alternative power distribution network material demand prediction method application environment according to an embodiment of the present invention;

FIG. 2 is a flow chart of an alternative method for predicting demand for materials of a power distribution network according to an embodiment of the present invention;

FIG. 3 is a flow chart of an alternative method for predicting demand for materials of a power distribution network according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of an alternative power distribution network material demand prediction apparatus according to an embodiment of the present invention;

fig. 5 is a schematic structural view of an alternative electronic device according to an embodiment of the present invention.

Detailed Description

In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

According to an aspect of the embodiment of the present invention, there is provided a method for predicting a demand for a power distribution network, optionally, as an optional implementation manner, the method for predicting a demand for a power distribution network may be applied, but not limited to, to an environment as shown in fig. 1.

In fig. 1, the terminal device 104 is responsible for human-computer interaction with the user 102, and the terminal device 104 includes a memory 106, a processor 108 and a display 110; terminal device 104 may interact with server 114 through network 112. Server 114 includes a database 116 and a processing engine 118; the terminal device 104 may send, to the terminal device 104, the usage record parameters of the historical distribution network materials, the historical planning data of the distribution network, and the economic development data of the distribution network region to the server 114 through the network 112, where the server 114 outputs the estimated material usage required by the target region to operate the power distribution network construction material project.

Alternatively, in the present embodiment, the terminal device 104 may be a terminal device configured with a target client, and may include, but is not limited to, at least one of the following: a mobile phone (e.g., an Android mobile phone, iOS mobile phone, etc.), a notebook computer, a tablet computer, a palm computer, a MID (Mobile Internet Devices, mobile internet device), a PAD, a desktop computer, a smart television, etc. The target client may be a video client, an instant messaging client, a browser client, an educational client, and the like. The network 112 may include, but is not limited to: a wired network, a wireless network, wherein the wired network comprises: local area networks, metropolitan area networks, and wide area networks, the wireless network comprising: bluetooth, WIFI, and other networks that enable wireless communications. The server 114 may be a single server, a server cluster including a plurality of servers, or a cloud server. The above is merely an example, and is not limited in any way in the present embodiment.

The construction and maintenance of the distribution network are one of important works of the power system, the related engineering of the distribution network is complex, a large amount of materials are consumed in the implementation, the traditional distribution network material demand method adopts a manual mode to predict, but the prediction result error is larger, and the distribution network material demand prediction method using mathematical modeling appears in the related technology. The method comprises the steps of screening and preprocessing historical data related to the distribution network material demands, guiding the data into a mathematical model for training, and then using the model after training for prediction of the distribution network material demands. However, these methods still have a large room for improvement in practical applications. In the aspect of a prediction algorithm, a prediction model with excellent effect can be realized through a neural network, but a large amount of preprocessing data and long training time are required, and the data amount and application conditions in actual engineering may not be allowed; the autoregressive moving average (Autoregressive Integrated Moving Average Model, ARIMA) algorithm has high data quality requirement, can be essentially only fit with a linear model, and cannot well capture the nonlinear relation between the material demand and the influencing factors. In terms of data, the method generally only uses the data of the material consumption of the power grid company, other influencing factors are not considered, and the accuracy of estimating the material demand of the power distribution network is not high.

Based on the above technical problem, optionally, as an optional implementation manner, as shown in fig. 2, the method for predicting the demand of the electric power distribution network material includes:

s202, preprocessing the use record parameters, the distribution network history planning data and the distribution network regional economic development data of the historical distribution network materials of the target region to obtain a preprocessing data set of the distribution network materials of the target region;

s204, dividing the preprocessing data set to obtain preprocessing sub-data sets; the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected;

s206, clustering the preprocessed sub-data sets to obtain a data cluster;

s208, inputting the data cluster into a material prediction model to output the predicted material usage amount required by the power distribution network construction material project of the target area operation; the material prediction model is a decision model for predicting the material usage amount, which is obtained by training sample material data for multiple times.

In step S202, in practical application, but not limited to, the historical distribution network material usage recording parameters of the grid company in the target area and the historical annual investment planning data of the grid may be used as model training data, and the historical annual main economic development index of the area covered by the distribution network is added as feature data, so that the prediction result of the power configuration material is more accurate. Because the amount of distribution network supplies is directly related to the grid planning, the grid planning is affected by local economic development and population. Therefore, the investment planning data, the economic development index and the population index of the power grid are important influencing factors for the prediction of the demand of the distribution network materials, and are all used as influencing data for the prediction of the demand of the distribution network materials. Therefore, not only is the historical distribution network material use record of the power grid company and the annual investment planning data of the power grid adopted as model training data, so that the access prediction data are more sufficient, but also the annual main economic development index of the area covered by the distribution network is added as characteristic data, so that the prediction result is more accurate.

Here, the historical distribution network material usage recording parameters, the power grid annual investment planning data and the distribution network regional economic development data may be in the form of a data table or in the form of an XLM format document, which is not limited herein. Preprocessing the historical distribution network material use recording parameters, the power grid annual investment planning data and the distribution network regional economic development data to obtain a preprocessing data set of the distribution network material of the target region; wherein, obtaining the preprocessing data set includes sorting the historical distribution network material usage recording parameters, the power grid annual investment planning data and the distribution network regional economic development data according to the date to form a data table, wherein the data table can include, but is not limited to, an EXCLE data table. For the missing and abnormal situations of the data in the data table, checking and processing are needed, wherein the missing data can be filled by adopting a mean method, or a method for filling the position from the median to the missing data can be adopted, and for the data spike phenomenon occurring in the data table, the data smoothing processing is carried out, so that the pretreatment data set of the distribution network material of the target area is further obtained.

In step S204, during actual application, the preprocessing dataset is divided, so as to obtain a preprocessing sub dataset; the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected; here, the attribute of the power distribution network construction material item may be an item type, an item name, an item department, a type of a material for delivery, a name of a material for delivery, or the like of the power construction, which is not limited herein. That is, for example, the preprocessing data set may be divided according to the above-described different attributes, so as to obtain a plurality of EXCLE data tables of project types, project names, project departments, ex-warehouse material types, and ex-warehouse materials as header types for power construction.

In step S206, in practical application, the preprocessed sub-data sets are clustered to obtain data clusters, for example, a project type, a project name, a project department, a material type for ex warehouse, and a plurality of EXCLE data tables with a header type for ex warehouse materials for power construction. And clustering the data items according to cosine distances among the table head items serving as distance measurement standards to form a plurality of clustering clusters, so as to obtain the data clustering clusters such as the attribute of the power distribution network construction material items with the highest similarity.

In step S208, in practical application, the material prediction model may use, but is not limited to, an extreme gradient lifting (eXtreme Gradient Boosting, XGBoost) prediction model, and the XGBoost prediction model has the advantages of fast training speed, low data quality requirement, good nonlinear factor capturing performance, capability of significantly reducing data preprocessing complexity, and obvious advantages in practical engineering. The XGBoost prediction model is a tree integration model, and the operation mode is that CART trees with the total quantity of K in the model are used for predicting the same data set, and the prediction results are summed to be used as final prediction values. Namely:

in the method, in the process of the invention,for data set x _i A corresponding prediction result; f (f) _k Is the predictive model of the kth tree.

In the XGBoost prediction model training process, each tree model adopts an objective function in the same form as a model precision evaluation index:

in the method, in the process of the invention,representing fitting precision of the model to the training set as a loss function; />As a function of complexity, too high a model complexity will lead to an overfitting phenomenon.

The square loss function may be used as the loss function in this embodiment, as follows:

for a single CART tree in the XGBoost prediction model, its complexity function is defined as:

wherein T is the total number of leaf nodes of the model, omega ² L2 norms of weight vectors of all leaf nodes; gamma and lambda are used as adjustable penalty coefficients to adjust the complexity of the XGBoost predictive model.

Optionally, but not limited to, a preprocessed data set may be used as a training set, and the partitioned preprocessed data subsets all include complete time, scheduling information of power material subclasses, and characteristic information such as economic development indexes of each area, and the XGBoost prediction model is used to train the material demand prediction model. And model parameter adjustment adopts a common grid searching mode of machine learning, parameters are adjusted item by item, and finally the optimal parameter combination of the electric power materials is determined.

In addition, the performance of the XGBoost prediction model can be judged. When the model training reaches the optimal or set requirement, taking the model parameter with the optimal objective function value as a final result; and otherwise, adjusting and retraining the model parameters. After the XGBoost prediction model is successfully trained, each parameter is stored in a server database to serve as a backup, and the XGBoost prediction model can be called by a web end or other applications. After a period of regular operation, a rolling updated training data set is read from a server database, and the regression prediction model is retrained so as to ensure the real-time performance of the XGBoost prediction model calculation data.

According to the historical demand conditions of the distribution network materials and the annual planning data of the power grid, the main economic development index of the distribution network coverage area in the annual is combined as a characteristic, the XGBoost prediction model is adopted to integrate a learning algorithm for parallel training, a high-precision power distribution network material demand prediction model is established, the power grid material departments are assisted to support material purchasing and material scheduling work, material shortage and material backlog risks are reduced, and enterprise funds are saved.

In one embodiment, step S202 includes: the acquired use record parameters, distribution network history planning data and distribution network regional economic development data of the historical distribution network materials are respectively sequenced according to the generation date, and then summarized to obtain a distribution network material data summary table; according to a preset data processing method, processing abnormal data in the distribution network material data table to obtain a preprocessing data set. Here, the date span is freely defined as the number of days, the number of months, the number of years, etc., according to the need, and is not limited herein.

In an embodiment, according to a preset data processing method, processing abnormal data in a distribution network material data table to obtain a preprocessed data set includes: under the condition that the abnormal data indicate missing data, filling an average value of non-missing data consistent with the data type of the missing data in an idle position where the missing data is located in a distribution network material data table, and obtaining a preprocessing data set; or under the condition that the abnormal data indicate missing data, filling the median of the non-missing data consistent with the data type of the missing data in the idle position of the missing data in the distribution network material data table, so as to obtain a preprocessing data set. By the means, the accuracy and the completeness of data acquisition can be improved.

In an embodiment, according to a preset data processing method, processing abnormal data in a distribution network material data table to obtain a preprocessing data set of the distribution network material includes: and under the condition that the abnormal data indicate missing data, zero padding processing is carried out in the idle position where the missing data are located in the distribution network material data table, so as to obtain a preprocessing data set. By the means, the data acquisition integrity can be improved.

In an embodiment, according to a preset data processing method, processing abnormal data in a distribution network material data table to obtain a preprocessed data set includes: under the condition that peak appears in the database record data of the distribution network material data table, smoothing the database record data of the peak position to obtain a preprocessing data set; the use record parameters of the historical distribution network materials comprise database exit record data. The database-exiting record data in the distribution network material data summary table may have peak phenomenon, and the common form is abnormal data with a plurality of orders of magnitude exceeding the normal level, so that the peak data is very unfavorable for model training and needs to be processed.

Optionally, the smoothing processing of the database record data of the spike phenomenon in the distribution network material data summary table includes: calculating the difference value of daily delivery and withdrawal records in the distribution network material data table to obtain a daily net delivery value; calculating an average value P of peak values of daily net delivery in the current natural month and an average value N of daily normal net delivery in the current natural month in the distribution network material data summary table; when P/N is less than or equal to 1.5, replacing each peak value in the daily net warehouse-out amount in the current natural month with the average value P; when P/N is more than 1.5, calculating an average value P ' of peak values of daily net delivery in a quarter where the current natural month is located and an average value N ' of daily normal net delivery in the quarter, and when P '/N ' is less than or equal to 1.5, replacing each peak value in the daily net delivery in the current natural month with the average value P '; when P '/N' > 1.5, each peak in the daily net inventory in the current natural month is deleted. By the means, the accuracy and the completeness of data acquisition can be improved.

In one embodiment, step S206 includes converting attribute information of a to-be-tested power distribution network construction material item in the preprocessed data set into a vector by adopting a one-bit efficient coding mode; the attribute information of the power distribution network construction material items to be tested comprises item types, item names and item departments; taking the vectorized attribute information of the to-be-detected power distribution network construction material project as the data item characteristic in the preprocessing data set; acquiring cosine distance matrixes between any two data item characteristics; based on the cosine distance matrix, clustering data items corresponding to the data item features to obtain a plurality of data clustering clusters corresponding to the data item features; each data cluster comprises an array set of data items corresponding to the data item characteristics.

In one embodiment, step S208 includes, before the step, obtaining a plurality of historical sample usage amounts generated in a target period of time; the historical sample usage amount comprises usage recording parameters of historical distribution network materials, distribution network historical planning data and distribution network regional economic development data; and training the initialized material pre-estimation model by using the plurality of historical sample usage amounts until obtaining the material pre-estimation model with the training result reaching the convergence condition.

In one embodiment, step S208 includes: constructing a characteristic information table taking data items as units according to the material types to be predicted and the corresponding time of the material types; dividing the characteristic information table to obtain a preprocessing sub-data set; the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected; clustering the preprocessed sub-data sets to obtain a data cluster; and inputting the data cluster into a material pre-estimation model to output the material usage amount corresponding to the characteristic information table.

It should be noted that, for simplicity of description, the foregoing method embodiments are all described as a series of acts, but it should be understood by those skilled in the art that the present invention is not limited by the order of acts described, as some steps may be performed in other orders or concurrently in accordance with the present invention. Further, those skilled in the art will also appreciate that the embodiments described in the specification are all preferred embodiments, and that the acts and modules referred to are not necessarily required for the present invention.

Based on the above embodiment, in an application embodiment, as shown in fig. 3, the above method for predicting the material demand of the power distribution network may include the following steps:

step S302, data access; the distribution network material use history (out-of-stock record), the power grid planning data and the regional economy development index are imported into a server database from a data source as history data. Distribution network material usage history and grid planning data are provided by an enterprise resource planning (Enterprise Resource Planning, abbreviated ERP) system of a grid company, and regional economic development indexes are collected from third party regional development data sources.

Step S304, data preprocessing; (1) Each item of history data in the target area is ordered according to date and then corresponds to each other one by one to form a data summary table D _ori Date spans are freely defined as times of day, times of month, times of year, etc. according to requirements. The actual material delivery quantity on each date is used as material demand quantity information, and the other items are used as characteristic information.

For the data missing and abnormal conditions in the table, the checking and processing are needed, and the processing steps are as follows:

(1) for the large-area continuous data loss in the table, the actual material mobilization condition of the power grid company in the corresponding time period needs to be investigated to judge whether the actual data is lost or the record is not caused by no material mobilization in the corresponding time period. For the loss of the authenticity data, the data needs to be recovered as much as possible, the missing part is complemented, and if large-area data is still missing after the completion, the row of the missing data is deleted, so that adverse effects on model training are avoided. And if no record exists, performing 0 supplementing processing.

(2) For the small-range jumping data missing in the table, common data filling methods (such as mean filling, median and the like) can be adopted to process the missing data. The actual indication that the total amount of the materials of the power grid company discharged and returned to the warehouse has certain regularity on a time scale, and the actual filling can be carried out according to the movement condition of the materials.

(3) The database record data can have peak phenomenon, and the common form is abnormal data with a plurality of orders of magnitude above normal level, and the peak data is very unfavorable for model training and needs to be processed. The peak data in the database record is smoothed by adopting the following mode:

1. calculating the difference of the warehouse-out records in the table to obtain daily net warehouse-out amount information;

2. calculating average peak value of daily net output in month _ave And average norm of the normal daily net inventory for the month _ave ；

3. When (when)At this time, peak can be used _ave Filling each peak value in the daily net warehouse-out amount of the month;

when (when)At that time, the average peak of the daily net-output peak data at the quarter of the month is recalculated _ave Average norm of the total daily net output for the season and other norms _ave And (5) judging again.

4. If the peak data cannot be eliminated by adopting the modes 2 and 3, filling or direct deleting can be carried out by adopting other common methods, so that the interference caused by model training is reduced. The processed data summary table is not subjected to overlay storage and is independently stored as a new data summary table D _tre 。

(3) The XGBoost algorithm uses a CART tree as a basic unit, does not need to normalize or standardize input data, and does not need to vectorize other non-digital features such as texts. In the invention, a one-bit effective one-hot coding mode is adopted to make the data summary table D _tre Non-numeric features such as item type and item name in (a) are converted into vectors.

Step S306, in this embodiment, the data set is divided according to the following steps:

(1) Taking the attribute information of vectorized project types, project names, project departments and other keywords and other power distribution network construction material projects as the characteristics of each data item to calculate a cosine distance matrix L between each data item _tre ；

(3) Clustering each data item by using a DB-Scan analysis algorithm and L as a distance measurement standard to form a plurality of cluster clusters, and recording a cluster center coordinate record C ₁ ，C ₂ ，C ₃ …. The data items in the same cluster are divided into data sets X ₁ ，X ₂ ，X ₃ …, and corresponding labels are added.

(3) Through the steps of D _tre Set-by-set partitioning into D _tra In the form of a format D _tra ＝{X ₁ ，X ₂ ，X ₃ …X _m Data items in each dataset have information such as the most similar item names and item types.

(4) Each dataset is then identified by the material subclass key, D _tra Each X in (a) _i Further dividing to form X _i1 ，X _i2 …X _is (i= … m). Finally divide the original data set into D _tra ＝{[X ₁₁ ，X ₁₂ …X _1s ]，[X ₂₁ ，X ₂₁ …X _2s ]，[X ₃₁ ，X ₃₂ …X ₃₂ …X _3s ]…[X _m1 …X _ms ]}。

Each data subset X _ij Separately training regression prediction model M _ij Parallel processing of large data sets is achieved, and prediction accuracy of a model aiming at specific material subclasses is enhanced. Data set D _tra Stored in a server database, and periodically updated by scrolling.

Step S308, model training; d is used in the present invention _tra As training set, partitioned data subset X _ij The method comprises the steps of including complete time, material subclass scheduling information, economic development indexes of various areas and other characteristic information, and training a material demand prediction model by using an XGBoost algorithm. The model parameter adjustment adopts a common grid searching mode of machine learning, parameters are adjusted item by item, and finally the optimal parameter combination is determined.

Step S310, judging whether the requirement is met; when the training of the material demand prediction model reaches the optimal or set requirement, taking the model parameter with the optimal objective function value as a final result; if the optimal model parameters of the objective function values are not reached, the process proceeds to step S308, where the material demand prediction model parameters are adjusted and retrained. After the material demand prediction model is successfully trained, each parameter is stored in a server database to serve as a backup, and the model can be called by a web end or other applications. After a period of regular operation, the rolling updated training data set D is read from the server database _tra Performing regression prediction modelRetraining to ensure real-time performance of the model.

Step S312, predicting distribution network material demands; (1) According to the material types and the corresponding time to be predicted, a complete characteristic information table is constructed by taking data items as units; (2) Preprocessing the characteristic information table in the second step to form D _tes Calculate D _tes Each data entry of D _tra Each cluster center C stored in _i Cosine distance table L of (2) _tes The method comprises the steps of carrying out a first treatment on the surface of the (3) Classifying the characteristic information according to the maximum cosine distance principle, and respectively transmitting the characteristic information to a corresponding material subclass prediction model M _ij And performing prediction calculation. For a specific material subclass a, the demand prediction result is corresponding to the prediction model [ M ] of each material subclass _1a ，M _2a …M _ma ]And the sum of the predicted results of (a) is used.

In the embodiment of the invention, the pretreatment data set of the distribution network materials of the target area is obtained by preprocessing the use recording parameters, the distribution network history planning data and the distribution network area economic development data of the historical distribution network materials of the target area; dividing the preprocessing data set to obtain preprocessing sub-data sets; wherein the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected; clustering the preprocessed sub-data sets to obtain a data cluster; the data cluster is input into a material prediction model to output the predicted material usage amount mode required by the target area to run the power distribution network construction material project, so that the purpose of reducing the prediction result and the actual error of the distribution network material is achieved, the accuracy of the distribution network material prediction result is improved, the prediction time is shortened, the work efficiency of a material management department is improved, and the technical problem of large subjective judgment error caused by the traditional distribution network material manual prediction is solved.

According to another aspect of the embodiment of the invention, a power distribution network material demand prediction device for implementing the power distribution network material demand prediction method is also provided. As shown in fig. 4, the apparatus includes:

a preprocessing unit 402, configured to preprocess usage record parameters of historical distribution network materials in a target area, distribution network historical planning data and distribution network area economic development data, so as to obtain a preprocessing dataset of the distribution network materials in the target area;

a dividing unit 404, configured to divide the preprocessed data set to obtain a preprocessed sub-data set; the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected;

a clustering unit 406, configured to cluster the preprocessed sub-data set to obtain a data cluster;

the estimating unit 408 is configured to input the data cluster into a material estimating model, so as to output an estimated material usage amount required by the power distribution network construction material project running in the target area; the material prediction model is a decision model for predicting the material usage amount, which is obtained by training sample material data for multiple times.

In the embodiment of the invention, but not limited to, historical distribution network material use recording parameters of a power grid company in a target area and power grid annual investment planning data are adopted as model training data, and an annual main economic development index of an area covered by the distribution network is added as characteristic data, so that a power configuration material prediction result is more accurate. Because the amount of distribution network supplies is directly related to the grid planning, the grid planning is affected by local economic development and population. Therefore, the investment planning data, the economic development index and the population index of the power grid are important influencing factors for the prediction of the demand of the distribution network materials, and are all used as influencing data for the prediction of the demand of the distribution network materials. Therefore, not only is the historical distribution network material use record of the power grid company and the annual investment planning data of the power grid adopted as model training data, so that the access prediction data are more sufficient, but also the annual main economic development index of the area covered by the distribution network is added as characteristic data, so that the prediction result is more accurate.

In the embodiment of the invention, the preprocessing data set is divided to obtain preprocessing sub-data sets; the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected; here, the attribute of the power distribution network construction material item may be an item type, an item name, an item department, a type of a material for delivery, a name of a material for delivery, or the like of the power construction, which is not limited herein. That is, for example, the preprocessing data set may be divided according to the above-described different attributes, so as to obtain a plurality of EXCLE data tables of project types, project names, project departments, ex-warehouse material types, and ex-warehouse materials as header types for power construction.

In the embodiment of the invention, the preprocessing sub-data sets are clustered to obtain data clusters, for example, a plurality of EXCLE data tables with the project types, project names, project departments, ex-warehouse material types and ex-warehouse materials of the header types of the power construction are clustered according to cosine distances among the header entries as distance measurement standards to form a plurality of clusters, and the data clusters such as the attributes of the power distribution network construction material projects with the highest similarity are obtained.

In the embodiment of the invention, the material prediction model can be used, but is not limited to an extreme gradient lifting (eXtreme Gradient Boosting, XGBoost) prediction model, the XGBoost prediction model has the advantages of high training speed, low data quality requirement, good nonlinear factor capturing performance, capability of remarkably reducing the complexity of data preprocessing and obvious advantages in practical engineering. XGBoost is a tree integrated model, and the operation mode is that CART trees with the total quantity of K in the model are used for predicting the same data set, and the prediction results are summed to be used as final prediction values. Namely:

In the XGBoost model training process, each tree model adopts an objective function in the same form as a model precision evaluation index:

for a single CART tree in the XGBoost model, its complexity function is defined as:

wherein T is the total number of leaf nodes of the model, omega ² Is the L2 norm of the weight vector for each leaf node. Gamma and lambda are used as adjustable penalty coefficients to adjust the complexity of the XGBoost model.

Optionally, but not limited to, a preprocessed data set is used as a training set, and the partitioned preprocessed data subsets all contain complete time, power material subclass scheduling information, economic development indexes of various areas and other characteristic information, and a material demand prediction model is trained by using an XGBoost algorithm. And model parameter adjustment adopts a common grid searching mode of machine learning, parameters are adjusted item by item, and finally the optimal parameter combination of the electric power materials is determined.

In addition, the performance of the XGBoost model can be judged. When the model training reaches the optimal or set requirement, taking the model parameter with the optimal objective function value as a final result; and otherwise, adjusting and retraining the model parameters. After the XGBoost model is successfully trained, each parameter is stored in a server database to serve as a backup, and the XGBoost model can be called by a web end or other applications to use. After a period of regular operation, a rolling updated training data set is read from a server database, and the regression prediction model is retrained so as to ensure the real-time performance of the XGBoost model calculation data.

According to the historical demand conditions of the distribution network materials and the annual planning data of the power grid, the XGBoost integrated learning algorithm is adopted to perform parallel training by combining the annual main economic development index of the coverage area of the distribution network as a characteristic, a high-precision power distribution network material demand prediction model is established, the power grid material departments are assisted to perform material purchasing and material scheduling work, material shortage and material backlog risks are reduced, and enterprise funds are saved.

According to still another aspect of the embodiment of the present invention, there is further provided an electronic device for implementing the above method for predicting a demand for materials of a power distribution network, where the electronic device may be the terminal device 104 or the server 114 shown in fig. 1. As shown in fig. 5, the electronic device comprises a memory 502 and a processor 504, the memory 502 having stored therein a computer program, the processor 504 being arranged to perform the steps of any of the method embodiments described above by means of the computer program.

Alternatively, in this embodiment, the electronic device may be located in at least one network device of a plurality of network devices of the computer network.

Alternatively, in the present embodiment, the above-described processor may be configured to execute the following steps by a computer program:

S1, preprocessing the use record parameters, the distribution network history planning data and the distribution network regional economic development data of the historical distribution network materials of a target region to obtain a preprocessing data set of the distribution network materials of the target region;

s2, dividing the preprocessed data set to obtain preprocessed sub-data sets; the preprocessing sub-data set comprises attribute information of a power distribution network construction material item to be detected;

s3, clustering the preprocessed sub-data sets to obtain a data cluster;

s4, inputting the data cluster into a material prediction model to output the predicted material usage amount required by the power distribution network construction material project of the target area operation; the material prediction model is a decision model for predicting the material usage amount, which is obtained by training sample material data for multiple times.

Alternatively, it will be understood by those skilled in the art that the structure shown in fig. 5 is only schematic, and the electronic device or apparatus may be a smart phone (such as an Android phone, iOS phone, etc.), a tablet computer, a palmtop computer, a mobile internet device (Mobile Internet Devices, MID), a PAD, etc. Fig. 5 is not limited to the structure of the electronic device and the electronic apparatus described above. For example, the electronics can also include more or fewer components (e.g., network interfaces, etc.) than shown in fig. 5, or have a different configuration than shown in fig. 5.

The memory 502 may be used to store software programs and modules, such as program instructions/modules corresponding to the power distribution network material demand prediction method and apparatus in the embodiment of the present invention, and the processor 504 executes the software programs and modules stored in the memory 502 to perform various functional applications and data processing, that is, implement the power distribution network material demand prediction method described above. Memory 502 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 502 may further include memory located remotely from processor 504, which may be connected to the terminal via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof. The memory 502 may be used to store information such as attribute information of virtual power distribution network construction material items, but is not limited to this. As an example, as shown in fig. 5, the memory 502 may include, but is not limited to, the preprocessing unit 402, the dividing unit 404, the clustering unit 406, and the estimating unit 408 in the power distribution network material demand prediction device. In addition, other module units in the power distribution network material demand prediction device may be included, but are not limited to, and are not described in detail in this example.

Optionally, the transmission device 506 is configured to receive or transmit data via a network. Specific examples of the network described above may include wired networks and wireless networks. In one example, the transmission device 506 includes a network adapter (Network Interface Controller, NIC) that may be connected to other network devices and routers via a network cable to communicate with the internet or a local area network. In one example, the transmission device 506 is a Radio Frequency (RF) module, which is used to communicate with the internet wirelessly.

In addition, the electronic device further includes: the display 508 is used for displaying the attribute information of the power distribution network construction material items; and a connection bus 510 for connecting the respective module parts in the above-described electronic device.

In other embodiments, the terminal device or the server may be a node in a distributed system, where the distributed system may be a blockchain system, and the blockchain system may be a distributed system formed by connecting the plurality of nodes through a network communication. Among them, the nodes may form a Peer-To-Peer (P2P) network, and any type of computing device, such as a server, a terminal, etc., may become a node in the blockchain system by joining the Peer-To-Peer network.

According to a further aspect of embodiments of the present invention, there is also provided a computer readable storage medium having a computer program stored therein, wherein the computer program is arranged to perform the steps of any of the method embodiments described above when run.

Alternatively, in the present embodiment, the above-described computer-readable storage medium may be configured to store a computer program for executing the steps of:

s3, clustering the preprocessed sub-data sets to obtain a data cluster;

Alternatively, in this embodiment, it will be understood by those skilled in the art that all or part of the steps in the methods of the above embodiments may be performed by a program for instructing a terminal device to execute the steps, where the program may be stored in a computer readable storage medium, and the storage medium may include: flash disk, read-Only Memory (ROM), random-access Memory (Random Access Memory, RAM), magnetic or optical disk, and the like.

The foregoing embodiment numbers of the present invention are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

The integrated units in the above embodiments may be stored in the above-described computer-readable storage medium if implemented in the form of software functional units and sold or used as separate products. Based on such understanding, the technical solution of the present invention may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, comprising several instructions for causing one or more computer devices (which may be personal computers, servers or network devices, etc.) to perform all or part of the steps of the method of the various embodiments of the present invention.

In the foregoing embodiments of the present invention, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.

In several embodiments provided in the present application, it should be understood that the disclosed client may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and are merely a logical functional division, and there may be other manners of dividing the apparatus in actual implementation, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.

Claims

1. The method for predicting the material demand of the power distribution network is characterized by comprising the following steps of:

acquiring a plurality of historical sample usage amounts generated in a target time period; the historical sample usage amount comprises usage recording parameters of historical distribution network materials of a target area, distribution network historical planning data and distribution network area economic development data;

preprocessing the use record parameters of the historical distribution network materials of the target area, the distribution network historical planning data and the distribution network area economic development data to obtain a preprocessing data set of the distribution network materials of the target area;

Dividing the preprocessing data set to obtain preprocessing sub-data sets of historical distribution network materials; wherein, the preprocessing sub-data set of the historical distribution network materials comprises attribute information of the power distribution network construction material items to be tested;

training the initialized material pre-estimation model by utilizing the preprocessing sub-data set of the historical distribution network material until obtaining a material pre-estimation model with a training result reaching a convergence condition;

constructing a characteristic information table taking data items as units according to the types of the materials to be predicted and the corresponding time of the types of the materials;

dividing the characteristic information table to obtain a preprocessing sub-data set of the materials to be predicted; wherein the pretreatment sub-data set of the materials to be predicted comprises attribute information of the power distribution network construction material items to be predicted;

converting attribute information of the to-be-detected power distribution network construction material items in the pretreatment sub-data set of the to-be-predicted material into vectors by adopting a one-bit effective coding mode; the attribute information of the power distribution network construction material items to be tested comprises an item type, an item name and an item department;

taking the vectorized attribute information of the power distribution network construction material items to be detected as the data item characteristics in the preprocessing sub-data set of the materials to be predicted;

Acquiring cosine distance matrixes between any two data item characteristics;

based on the cosine distance matrix, clustering the data items corresponding to the data item features to obtain a plurality of data clustering clusters corresponding to the data item features; each data cluster comprises an array set of data items corresponding to the data item characteristics;

and inputting the data cluster into the material pre-estimation model to output the material usage amount corresponding to the characteristic information table.

2. The method of claim 1, wherein preprocessing the usage record parameters of the historical distribution network materials of the target area, the distribution network historical planning data and the distribution network area economic development data to obtain a preprocessing dataset of the distribution network materials of the target area comprises:

the acquired use record parameters of the historical distribution network materials, the distribution network historical planning data and the distribution network regional economic development data are respectively sequenced according to the generation date, and then summarized to obtain a distribution network material data summary table;

and processing the abnormal data in the distribution network material data summary table according to a preset data processing method to obtain the preprocessing data set.

3. The method according to claim 2, wherein the processing the abnormal data in the distribution network material data table according to a preset data processing method to obtain the preprocessed data set includes:

under the condition that the abnormal data indicate missing data, filling an average value of non-missing data consistent with the data type of the missing data in an idle position where the missing data are located in the distribution network material data table, and obtaining the preprocessing data set; or alternatively

And under the condition that the abnormal data indicate missing data, filling the median of non-missing data with the same data type as the missing data in the idle position of the missing data in the distribution network material data table, so as to obtain the preprocessing data set.

4. The method according to claim 2, wherein the processing the abnormal data in the distribution network material data table according to the preset data processing method to obtain the preprocessing data set of the distribution network material includes:

and under the condition that the abnormal data indicate missing data, zero padding processing is carried out in an idle position where the missing data are located in the distribution network material data table, so that the preprocessing data set is obtained.

5. The method according to claim 2, wherein the processing the abnormal data in the distribution network material data table according to a preset data processing method to obtain the preprocessed data set includes:

under the condition that peak appears in the database record data of the distribution network material data table, smoothing the database record data of the peak position to obtain the preprocessing data set; the usage record parameters of the historical distribution network materials comprise the database exit record data.

6. The utility model provides a power distribution network material demand prediction device which characterized in that includes:

the preprocessing unit is used for preprocessing the use record parameters, the distribution network history planning data and the distribution network regional economic development data of the historical distribution network materials of the target region to obtain a preprocessing data set of the distribution network materials of the target region;

the dividing unit is used for dividing the preprocessing data set to obtain a preprocessing sub-data set of the historical distribution network material; wherein, the preprocessing sub-data set of the historical distribution network materials comprises attribute information of the power distribution network construction material items to be tested;

the clustering unit is used for converting the attribute information of the to-be-detected power distribution network construction material items in the pretreatment sub-data set of the materials to be predicted into vectors by adopting a one-bit effective coding mode; the attribute information of the power distribution network construction material items to be tested comprises an item type, an item name and an item department; taking the vectorized attribute information of the power distribution network construction material items to be detected as the data item characteristics in the preprocessing sub-data set of the materials to be predicted; acquiring cosine distance matrixes between any two data item characteristics; based on the cosine distance matrix, clustering the data items corresponding to the data item features to obtain a plurality of data clustering clusters corresponding to the data item features; each data cluster comprises an array set of data items corresponding to the data item characteristics;

The estimating unit is used for inputting the data cluster into a material estimating model so as to output the material usage amount corresponding to the characteristic information table

The device is further used for acquiring a plurality of historical sample usage amounts generated in a target time period before preprocessing the usage recording parameters of the historical distribution network materials of the target area, the distribution network historical planning data and the distribution network area economic development data; wherein the historical sample usage amount comprises the usage record parameters of the historical distribution network materials of the target area, the distribution network historical planning data and the distribution network area economic development data; before the attribute information of the to-be-detected power distribution network construction material items in the pretreatment sub-data set of the materials to be predicted is converted into vectors by adopting a one-bit effective coding mode, training an initialized material estimation model by utilizing the pretreatment sub-data set of the historical distribution network materials until the material estimation model with the training result reaching a convergence condition is obtained; constructing a characteristic information table taking data items as units according to the types of the materials to be predicted and the corresponding time of the types of the materials; dividing the characteristic information table to obtain a preprocessing sub-data set of the materials to be predicted; the pretreatment sub-data set of the materials to be predicted comprises attribute information of the power distribution network construction material items to be detected.

7. A computer-readable storage medium, characterized in that the computer-readable storage medium comprises a stored program, wherein the program when run performs the method of any one of claims 1 to 5.

8. An electronic device comprising a memory and a processor, characterized in that the memory has stored therein a computer program, the processor being arranged to execute the method according to any of claims 1 to 5 by means of the computer program.