CN110309968A - A kind of Dynamic Pricing System and method based on pile group prediction charge volume - Google Patents
A kind of Dynamic Pricing System and method based on pile group prediction charge volume Download PDFInfo
- Publication number
- CN110309968A CN110309968A CN201910571591.0A CN201910571591A CN110309968A CN 110309968 A CN110309968 A CN 110309968A CN 201910571591 A CN201910571591 A CN 201910571591A CN 110309968 A CN110309968 A CN 110309968A
- Authority
- CN
- China
- Prior art keywords
- data
- charge volume
- model
- deep learning
- learning model
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000013136 deep learning model Methods 0.000 claims abstract description 43
- 230000000737 periodic effect Effects 0.000 claims abstract description 35
- 238000012545 processing Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 36
- 238000013528 artificial neural network Methods 0.000 claims description 29
- 238000012360 testing method Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 230000007774 longterm Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims 1
- 210000005036 nerve Anatomy 0.000 claims 1
- 230000005611 electricity Effects 0.000 abstract description 4
- 230000001737 promoting effect Effects 0.000 abstract description 2
- 238000012546 transfer Methods 0.000 abstract description 2
- 239000011159 matrix material Substances 0.000 description 8
- 230000006870 function Effects 0.000 description 6
- 210000004027 cell Anatomy 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 229940050561 matrix product Drugs 0.000 description 3
- 210000000653 nervous system Anatomy 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0206—Price or cost determination based on market factors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Primary Health Care (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of Dynamic Pricing Systems based on pile group prediction charge volume, including data module, data module includes time granularity division to data processing and is standardized to data, data module is connected with prediction model unit, prediction model unit includes periodic depth learning model, time series deep learning model and surface model, treated data by periodic depth learning model and time series deep learning Model Transfer are connected to surface model by data module, surface model is externally connected with pricing module, pricing data after pricing module calculates is exported by output module;The present invention realizes Dynamic Pricing using the frequency of use of charging pile each time point, solving existing charging pile all is fixed electricity price method, and no matter the utilization rate of pile group just, uses unified price, it is unfavorable for using and promoting for charging pile, is unfavorable for charging pile and realizes the pricing problem optimized.
Description
Technical field
The invention belongs to electric automobile charging pile pricing techniques fields, and in particular to a kind of to predict charge volume based on pile group
Dynamic Pricing System and method.
Background technique
At this stage, charging platform mainly takes fixed electricity price method, and no matter the utilization rate of pile group just, uses unification
Price.The present invention proposes the utilization rate according to charging pile, Dynamic Pricing is realized, to improve the integral benefit of system.
Periodic history average is mainly based upon to the method for charge volume data prediction.This method takes history with for the moment
The average value of charge volume is carved as predicted value.For example to predict 12 points of charge volume today, then take the last week or one month daily
The average value of 12 points of charge volume is as predicted value.History average can preferably be carried out using the periodic feature of data pre-
It surveys, prediction result is relatively preferable, but prediction result has hysteresis quality.When having special event generation in short term, for example, charging is excellent
Favour activity, charge volume is significantly increased in a short time, and history average is difficult to capture this variation, and prediction effect is non-in this case
It is often poor.Moreover, the unpredictable future prolonged charge volume of history average.
Neural network is most initially the inspiration by biological nervous system, is occurred to simulate biological nervous system, by
Composition is coupled to each other between a large amount of node (or neuron).Neural network is adjusted weight according to the variation of input,
The behavior of improvement system, automatic study are able to solve the model of problem to one.LSTM (length memory network) is RNN (circulation
Neural network) a kind of special shape, efficiently solve multilayer neural network training gradient disappear and gradient explosion issues, energy
Time Dependent sequence when enough handling long.LSTM can capture the time series characteristic of charge volume data, can using LSTM model
Effectively improve precision of prediction.
LSTM network is made of LSTM unit, and LSTM unit is by unit, input gate, out gate and forgetting door composition.
Forget door: determining how much information abandoned from the output state of a upper unit, formula is as follows:
Wherein,It is the output for forgeing door,It is list entries,It is the output of a upper unit,Indicate sigmoid
Function,Indicate the weight parameter matrix of input,Indicate the weight parameter matrix of unit output,Indicate deviation
Parameter vector.
Input gate: it determines that how many new information is allowed to be added in Cell state, and location mode C is updated, formula
It is as follows:
Wherein,Indicate the location mode of active cell,WithIndicate sigmoid function,Representing matrix product,Table
Show the weight parameter matrix of input,Indicate the weight parameter matrix of unit output,Indicate straggling parameter vector,
It is the output for forgeing door,It is the location mode of a upper unit,Representing matrix product,Indicate the weight parameter of input
Matrix,Indicate the weight parameter matrix of unit output,Indicate straggling parameter vector.
Out gate: result is exported based on current location mode.
Wherein,Indicate the output of active cell,WithIndicate sigmoid function,Representing matrix product,Table
Show the weight parameter matrix of input,Indicate the weight parameter matrix of unit output,Indicate straggling parameter vector.
Summary of the invention
It is all fixed electricity price method the invention solves existing charging pile, no matter the utilization rate of pile group just, uses
Unified price is unfavorable for using and promoting for charging pile, is unfavorable for charging pile and realizes the pricing problem optimized, provides thus
A kind of Dynamic Pricing System based on pile group prediction charge volume.
The technical solution used to solve the technical problems of the present invention is that:
A kind of Dynamic Pricing System based on pile group prediction charge volume, including data module, the data module is to data processing
It is divided including time granularity and data is standardized, the data module is connected with prediction model unit, the prediction mould
Type unit includes periodic deep learning model, time series deep learning model and surface model, the data module
Treated data by periodic deep learning model and time series deep learning Model Transfer are connected to surface
Model, the surface model are externally connected with pricing module, and the pricing data after the pricing module calculates passes through output mould
Block output.
Further, the periodic deep learning model and time series deep learning model include encoder, solution
Code device and neural network unit, the encoder are connected to decoder, and the decoder is connected to neural network unit.
Further, the surface model includes three layers of full Connection Neural Network unit, described three layers full connection mind
Data characteristics module is also connected with through network unit.
A method of the Dynamic Pricing System based on pile group prediction charge volume, comprising the following steps:
Step 1: charge volume is pre-processed, hour granularity division is carried out to the time by data processing unit, to data into
Row standardization;
Step 2: pretreated charge volume data are divided into training set, verifying collection and test set;
Step 3: the charge volume data after division are passed through prediction model building unit prediction model, including periodic depth
Practise model, time series deep learning model and surface model;
Step 4: pre-training is carried out using pre-training part of the training set data to time series deep learning model, it is excellent in advance
The parameter for changing time series deep learning model is avoided parameter optimization to local best points in integrally training;
Step 5: carrying out whole training to 3 kinds of models that step 3 is established using training set data and verifying collection data;
Step 6: carrying out short-term prediction using test set data and using the trained model of step 5;
Step 7: carrying out long-term prediction using test set data and using the trained model of step 5;
Step 8: realizing Dynamic Pricing using prediction result.
The calculation method of the step 1 are as follows:
Wherein,Indicate charge volume original value,Indicate the minimum value of charge volume original value,Indicate that charge volume is original
The maximum value of value,For normalized upper limit value,For normalized lower limit value,After indicating normalization
Section,For the result after standardization.
The calculation method of the step 2 and step 3 are as follows:
Time series data is used in time series deep learning model:
;
Periodic sequence data are used in periodic deep learning model:
;
Surface data are used in surface model:;
Wherein, n indicates current time, and t indicates the step-length of time series, and p indicates the step-length of periodic sequence.It indicates at n-th
The charge volume at quarter,Indicate the charge volume of the first i days phases at the n-th moment in the same time,I weeks certain day the n-th moment before indicating
Charge volume;Indicate the charging duration set at the preceding t moment including the n-th moment,It indicates including the preceding p on the day of the n-th moment
The charging duration set at its in a few days identical moment,Indicate the surface at the n-th moment, including festivals or holidays, the band of position, day
Gas.
Compared with prior art, beneficial effects of the present invention are as follows: the present invention is not first merely with the week of charge volume data
Phase property feature also utilizes the time series feature of charge volume data, the utilization to data and precision of prediction greatly improved;Its
It is secondary to consider the influence of surface festivals or holidays, charging pile region to charge volume, it adds it in prediction model, substantially
Improve precision of prediction and the prediction to particular value;Dynamic Pricing strategy is finally designed based on prediction result, greatly improves system
Income.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples.
Fig. 1 is the flow chart of the embodiment of the present invention;
The most important component symbol of the embodiment of the present invention is as follows:
It is data module -1, prediction model unit -2, periodic deep learning model -3, time series deep learning model -4, outer
- 5, three layers of portion's characteristic model full Connection Neural Network unit -501, encoder -6, decoder -7, neural network unit -8, price
Module -9, output module -10, data characteristics module -11.
Specific embodiment
In the description of the present invention, it is to be understood that, term " center ", " longitudinal direction ", " transverse direction ", " length ", " width ",
" thickness ", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom" "inner", "outside", " axial direction ",
The orientation or positional relationship of the instructions such as " radial direction ", " circumferential direction " is to be based on the orientation or positional relationship shown in the drawings, merely to just
In description the present invention and simplify description, rather than the device or element of indication or suggestion meaning must have a particular orientation, with
Specific orientation construction and operation, therefore be not considered as limiting the invention.
Such as Fig. 1, a kind of Dynamic Pricing System based on pile group prediction charge volume, including data module 1, the data module
1 pair of data processing includes that time granularity is divided and is standardized to data, and the data module 1 is connected with prediction model unit
2, the prediction model unit 2 includes periodic deep learning model 3, time series deep learning model 4 and surface mould
Type 5, by treated, data pass through periodic deep learning model 3 and time series deep learning model 4 to the data module 1
Transmitting is connected to surface model 5, and the surface model 5 is externally connected with pricing module 9, and the pricing module 9 calculates
Pricing data afterwards is exported by output module 10.
The periodic deep learning model 3 and time series deep learning model 4 include encoder 6,7 and of decoder
Neural network unit 8, the encoder 6 are connected to decoder 7, and the decoder 7 is connected to neural network unit 8.
The surface model 5 includes three layers of full Connection Neural Network unit 501, three layers of full Connection Neural Network
Unit 501 is also connected with data characteristics module 11.
A method of based on pile group prediction charge volume Dynamic Pricing System, feature the following steps are included:
Step 1: charge volume data are pre-processed
1) time granularity divides: all charge volume data are handled by one hour time granularity as charge volume number hourly
According to;
2) data are standardized: charge volume data hourly are standardized using minimum value maximum value, formula is such as
Under:
Wherein,Indicate charge volume original value,Indicate the minimum value of charge volume original value,Indicate that charge volume is original
The maximum value of value,For normalized upper limit value,For normalized lower limit value,After indicating normalization
Section,For the result after standardization.
Step 2: pretreated charge volume data are divided into training set, verifying collection and test set.In each data set
In, the data that different models use have following several types:
Time series data is used in time series deep learning model 4:
;
Periodic sequence data are used in periodic deep learning model 3:
;
Surface data are used in surface model 5:;
Wherein, n indicates current time, and t indicates the step-length of time series, and p indicates the step-length of periodic sequence.It indicates at n-th
The charge volume at quarter,Indicate the charge volume of the first i days phases at the n-th moment in the same time,I weeks certain day the n-th moment before indicating
Charge volume.Indicate the charging duration set at the preceding t moment including the n-th moment,It indicates including the preceding p on the day of the n-th moment
The charging duration set at its in a few days identical moment,Indicate the surface at the n-th moment, including festivals or holidays, the band of position, day
Gas.
Step 3: building prediction model unit 2, prediction model unit 2 includes periodic deep learning model 3, time sequence
Column deep learning model 4 and surface model 5, the structure and training mechanism of each model are as follows:
1) periodic deep learning model 3: being the multilayer length memory network LSTM based on attention Attention mechanism
Model, hidden layer include the decoding of the encoder Encoder of two layers of length memory network LSTM, two layers length memory network LSTM
Device Decoder and three layers of full Connection Neural Network unit 501, the structure of decoder Decoder is identical as encoder Encoder.
The realization details of periodic deep learning model 3: by periodic sequence dataInput coding device Encoder, takes volume
All outputs of code device Encoder, and different weights is distributed to all outputs, it is re-used as the input of decoder Decoder, is solved
All outputs of code device Decoder are recently entered to three layers of full Connection Neural Network unit 501, three layers of full Connection Neural Network list
The output of member 501 is the output of periodic deep learning model 3.
2) time series deep learning model 4: being one based on new coder-decoder Encoder-Decoder machine
The multilayer length memory network LSTM model of system includes pre-training part and predicted portions.Pre-training part includes two layers of length
The decoder Decoder and one layer of full articulamentum of the encoder Encoder of memory network LSTM, two layers length memory network LSTM
Neural network, predicted portions include two layers of length memory network LSTM of pre-training part encoder Encoder and three layers it is complete
Connection Neural Network unit 501.
The realization details of pre-training part: by time series dataInput coding device Encoder, encoder Encoder
The last one unit location mode as decoder Decoder first unit initial cell state, decoder
All outputs of Decoder are input to one layer of full articulamentum neural network, even if the output of obtained output pre-training.
The realization details of predicted portions: by time series dataInput coding device Encoder, encoder Encoder's
The output of the last one unit is input to three layers of full Connection Neural Network unit 501 again, three layers of full articulamentum neural network it is defeated
Out be predicted portions output as a result, and time series deep learning model 4 output.
3) surface model 5: including three layers of full Connection Neural Network unit 501;It realizes details: taking periodic depth
Practise the output and surface data of model 3, time series deep learning model 4It is combined into an one-dimensional vector, it will
The one-dimensional vector is inputted as three layers of full Connection Neural Network unit 501, by three layers of full Connection Neural Network unit 501, most
Output prediction result eventually.
Step 4: carrying out pre-training using pre-training part of the training set data to time series deep learning model 4, mention
The parameter of preceding optimization time series deep learning model 4 is avoided parameter optimization to local best points in integrally training;
The location mode that first unit of encoder Encoder is initialized with nought state, by input data, i.e., preceding t1It is small
When charge volume be input to encoder Encoder, obtain the output of encoder Encoder;Take encoder Encoder the last one
Initial cell state of the location mode of unit as first unit of decoder Decoder, then by time series data,
Before i.e.The charge volume of hour is input to decoder Decoder, obtains all outputs of decoder Decoder, reuses one layer
Full Connection Neural Network, which calculate to all outputs of Decoder, finally obtains predicted value;It calculates predicted value and charge volume is true
The RMSE root-mean-square error of value minimizes RMSE using Adam method, by the parameter training of model to suitable value.Trained damage
Mistake function is RMSE root-mean-square error, and formula is as follows:
Wherein,Indicate the charge volume true value at the i-th moment,The charge volume predicted value for indicating for the i-th moment, when pre-training portion
The loss function value divided stops pre-training when being reduced to 0.05 or less;
Step 5: carrying out whole training to 3 kinds of models that step 3 is established using training set data and verifying collection data
Input data is input in 3 kinds of models, while whole training is carried out to 3 kinds of models, what surface model 5 was exported
It as a result is exactly final predicted value.The loss function value of predicted value and charge volume true value after calculating training every time, minimizes damage
Functional value is lost, by the parameter training of model to target value.According to effect of the model on training set, verifying collection, constantly debugging mould
The hyper parameter of type improves precision of prediction under conditions of reducing over-fitting.
The input data includes: time series data, periodic sequence data, surface dataOutside festivals or holidays at the n-th moment, region, weather and temperature
Portion's characteristic and charge volume true value areThe charging magnitude at next moment.
Step 6: carrying out short-term prediction using test set data and using the trained model of step 5.
Input data are as follows: time series data, periodic sequence data, surface dataIt is true with charge volume
Value is。
The result that surface model 5 is exported is exactly final short-term prediction value.
Step 7: carrying out long-term prediction using test set data and using the trained model of step 5.
Input data are as follows: time series data, periodic sequence data, surface data;
After once predicting, the charge volume at next moment is obtained, by predicted valueIt is put into as true value defeated
Enter in data, reconfigure input data: time series data, periodic sequence data, surface data, then input data is input in model, prediction obtains the charge volume at next moment。
By predicted valueIt is put into input data as true value, reconfigures input data, repeatedly, until
Obtain the predicted value in object time section.
The weighing computation method: output state h=< h of encoder Encoder is set1, h2, h3... ..., ht>, hiIt is
The output of i-th of unit of encoder Encoder, t are the step-lengths of encoder Encoder;The output state of decoder Decoder is,It is the output of i-th of unit of decoder Decoder, p is decoder
The step-length of Decoder.
The first step, fusion Encoder output and Decoder state output.
It is two layers of full Connection Neural Network.
Second step calculates eachWeight.
After being handled using Softmax, all weights and be 1.
Step 8: realizing Dynamic Pricing using prediction result.
Input data is the charge volume of prediction of next momentIf predicted value is lower than the ideal utilization rate of pile group,
Then reduce charging price.Pile group utilization rate is improved by reducing price, utilizes the method training price and pile group of machine learning
The relationship of utilization rate, to formulate optimal pricing strategy.Different pricing methods are as follows:
1) two-value pricing strategy: if prediction charge volume is less than a certain given threshold value, low electricity price P is used;Otherwise normal fixed
Valence;
2) ladder pricing strategy: ladder pricing method is taken according to prediction utilization rate, utilization rate is lower, then price is lower.
Taking the above-mentioned ideal embodiment according to the present invention as inspiration, through the above description, relevant staff is complete
Various changes and amendments can be carried out without departing from the scope of the technological thought of the present invention' entirely.The technology of this invention
Property range is not limited to the contents of the specification, it is necessary to which the technical scope thereof is determined according to the scope of the claim.
Claims (6)
1. a kind of Dynamic Pricing System based on pile group prediction charge volume, including data module (1), the data module (1) are right
Data processing includes that time granularity is divided and is standardized to data, it is characterised in that: the data module (1) is connected with pre-
It surveys model unit (2), the prediction model unit (2) includes periodic deep learning model (3), time series deep learning mould
Type (4) and surface model (5), by treated, data pass through periodic deep learning model (3) to the data module (1)
It is connected to surface model (5) with time series deep learning model (4) transmitting, the surface model (5) connects outside
Module that prices are fixed (9), the pricing data after the pricing module (9) calculates are exported by output module (10).
2. a kind of Dynamic Pricing System based on pile group prediction charge volume according to claim 1, it is characterised in that: described
Periodic deep learning model (3) and time series deep learning model (4) include encoder (6), decoder (7) and nerve
Network unit (8), the encoder (6) are connected to decoder (7), and the decoder (7) is connected to neural network unit (8).
3. a kind of Dynamic Pricing System based on pile group prediction charge volume according to claim 2, it is characterised in that: described
Surface model (5) includes three layers of full Connection Neural Network unit (501), described three layers full Connection Neural Network unit
(501) data characteristics module (11) are also connected with.
4. a kind of method of Dynamic Pricing System based on pile group prediction charge volume according to claim 1, feature packet
Include following steps:
Step 1: charge volume is pre-processed, hour granularity division is carried out to the time by data processing unit, to data into
Row standardization;
Step 2: pretreated charge volume data are divided into training set, verifying collection and test set;
Step 3: the charge volume data after division are constructed prediction model, including periodic depth by prediction model unit (2)
Learning model (3), time series deep learning model (4) and surface model (5);
Step 4: carrying out pre-training using pre-training part of the training set data to time series deep learning model (4), in advance
The parameter for optimizing time series deep learning model (4) is avoided parameter optimization to local best points in integrally training;
Step 5: carrying out whole training to 3 kinds of models that step 3 is established using training set data and verifying collection data;
Step 6: carrying out short-term prediction using test set data and using the trained model of step 5;
Step 7: carrying out long-term prediction using test set data and using the trained model of step 5;
Step 8: realizing Dynamic Pricing using prediction result.
5. a kind of method of Dynamic Pricing System based on pile group prediction charge volume according to claim 4, feature exist
In: the calculation method of the step 1 are as follows:
Wherein,Indicate charge volume original value,Indicate the minimum value of charge volume original value,Indicate that charge volume is original
The maximum value of value,For normalized upper limit value,For normalized lower limit value,After indicating normalization
Section,For the result after standardization.
6. a kind of method of Dynamic Pricing System based on pile group prediction charge volume according to claim 4, feature exist
In: the calculation method of the step 2 and step 3 are as follows:
Time series data is used in time series deep learning model (4):
;
Periodic sequence data are used in periodic deep learning model (3):
;
Surface data are used in surface model (5):;
Wherein, n indicates current time, and t indicates the step-length of time series, and p indicates the step-length of periodic sequence;It indicates at n-th
The charge volume at quarter indicates the charge volume of the first i days phases at the n-th moment in the same time,I weeks certain day the n-th moment before indicating
Charge volume;Indicate the charging duration set at the preceding t moment including the n-th moment,It indicates including first p days on the day of the n-th moment
In a few days charging duration set mutually in the same time,Indicate the surface at the n-th moment, including festivals or holidays, the band of position, weather.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910571591.0A CN110309968A (en) | 2019-06-28 | 2019-06-28 | A kind of Dynamic Pricing System and method based on pile group prediction charge volume |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910571591.0A CN110309968A (en) | 2019-06-28 | 2019-06-28 | A kind of Dynamic Pricing System and method based on pile group prediction charge volume |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110309968A true CN110309968A (en) | 2019-10-08 |
Family
ID=68077980
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910571591.0A Pending CN110309968A (en) | 2019-06-28 | 2019-06-28 | A kind of Dynamic Pricing System and method based on pile group prediction charge volume |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110309968A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113866391A (en) * | 2021-09-29 | 2021-12-31 | 天津师范大学 | Deep learning model prediction factor interpretation method and application thereof in soil water content prediction |
CN114619907A (en) * | 2020-12-14 | 2022-06-14 | 中国科学技术大学 | Coordinated charging method and coordinated charging system based on distributed deep reinforcement learning |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001325582A (en) * | 2000-05-17 | 2001-11-22 | Chugoku Electric Power Co Inc:The | Learning and predicting device for time-series data |
CN104268769A (en) * | 2014-03-11 | 2015-01-07 | 南方电网科学研究院有限责任公司 | Time-of-use electricity price design method based on model clustering method |
US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
CN106786692A (en) * | 2016-12-19 | 2017-05-31 | 湘潭大学 | One kind is based on the orderly charge control method of distributed electric automobile |
CN108001282A (en) * | 2017-12-16 | 2018-05-08 | 深圳供电局有限公司 | A kind of charging unit and method that electricity price dynamic adjustment is realized based on big data |
CN109146156A (en) * | 2018-08-03 | 2019-01-04 | 大连理工大学 | A method of for predicting charging pile system charge volume |
CN109829587A (en) * | 2019-02-12 | 2019-05-31 | 国网山东省电力公司电力科学研究院 | Zonule grade ultra-short term and method for visualizing based on depth LSTM network |
-
2019
- 2019-06-28 CN CN201910571591.0A patent/CN110309968A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001325582A (en) * | 2000-05-17 | 2001-11-22 | Chugoku Electric Power Co Inc:The | Learning and predicting device for time-series data |
CN104268769A (en) * | 2014-03-11 | 2015-01-07 | 南方电网科学研究院有限责任公司 | Time-of-use electricity price design method based on model clustering method |
US20170091615A1 (en) * | 2015-09-28 | 2017-03-30 | Siemens Aktiengesellschaft | System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies |
CN106786692A (en) * | 2016-12-19 | 2017-05-31 | 湘潭大学 | One kind is based on the orderly charge control method of distributed electric automobile |
CN108001282A (en) * | 2017-12-16 | 2018-05-08 | 深圳供电局有限公司 | A kind of charging unit and method that electricity price dynamic adjustment is realized based on big data |
CN109146156A (en) * | 2018-08-03 | 2019-01-04 | 大连理工大学 | A method of for predicting charging pile system charge volume |
CN109829587A (en) * | 2019-02-12 | 2019-05-31 | 国网山东省电力公司电力科学研究院 | Zonule grade ultra-short term and method for visualizing based on depth LSTM network |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114619907A (en) * | 2020-12-14 | 2022-06-14 | 中国科学技术大学 | Coordinated charging method and coordinated charging system based on distributed deep reinforcement learning |
CN114619907B (en) * | 2020-12-14 | 2023-10-20 | 中国科学技术大学 | Coordinated charging method and coordinated charging system based on distributed deep reinforcement learning |
CN113866391A (en) * | 2021-09-29 | 2021-12-31 | 天津师范大学 | Deep learning model prediction factor interpretation method and application thereof in soil water content prediction |
CN113866391B (en) * | 2021-09-29 | 2024-03-08 | 天津师范大学 | Deep learning model prediction factor interpretation method and application thereof in soil water content prediction |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | A deep learning based hybrid framework for day-ahead electricity price forecasting | |
CN110766212B (en) | Ultra-short-term photovoltaic power prediction method for historical data missing electric field | |
CN110348624B (en) | Sand storm grade prediction method based on Stacking integration strategy | |
CN111047085B (en) | Hybrid vehicle working condition prediction method based on meta-learning | |
CN109146156B (en) | Method for predicting charging amount of charging pile system | |
CN109754113A (en) | Load forecasting method based on dynamic time warping Yu length time memory | |
CN107590567A (en) | A kind of Recognition with Recurrent Neural Network short-term load forecasting method based on comentropy cluster and notice mechanism | |
CN106952181A (en) | Electric Load Prediction System based on long Memory Neural Networks in short-term | |
CN111091196B (en) | Passenger flow data determination method and device, computer equipment and storage medium | |
CN110837888A (en) | Traffic missing data completion method based on bidirectional cyclic neural network | |
CN108539738A (en) | A kind of short-term load forecasting method promoting decision tree based on gradient | |
CN109583565A (en) | Forecasting Flood method based on the long memory network in short-term of attention model | |
CN112001556B (en) | Reservoir downstream water level prediction method based on deep learning model | |
CN110309968A (en) | A kind of Dynamic Pricing System and method based on pile group prediction charge volume | |
Hua et al. | Forecasting usage and bike distribution of dockless bike‐sharing using journey data | |
CN114792156A (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
CN112766603A (en) | Traffic flow prediction method, system, computer device and storage medium | |
CN114330934A (en) | Model parameter self-adaptive GRU new energy short-term power generation power prediction method | |
Brahimi et al. | Modelling on Car‐Sharing Serial Prediction Based on Machine Learning and Deep Learning | |
Zhu et al. | Multiscale deep network based multistep prediction of high‐dimensional time series from power transmission systems | |
CN111697560B (en) | Method and system for predicting load of power system based on LSTM | |
CN117273217A (en) | Space debris track forecasting method and system and electronic equipment | |
CN110543942A (en) | Multi-space-time long and short memory depth network accurate prediction method | |
Wai et al. | Design of intelligent solar PV power generation forecasting mechanism combined with weather information under lack of real-time power generation data | |
CN116822722A (en) | Water level prediction method, system, device, electronic equipment and medium |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191008 |
|
RJ01 | Rejection of invention patent application after publication |