CN116227738B - Method and system for predicting traffic interval of power grid customer service - Google Patents
Method and system for predicting traffic interval of power grid customer service Download PDFInfo
- Publication number
- CN116227738B CN116227738B CN202310483798.9A CN202310483798A CN116227738B CN 116227738 B CN116227738 B CN 116227738B CN 202310483798 A CN202310483798 A CN 202310483798A CN 116227738 B CN116227738 B CN 116227738B
- Authority
- CN
- China
- Prior art keywords
- data
- telephone traffic
- complete
- traffic
- influence factor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000000605 extraction Methods 0.000 claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 19
- 238000004364 calculation method Methods 0.000 claims description 17
- 230000001502 supplementing effect Effects 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 3
- 238000011002 quantification Methods 0.000 abstract description 6
- 230000006870 function Effects 0.000 description 34
- 230000009469 supplementation Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 230000001419 dependent effect Effects 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000015654 memory Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 108090000672 Annexin A5 Proteins 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Mathematical Optimization (AREA)
- Human Resources & Organizations (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Finance (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Biology (AREA)
- Accounting & Taxation (AREA)
- Algebra (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Entrepreneurship & Innovation (AREA)
- Quality & Reliability (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Telephonic Communication Services (AREA)
Abstract
The application discloses a method and a system for predicting a traffic interval of power grid customer service, which are used for carrying out similar daily clustering on the traffic influence factor data after the missing value is supplemented by acquiring historical traffic data and traffic influence factor data. After feature extraction is carried out on the processed data and the clustering result by using the prediction model to obtain feature data, a quantile regression model is adopted to calculate the feature data to obtain a condition quantile, a non-parameter kernel density estimation method is adopted to calculate the condition quantile to obtain an interval prediction result, and the interval prediction result is sent to a power grid customer service scheduling system so that the power grid customer service scheduling system carries out telephone traffic scheduling according to the interval prediction result. According to the method, the telephone traffic probability of the power grid customer service is predicted by combining telephone traffic influencing factors, so that the effective quantification of the uncertainty of the prediction of a modern power supply service system can be realized, and the accuracy of the prediction of the telephone traffic interval of the power grid customer service is improved.
Description
Technical Field
The application relates to the field of traffic interval prediction, in particular to a method and a system for predicting a power grid customer service traffic interval.
Background
With the continuous development of society, the electric power customer service center is continuously enlarged in scale, plays an increasingly important role as an important bridge for enterprises to communicate with clients, and the operation management mode of the electric power customer service center is also required to be continuously updated along with the social and economic development. The current telephone traffic prediction management application of the electric power customer service center mainly depends on historical experience, and has the problems of low scheduling efficiency, poor fitting degree between a required value and a scheduling person, large human resource input, unattainable service level and the like. The traditional telephone traffic prediction technology cannot adapt to the complex service scene containing the value-added service in the modern power supply service at present, so that the utilization of human resources of the telephone traffic service is unreasonable. At present, main traffic prediction models comprise inertial prediction, kalman filtering, traffic OLAP analysis and the like, and all belong to point prediction.
The interval prediction can realize the effective quantification of the prediction uncertainty of a modern power supply service system, and can provide more comprehensive prediction information compared with the classical prediction uncertainty, thereby providing key data support for analysis and decision-making of an electric power customer service system.
Disclosure of Invention
The application provides a method and a system for predicting a power grid customer service traffic interval, which can realize the effective quantification of the prediction uncertainty of a modern power supply service system and improve the accuracy of the power grid customer service traffic interval prediction.
In order to solve the above technical problems, an embodiment of the present application provides a method for predicting a traffic interval of a customer service of a power grid, including:
acquiring historical telephone traffic data and telephone traffic influence factor data, and supplementing missing values of the historical telephone traffic data and the telephone traffic influence factor data to obtain complete historical telephone traffic data and complete telephone traffic influence factor data;
performing similar daily clustering on the complete telephone traffic influence factor data to obtain a clustering result of the complete telephone traffic influence factor data;
inputting the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data into a constructed prediction model, carrying out feature extraction on the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data by the prediction model to obtain feature data, calculating the feature data by adopting a quantile regression model to obtain the conditional quantiles of different quantile points, calculating the conditional quantiles by adopting a non-parameter kernel density estimation method to obtain an interval prediction result, and sending the interval prediction result to a grid customer service scheduling system so that the grid customer service scheduling system carries out telephone traffic scheduling according to the interval prediction result.
According to the embodiment, historical telephone traffic data and telephone traffic influence factor data are obtained, missing value supplementation is carried out on the historical telephone traffic data and telephone traffic influence factor data, complete historical telephone traffic data and complete telephone traffic influence factor data are obtained, similar daily clustering is carried out on the complete telephone traffic influence factor data to obtain a clustering result of the complete telephone traffic influence factor data, the clustering result of the complete historical telephone traffic data and the complete telephone traffic influence factor data is input into a built prediction model, after feature extraction is carried out on the clustering result of the complete historical telephone traffic data and the complete telephone traffic influence factor data by the prediction model to obtain feature data, a quantile regression model is adopted to calculate the feature data to obtain a conditional quantile, then a non-parameter kernel density estimation method is adopted to calculate the conditional quantile to obtain an interval prediction result, and the interval prediction result is sent to a grid customer service scheduling system, so that the grid customer service scheduling system carries out telephone traffic scheduling according to the interval prediction result. According to the method, the telephone traffic probability of the power grid customer service is predicted by combining telephone traffic influencing factors, so that the effective quantification of the uncertainty of the prediction of a modern power supply service system can be realized, and the accuracy of the prediction of the telephone traffic interval of the power grid customer service is improved.
As a preferred scheme, missing value supplementation is carried out on historical telephone traffic data and telephone traffic influencing factor data to obtain complete historical telephone traffic data and complete telephone traffic influencing factor data, which concretely comprises the following steps:
obtaining an approximation of historical traffic data and an approximation of traffic impact factor data by fitting a polynomial using adjacent normal data;
and supplementing the approximate value as a missing value to the historical telephone traffic data and the telephone traffic influencing factor data to obtain complete historical telephone traffic data and complete telephone traffic influencing factor data.
According to the embodiment, the approximation value of the historical telephone traffic data and the approximation value of the telephone traffic influence factor data are obtained by utilizing the adjacent normal data fitting polynomials, the approximation value is used as a missing value to be supplemented into the historical telephone traffic data and the telephone traffic influence factor data, the complete historical telephone traffic data and the complete telephone traffic influence factor data are obtained, the missing value supplementation is carried out on the data, the abnormal value of the data is cleaned, the integrity and the accuracy of the data are guaranteed, and the prediction precision is improved.
As a preferred scheme, the complete telephone traffic influence factor data is clustered on a similar day to obtain a clustering result of the complete telephone traffic influence factor data, specifically:
after normalizing the complete telephone traffic influence factor data, selecting a plurality of center points;
and distributing the data of each complete telephone traffic influencing factor to a central point smaller than a preset distance to obtain a plurality of clusters, and recalculating the central point of each cluster to serve as a clustering result of the data of the complete telephone traffic influencing factor.
As a preferred scheme, the quantile regression model is adopted to calculate the characteristic data to obtain the conditional quantiles of different quantile points, and the method specifically comprises the following steps:
solving the characteristic data by using a quantile regression function to obtain regression coefficient vectors of different quantile points, wherein the quantile regression function is as follows:
wherein,representing dependent variablesP p Is the first of (2)τThe number of digits of the individual condition is calculated,τthe range of (1, 0),β(τ) Is a vector of regression coefficients, +.>Represented as feature data;
solving regression coefficient vectors of different quantile points to obtain conditional quantile of the different quantile points, wherein a solving formula is as follows:
wherein L is the neural network loss function value,represented as characteristic data>Expressed as an asymmetric function:
where s is the argument of the function γ.
As a preferred scheme, a non-parameter kernel density estimation method is adopted to calculate the conditional quantile to obtain an interval prediction result, and the method specifically comprises the following steps:
calculating the conditional quantiles by using a non-parameter kernel density estimation function to obtain an interval prediction result, wherein the kernel density estimation function is as follows:
wherein,V(u a ) For the probability distribution of the traffic on the a-th day to be predicted,u a as an independent variable of the probability function,N a for a similar number of days for the a moments to be predicted,hfor the bandwidth, takeh=0.01，K() As a kernel function, taking the kernel function as a Gaussian function,G P in the case of a vector of samples,G P =[,/>,/>…,/>]。
preferably, the prediction model is obtained through training, specifically:
dividing the clustering result of the complete historical telephone traffic data and the complete telephone traffic influence factor data into a training set and a testing set;
constructing a prediction model, and inputting the clustering result of the complete historical telephone traffic data and the complete telephone traffic influence factor data into the prediction model for calculation to obtain an actual interval prediction result;
and comparing the actual interval prediction result with the expected interval prediction result according to the error function to obtain an error, and updating model parameters according to the error until the training times are reached to obtain a trained prediction model.
In order to solve the same technical problems, the embodiment of the application also provides a system for predicting the traffic interval of the power grid customer service, which comprises a data acquisition module, a clustering module and a prediction interval calculation module,
the data acquisition module is used for acquiring historical telephone traffic data and telephone traffic influence factor data, and supplementing missing values of the historical telephone traffic data and the telephone traffic influence factor data to obtain complete historical telephone traffic data and complete telephone traffic influence factor data;
the clustering module is used for carrying out similar daily clustering on the complete telephone traffic influence factor data to obtain a clustering result of the complete telephone traffic influence factor data;
the prediction interval calculation module is used for inputting the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data into a constructed prediction model, so that after the prediction model performs feature extraction on the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data to obtain feature data, the feature data is calculated by adopting a quantile regression model to obtain the conditional quantiles of different quantiles, then the conditional quantiles are calculated by adopting a non-parameter kernel density estimation method to obtain an interval prediction result, and the interval prediction result is sent to the grid customer service scheduling system so that the grid customer service scheduling system performs telephone traffic scheduling according to the interval prediction result.
Preferably, the data acquisition module comprises an approximation calculation unit and a filling unit,
the approximate value calculating unit is used for obtaining the approximate value of the historical telephone traffic data and the approximate value of the telephone traffic influencing factor data by fitting a polynomial by using adjacent normal data;
the filling unit is used for supplementing the approximate value as a missing value to the historical telephone traffic data and the telephone traffic influencing factor data to obtain complete historical telephone traffic data and complete telephone traffic influencing factor data.
Preferably, the clustering module comprises a central point selecting unit and a clustering result generating unit,
the central point selecting unit is used for selecting a plurality of central points after normalizing the complete telephone traffic influencing factor data;
the clustering result generating unit is used for distributing the telephone traffic influence factor data of each complete telephone traffic influence factor data to a central point smaller than a preset distance to obtain a plurality of clusters and recalculate the central point of each cluster to be used as a clustering result of the complete telephone traffic influence factor data.
Drawings
Fig. 1: a flow diagram of one embodiment of the power grid customer service traffic interval prediction method provided by the application;
fig. 2: the telephone traffic similar day clustering result schematic diagram of one embodiment of the telephone traffic interval prediction method of the power grid customer service provided by the application;
fig. 3: the application provides a system structure schematic diagram of another embodiment of a power grid customer service traffic interval prediction method.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Example 1
Referring to fig. 1, in order to provide a method for predicting a traffic interval of a customer service in a power grid according to an embodiment of the present application, the method for predicting a traffic interval of a customer service in a power grid includes steps 101 to 104, where the steps are as follows:
step 101: and acquiring historical telephone traffic data and telephone traffic influence factor data, and supplementing missing values of the historical telephone traffic data and the telephone traffic influence factor data to obtain complete historical telephone traffic data and complete telephone traffic influence factor data.
Optionally, missing value supplementation is performed on historical traffic data and traffic volume influence factor data to obtain complete historical traffic data and complete traffic volume influence factor data, which specifically comprises:
obtaining an approximation of historical traffic data and an approximation of traffic impact factor data by fitting a polynomial using adjacent normal data;
and supplementing the approximate value as a missing value to the historical telephone traffic data and the telephone traffic influencing factor data to obtain complete historical telephone traffic data and complete telephone traffic influencing factor data.
In this embodiment, the historical traffic data is derived from the grid customer service system on an hour scale, two years, and one hour intervals. Then, the historical weather information, the historical power failure information and the holiday condition are led out from a weather bureau officer network, wherein the historical weather information specifically comprises rainfall (light rain, medium rain and heavy rain), snowfall (light snow, medium snow and heavy snow), wind power (1-9 grades and typhoons), temperature (high-temperature yellow early warning, high Wen Chengse early warning and high-temperature red early warning); the historical power outage information specifically comprises the power outage type (scheduled maintenance power outage and fault temporary power outage), the power outage duration and the number of users involved in the power outage range, and the holiday situation specifically comprises national legal holidays and weekends.
Then, a Lagrange interpolation method is adopted, and an approximation value is obtained by fitting a polynomial by using adjacent normal data to supplement the missing value, specifically:
missing value replenishment is performed by fitting polynomials with neighboring normal data to obtain approximations:
wherein,is the firstiDay missing data timing of the day, +.>Is the firstiData missing on day.
Interpolation polynomialThe construction process is as follows: taking the missing data for two yearsnData points>As a result ofnPersonal (S)n-polynomial of degree 1->：
Wherein,is the firstiData of day loss, ->Is the firstiData missing on day.
Finally obtaining interpolation functionyAccording to the interpolation function, the approximation value is obtained by calculation, and the expression of the interpolation function is as follows:
wherein,is the firstiData missing on day,/->Representingn-polynomial of degree 1.
Step 102: and carrying out similar daily clustering on the complete telephone traffic influence factor data to obtain a clustering result of the complete telephone traffic influence factor data.
Optionally, similar daily clustering is performed according to the traffic influencing factor data to obtain a traffic influencing factor data clustering result, which specifically comprises:
after the telephone traffic influencing factor data are standardized, a plurality of centers are randomly selected;
and carrying out iterative computation on the daily traffic volume influence factor data by using a preset loss function to obtain a plurality of clusters, and calculating the central point of each cluster.
In this embodiment, K-means is used to perform similar day clustering based on the data characteristics of traffic volume influencing factor variables.
First, normalization processing is performed on the original data:
wherein,for the original data +.>Is normalized data.
Randomly selectqCenter, and define a loss function, wherein,qthe centers are denoted asThe defined loss function is:
order theFor the number of iterative steps, the following procedure is repeated until convergence: sample->It is assigned to the center nearest to it +.>：
For the center of each class, the center point is recalculated, with the calculation formula:
wherein,representing sample data, ++>Representing a randomly selected center point.
Finally, obtaining a plurality of clusters and recalculating the central point of each cluster, as a clustering result of the complete traffic influencing factor data, wherein the clustering result is shown in figure 2, the approximation value of the historical traffic data and the approximation value of the traffic influencing factor data are obtained by utilizing the fitting polynomial of adjacent normal data, the approximation value is used as a missing value to be supplemented into the historical traffic data and the traffic influencing factor data, the complete historical traffic data and the complete traffic influencing factor data are obtained, the missing value supplementation is carried out on the data, and meanwhile, the abnormal value of the data is cleaned, so that the integrity and the accuracy of the data are guaranteed, and the prediction accuracy is improved.
Step 103: inputting the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data into a constructed prediction model, carrying out feature extraction on the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data by the prediction model to obtain feature data, calculating the feature data by adopting a quantile regression model to obtain the conditional quantiles of different quantile points, calculating the conditional quantiles by adopting a non-parameter kernel density estimation method to obtain an interval prediction result, and sending the interval prediction result to a grid customer service scheduling system so that the grid customer service scheduling system carries out telephone traffic scheduling according to the interval prediction result.
Optionally, the quantile regression model is adopted to calculate the feature data to obtain the conditional quantiles of different quantile points, which specifically comprises the following steps:
solving the characteristic data by using a quantile regression function to obtain regression coefficient vectors of different quantile points, wherein the quantile regression function is as follows:
wherein the method comprises the steps of，Representing dependent variablesP p Is the first of (2)τThe number of digits of the individual condition is calculated,τthe range of (1, 0),β(τ) Is a vector of regression coefficients, +.>Represented as feature data;
solving regression coefficient vectors of different quantile points to obtain conditional quantile of the different quantile points, wherein a solving formula is as follows:
wherein L is the neural network loss function value,represented as characteristic data>Expressed as an asymmetric function:
where s is the argument of the function γ.
In this embodiment, the historical traffic data and the traffic influencing factor data sequence are input into the prediction model, and it should be noted that, the prediction model may be preferably constructed based on a long-short memory neural network, and the long-short memory neural network is adopted to process the traffic data on similar days, where the processing steps are as follows:
time of daytThe input of the corresponding neuron has three parts: at this point in timepTraffic data of (a)z p Output at the previous timeo p-1 State value of previous timeS p-1 At the same time, time of daypWill also output the neurono p Sum state valueS p To the next neuron;
the long-term memory neural network introduces three control gates: input doord p Output doore p Forgetful doorf p . The three control gates are all between 0 and 1]The coefficients of the interval and the calculation formulas of the three control gates are as follows:
wherein,W d 、W e 、W f respectively a weight matrix of three control gates,b d 、b e 、b f respectively, the corresponding offset amounts are set,is a sigmoid function and then based on the input at the current timez p And the output of the last timeo p-1 To calculate candidate state value of the current neuron +.>Wherein->The expression of (2) is:
wherein,W s 、b s respectively representing a weight matrix and a bias of the candidate state;
the state value at the current moment is calculated from the state value at the previous moment and the current waiting timeThe state value is selected and obtained by forgetting the doorf p And an input doord p To determine the corresponding ratio, representing the multiplication by element:
finally, calculating the output value at the current momento p ：
Wherein,representing an output gate.
In the prediction model, firstly, feature extraction is carried out on clustering results of complete historical telephone traffic data and complete telephone traffic influence factor data to obtain feature data, and then a quantile regression function is utilized to solve the feature data to obtain regression coefficient vectors of different quantile points, wherein the quantile regression function is as follows:
wherein,representing dependent variablesP p Is the first of (2)τThe number of digits of the individual condition is calculated,τthe range of (1, 0),β(τ) Is a vector of regression coefficients, +.>Represented as feature data;
solving regression coefficient vectors of different quantile points to obtain conditional quantile of the different quantile points, wherein a solving formula is as follows:
wherein L is the neural network loss function value,represented as characteristic data>Expressed as an asymmetric function:
where s is the argument of the function γ.
Calculating the conditional quantiles by using a non-parameter kernel density estimation function to obtain an interval prediction result, wherein the kernel density estimation function is as follows:
wherein,V(u a ) For the probability distribution of the traffic on the a-th day to be predicted,u a as an independent variable of the probability function,N a for a similar number of days for the a moments to be predicted,hfor the bandwidth, takeh=0.01，K() As a kernel function, taking the kernel function as a Gaussian function,G P in the case of a vector of samples,G P =[,/>,/>…,/>]。
and sending the obtained interval prediction result to a power grid customer service scheduling system so that the power grid customer service scheduling system performs traffic scheduling according to the interval prediction result.
As an example of the embodiment, taking traffic data of the electric power customer service center 2020, 1 st 2021 nd 12 nd 31 st in a certain city in the south as an example data set, the sampling period is 1 hour; traffic data from 1 st 2020 to 12 nd 20 st 2021 is used as a training set, and traffic data from 21 st 2021 to 31 st 2021 is used as a test set; comparing three methods of similar daily cluster-kernel density estimation (method one), CNN-kernel density estimation (method two) and LSTM-Gaussian (method three) with the method provided by the application, the CNN network parameters used in the calculation example are as follows: the two convolution layers, the two pooling layers and one full-connection layer are respectively provided with convolution kernels with the size of 2 multiplied by 2, the number of the convolution kernels is respectively 12 and 16, the pooling window of the pooling layers is respectively provided with the size of 2 multiplied by 2, the step length is 12, and the number of neurons of the two full-connection layers is respectively 100 and 120; LSTM parameters are: the number of network layers is 3, and the number of hidden layer nodes is 12.
The reliability and sharpness performance of the prediction result were evaluated by selecting the interval coverage (PICP) and the Prediction Interval Average Width (PIAW), and the results are shown in table 1:
table 1 comparison of the evaluation results of the methods at different confidence intervals
Under the confidence levels of 95%, 90% and 80%, the interval prediction method provided by the application has higher coverage rate of the prediction interval and higher sensitivity of the prediction interval.
The application has the following beneficial effects:
obtaining historical telephone traffic data and telephone traffic influencing factor data, supplementing missing values of the historical telephone traffic data and telephone traffic influencing factor data to obtain complete historical telephone traffic data and complete telephone traffic influencing factor data, carrying out similar daily clustering on the complete telephone traffic influencing factor data to obtain a clustering result of the complete telephone traffic influencing factor data, inputting the clustering result of the complete historical telephone traffic data and the complete telephone traffic influencing factor data into a constructed prediction model, carrying out feature extraction on the clustering result of the complete historical telephone traffic data and the complete telephone traffic influencing factor data by the prediction model to obtain feature data, calculating the feature data by a quantile regression model to obtain a conditional quantile, calculating the conditional quantile by a non-parameter kernel density estimation method to obtain a section prediction result, and sending the section prediction result to a grid customer service scheduling system so that the grid customer service scheduling system carries out telephone traffic scheduling according to the section prediction result. According to the method, the telephone traffic probability of the power grid customer service is predicted by combining telephone traffic influencing factors, so that the effective quantification of the uncertainty of the prediction of a modern power supply service system can be realized, and the accuracy of the prediction of the telephone traffic interval of the power grid customer service is improved.
Example two
Correspondingly, referring to fig. 3, fig. 3 is a schematic structural diagram of a system for predicting a traffic interval of a customer service of a power grid, as shown in the drawing, the system for predicting the traffic interval of the customer service of the power grid includes a data acquisition module 301, a clustering module 302, and a prediction interval calculation module 303, wherein specific units of each module are as follows:
the data acquisition module 301 is configured to acquire historical traffic data and traffic influencing factor data, and perform missing value supplementation on the historical traffic data and the traffic influencing factor data to obtain complete historical traffic data and complete traffic influencing factor data;
the clustering module 302 is configured to perform similar daily clustering on complete traffic volume influence factor data to obtain a clustering result of the complete traffic volume influence factor data;
the prediction interval calculation module 303 is configured to input the clustering result of the complete historical traffic data and the complete traffic influencing factor data into a constructed prediction model, so that the prediction model performs feature extraction on the clustering result of the complete historical traffic data and the complete traffic influencing factor data to obtain feature data, then calculates the feature data by adopting a quantile regression model to obtain the conditional quantiles of different quantiles, calculates the conditional quantiles by adopting a non-parametric kernel density estimation method to obtain an interval prediction result, and sends the interval prediction result to the grid customer service scheduling system so that the grid customer service scheduling system performs traffic scheduling according to the interval prediction result.
Alternatively, the data acquisition module 301 includes an approximation calculation unit 3011 and a padding unit 3012,
the approximate value calculation unit 3011 is used for obtaining an approximate value of historical traffic data and an approximate value of traffic volume influence factor data by fitting a polynomial with adjacent normal data;
the filling unit 3012 is configured to supplement the approximate value as a missing value to the historical traffic data and the traffic influencing factor data, so as to obtain complete historical traffic data and complete traffic influencing factor data.
The clustering module 302 comprises a central point selection unit 3021 and a clustering result generation unit 3022,
the central point selecting unit 3021 is configured to normalize the complete traffic volume influencing factor data and then select a plurality of central points;
the clustering result generating unit 3022 is configured to distribute each complete traffic volume influencing factor data to a central point smaller than a preset distance, obtain a plurality of clusters, and recalculate the central point of each cluster as a clustering result of the complete traffic volume influencing factor data.
The above-mentioned system for predicting the traffic interval of the power grid customer service can implement the method for predicting the traffic interval of the power grid customer service in the embodiment of the method. The options in the method embodiments described above are also applicable to this embodiment and will not be described in detail here. The rest of the embodiments of the present application may refer to the content of the above method embodiments, and in this embodiment, no further description is given.
Compared with the prior art, the method has the advantages that the historical telephone traffic data and the telephone traffic influence factor data are obtained, missing value supplementation is carried out on the historical telephone traffic data and the telephone traffic influence factor data, the complete historical telephone traffic data and the complete telephone traffic influence factor data are obtained, similar daily clustering is carried out on the complete telephone traffic influence factor data to obtain the clustering result of the complete telephone traffic influence factor data, the clustering result of the complete historical telephone traffic data and the clustering result of the complete telephone traffic influence factor data is input into a built prediction model, after feature extraction is carried out on the clustering result of the complete historical telephone traffic data and the complete telephone traffic influence factor data by the prediction model to obtain feature data, the feature data is calculated by adopting a quantile regression model to obtain a conditional quantile, then the conditional quantile is calculated by adopting a non-parameter kernel density estimation method to obtain an interval prediction result, and the interval prediction result is sent to a grid customer service scheduling system, so that the grid customer service scheduling system carries out telephone traffic scheduling according to the interval prediction result. According to the method, the telephone traffic probability of the power grid customer service is predicted by combining telephone traffic influencing factors, so that the effective quantification of the uncertainty of the prediction of a modern power supply service system can be realized, and the accuracy of the prediction of the telephone traffic interval of the power grid customer service is improved.
The foregoing embodiments have been provided for the purpose of illustrating the general principles of the present application, and are not to be construed as limiting the scope of the application. It should be noted that any modifications, equivalent substitutions, improvements, etc. made by those skilled in the art without departing from the spirit and principles of the present application are intended to be included in the scope of the present application.
Claims (5)
1. The utility model provides a power grid customer service traffic interval prediction method which is characterized by comprising the following steps:
acquiring historical telephone traffic data and telephone traffic influence factor data, and supplementing missing values of the historical telephone traffic data and the telephone traffic influence factor data to obtain complete historical telephone traffic data and complete telephone traffic influence factor data;
performing similar daily clustering on the complete telephone traffic influence factor data to obtain a clustering result of the complete telephone traffic influence factor data;
inputting the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data into a constructed prediction model, so that the prediction model performs feature extraction on the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data to obtain feature data, then calculates the feature data by adopting a quantile regression model to obtain the conditional quantiles of different quantile points, calculates the conditional quantiles by adopting a non-parameter kernel density estimation method to obtain an interval prediction result, and sends the interval prediction result to a power grid customer service scheduling system so that the power grid customer service scheduling system performs telephone traffic scheduling according to the interval prediction result;
the clustering result of the complete telephone traffic influence factor data is obtained by carrying out similar daily clustering on the complete telephone traffic influence factor data, and specifically comprises the following steps:
after normalizing the complete telephone traffic influence factor data, selecting a plurality of center points;
distributing the complete telephone traffic influence factor data to a center point smaller than a preset distance to obtain a plurality of clusters, and recalculating the center point of each cluster to serve as a clustering result of the complete telephone traffic influence factor data;
calculating the characteristic data by adopting a quantile regression model to obtain the conditional quantiles of different quantile points, wherein the conditional quantiles are specifically as follows:
solving the characteristic data by using a quantile regression function to obtain regression coefficient vectors of different quantile points, wherein the quantile regression function is as follows:
wherein,P p representing the probability of the P-th feature data,representation ofP p Is the first of (2)τThe number of digits of the individual condition is calculated,τthe range of (1, 0),β(τ) Is a vector of regression coefficients, +.>Representing the p-th feature data;
solving the regression coefficient vector of the different quantile points to obtain the conditional quantile of the different quantile points, wherein the solving formula is as follows:
wherein L is the neural network loss function value,represents the p-th characteristic data, N represents the total number of characteristic data, +.>Expressed as an asymmetric function:
where s represents the argument of the function gamma randomly generated,representing regression coefficients;
calculating the conditional quantiles by adopting a non-parameter kernel density estimation method to obtain an interval prediction result, wherein the method specifically comprises the following steps of:
calculating the conditional quantiles by using a non-parameter kernel density estimation function to obtain an interval prediction result, wherein the kernel density estimation function is as follows:
wherein,V(u a ) For the probability distribution of the traffic on the a-th day to be predicted,u a as an independent variable of the probability function,N a for a similar number of days at a time instant to be predicted, h is the bandwidth, taking h=0.01,K() As a kernel function, taking the kernel function as a Gaussian function,G P in the case of a vector of samples,G P =[, />,/>…,/>]，/> 、/> 、/> …/>regression coefficients of the 1 st, 2 nd, 3 rd and … th feature data are shown.
2. The method for predicting a customer service traffic interval of a power grid according to claim 1, wherein the step of supplementing the historical traffic data and the traffic influencing factor data by missing values to obtain complete historical traffic data and complete traffic influencing factor data comprises the following steps:
obtaining an approximation of the historical traffic data and an approximation of the traffic impact factor data by fitting a polynomial with adjacent normal data;
and supplementing the approximate value as a missing value to the historical telephone traffic data and the telephone traffic influence factor data to obtain complete historical telephone traffic data and complete telephone traffic influence factor data.
3. The method for predicting a traffic volume of customer service in a power grid according to claim 1, wherein the following steps are performed
The prediction model is obtained through training, and specifically comprises the following steps:
dividing the clustering result of the complete historical telephone traffic data and the complete telephone traffic influence factor data into a training set and a testing set;
constructing a prediction model, and inputting the clustering result of the complete historical telephone traffic data and the complete telephone traffic influence factor data into the prediction model for calculation to obtain an actual interval prediction result;
and comparing the actual interval prediction result with the expected interval prediction result according to the error function to obtain an error, and updating model parameters according to the error until the training times are reached to obtain a trained prediction model.
4. A prediction system for a traffic interval of power grid customer service is characterized by comprising a data acquisition module, a clustering module and a prediction interval calculation module, wherein,
the data acquisition module is used for acquiring historical telephone traffic data and telephone traffic influence factor data, and supplementing missing values of the historical telephone traffic data and the telephone traffic influence factor data to obtain complete historical telephone traffic data and complete telephone traffic influence factor data;
the clustering module is used for carrying out similar daily clustering on the complete telephone traffic influence factor data to obtain a clustering result of the complete telephone traffic influence factor data;
the prediction interval calculation module is used for inputting the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data into a constructed prediction model, so that the prediction model performs feature extraction on the clustering results of the complete historical telephone traffic data and the complete telephone traffic influence factor data to obtain feature data, then calculates the feature data by adopting a quantile regression model to obtain the conditional quantiles of different quantiles, calculates the conditional quantiles by adopting a non-parametric kernel density estimation method to obtain an interval prediction result, and sends the interval prediction result to a power grid customer service scheduling system so that the power grid customer service scheduling system performs telephone traffic scheduling according to the interval prediction result;
the clustering module comprises a central point selecting unit and a clustering result generating unit,
the central point selecting unit is used for selecting a plurality of central points after normalizing the complete telephone traffic influencing factor data;
the clustering result generating unit is used for distributing the complete telephone traffic influence factor data to a center point smaller than a preset distance to obtain a plurality of clusters and recalculate the center point of each cluster to be used as a clustering result of the complete telephone traffic influence factor data;
calculating the characteristic data by adopting a quantile regression model to obtain the conditional quantiles of different quantile points, wherein the conditional quantiles are specifically as follows:
solving the characteristic data by using a quantile regression function to obtain regression coefficient vectors of different quantile points, wherein the quantile regression function is as follows:
wherein,P p representing the probability of the P-th feature data,representation ofP p Is the first of (2)τThe number of digits of the individual condition is calculated,τthe range of (1, 0),β(τ) Is a vector of regression coefficients, +.>Representing the p-th feature data;
solving the regression coefficient vector of the different quantile points to obtain the conditional quantile of the different quantile points, wherein the solving formula is as follows:
wherein L is the neural network loss function value,represents the p-th characteristic data, N represents the total number of characteristic data, +.>Expressed as an asymmetric function:
where s represents the argument of the function gamma randomly generated,representing regression coefficients;
calculating the conditional quantiles by adopting a non-parameter kernel density estimation method to obtain an interval prediction result, wherein the method specifically comprises the following steps of:
calculating the conditional quantiles by using a non-parameter kernel density estimation function to obtain an interval prediction result, wherein the kernel density estimation function is as follows:
wherein,V(u a ) For the probability distribution of the traffic on the a-th day to be predicted,u a as an independent variable of the probability function,N a for a similar number of days for the a moments to be predicted,hfor the bandwidth, takeh=0.01，K() As a kernel function, taking the kernel function as a Gaussian function,G P in the case of a vector of samples,G P =[,/>,/>…, />]，/> 、/> 、/> …/>regression coefficients of the 1 st, 2 nd, 3 rd and … th feature data are shown.
5. The grid customer service traffic interval prediction system as recited in claim 4, wherein the data acquisition module comprises an approximation calculation unit and a filling unit,
the approximation calculation unit is used for obtaining the approximation of the historical traffic data and the approximation of the traffic volume influence factor data by fitting a polynomial by using adjacent normal data;
and the filling unit is used for supplementing the approximate value as a missing value to the historical telephone traffic data and the telephone traffic influencing factor data to obtain complete historical telephone traffic data and complete telephone traffic influencing factor data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310483798.9A CN116227738B (en) | 2023-05-04 | 2023-05-04 | Method and system for predicting traffic interval of power grid customer service |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310483798.9A CN116227738B (en) | 2023-05-04 | 2023-05-04 | Method and system for predicting traffic interval of power grid customer service |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116227738A CN116227738A (en) | 2023-06-06 |
| CN116227738B true CN116227738B (en) | 2023-12-08 |
Family
ID=86577214
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310483798.9A Active CN116227738B (en) | 2023-05-04 | 2023-05-04 | Method and system for predicting traffic interval of power grid customer service |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116227738B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101159976A (en) * | 2007-08-24 | 2008-04-09 | 中国移动通信集团设计院有限公司 | Method and device of predicting telephone traffic and channel configuration |
| CN103002165A (en) * | 2012-11-21 | 2013-03-27 | 江苏省电力公司电力科学研究院 | Method for predicting short-term telephone traffic of power supply service center |
| CN105847598A (en) * | 2016-04-05 | 2016-08-10 | 浙江远传信息技术股份有限公司 | Method and device for call center multifactorial telephone traffic prediction |
| CN106713677A (en) * | 2016-05-24 | 2017-05-24 | 国家电网公司客户服务中心 | Prediction method for incoming call traffic of power client service center |
| CN109063970A (en) * | 2018-07-06 | 2018-12-21 | 郑州天迈科技股份有限公司 | Two-way automatic scheduling system based on passenger flow simulation analysis |
| CN109978201A (en) * | 2017-12-27 | 2019-07-05 | 深圳市景程信息科技有限公司 | Probability load prediction system and method based on Gaussian process quantile estimate model |
| CN113973156A (en) * | 2021-12-22 | 2022-01-25 | 杭州远传新业科技有限公司 | Telephone traffic prediction method and system and telephone traffic prediction device |
| CN115271041A (en) * | 2022-07-25 | 2022-11-01 | 国家电网有限公司客户服务中心 | Method for predicting telephone traffic of power service |
-
2023
- 2023-05-04 CN CN202310483798.9A patent/CN116227738B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101159976A (en) * | 2007-08-24 | 2008-04-09 | 中国移动通信集团设计院有限公司 | Method and device of predicting telephone traffic and channel configuration |
| CN103002165A (en) * | 2012-11-21 | 2013-03-27 | 江苏省电力公司电力科学研究院 | Method for predicting short-term telephone traffic of power supply service center |
| CN105847598A (en) * | 2016-04-05 | 2016-08-10 | 浙江远传信息技术股份有限公司 | Method and device for call center multifactorial telephone traffic prediction |
| CN106713677A (en) * | 2016-05-24 | 2017-05-24 | 国家电网公司客户服务中心 | Prediction method for incoming call traffic of power client service center |
| CN109978201A (en) * | 2017-12-27 | 2019-07-05 | 深圳市景程信息科技有限公司 | Probability load prediction system and method based on Gaussian process quantile estimate model |
| CN109063970A (en) * | 2018-07-06 | 2018-12-21 | 郑州天迈科技股份有限公司 | Two-way automatic scheduling system based on passenger flow simulation analysis |
| CN113973156A (en) * | 2021-12-22 | 2022-01-25 | 杭州远传新业科技有限公司 | Telephone traffic prediction method and system and telephone traffic prediction device |
| CN115271041A (en) * | 2022-07-25 | 2022-11-01 | 国家电网有限公司客户服务中心 | Method for predicting telephone traffic of power service |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度学习分位数回归模型的风电功率概率密度预测;李彬等;《电力自动化设备》;第38卷(第9期);第15-20页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116227738A (en) | 2023-06-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110705743B (en) | New energy consumption electric quantity prediction method based on long-term and short-term memory neural network | |
| CN111260136A (en) | Building short-term load prediction method based on ARIMA-LSTM combined model | |
| CN111861013B (en) | Power load prediction method and device | |
| CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
| CN109583565A (en) | Forecasting Flood method based on the long memory network in short-term of attention model | |
| CN109492748B (en) | Method for establishing medium-and-long-term load prediction model of power system based on convolutional neural network | |
| CN112766600B (en) | Urban area crowd flow prediction method and system | |
| CN111008726B (en) | Class picture conversion method in power load prediction | |
| CN113112791A (en) | Traffic flow prediction method based on sliding window long-and-short term memory network | |
| CN114792156A (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| CN114330934A (en) | Model parameter self-adaptive GRU new energy short-term power generation power prediction method | |
| CN110807508B (en) | Bus peak load prediction method considering complex weather influence | |
| CN113868938A (en) | Short-term load probability density prediction method, device and system based on quantile regression | |
| CN116050595A (en) | Attention mechanism and decomposition mechanism coupled runoff amount prediction method | |
| CN115204477A (en) | Bicycle flow prediction method of context awareness graph recursive network | |
| CN115186923A (en) | Photovoltaic power generation power prediction method and device and electronic equipment | |
| CN115238854A (en) | Short-term load prediction method based on TCN-LSTM-AM | |
| CN114897264A (en) | Photovoltaic output interval prediction method under small sample scene based on transfer learning | |
| CN114721898A (en) | Edge cloud server utilization rate prediction method and device based on boosting algorithm and storage medium | |
| CN114357670A (en) | Power distribution network power consumption data abnormity early warning method based on BLS and self-encoder | |
| CN114757330A (en) | Urban instantaneous water consumption prediction method based on LSTM | |
| CN116227738B (en) | Method and system for predicting traffic interval of power grid customer service | |
| CN115713044B (en) | Method and device for analyzing residual life of electromechanical equipment under multi-condition switching | |
| CN109447843B (en) | Photovoltaic output prediction method and device | |
| Lai et al. | Short‐term passenger flow prediction for rail transit based on improved particle swarm optimization algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |