CN107704970A - A kind of Demand-side load forecasting method based on Spark - Google Patents
A kind of Demand-side load forecasting method based on Spark Download PDFInfo
- Publication number
- CN107704970A CN107704970A CN201711004651.8A CN201711004651A CN107704970A CN 107704970 A CN107704970 A CN 107704970A CN 201711004651 A CN201711004651 A CN 201711004651A CN 107704970 A CN107704970 A CN 107704970A
- Authority
- CN
- China
- Prior art keywords
- key
- value
- spark
- data
- load
- 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
- 238000013277 forecasting method Methods 0.000 title claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 238000013480 data collection Methods 0.000 claims description 5
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 claims description 4
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 claims description 4
- 238000000605 extraction Methods 0.000 claims description 4
- 238000010977 unit operation Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 2
- 230000001133 acceleration Effects 0.000 claims 1
- 238000000034 method Methods 0.000 abstract description 8
- 230000007935 neutral effect Effects 0.000 abstract description 5
- 241001269238 Data Species 0.000 abstract description 3
- 238000007418 data mining Methods 0.000 abstract description 3
- 230000001537 neural effect Effects 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract description 2
- 230000004913 activation Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 241001123248 Arma Species 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013523 data management Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000009466 transformation Effects 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
- 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
-
- 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)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Entrepreneurship & Innovation (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Development Economics (AREA)
- Data Mining & Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The purpose of the present invention is：Spark big datas platform is combined with LSTM neutral net load forecasting methods, avoids unit data mining pattern problem encountered.In order to achieve the above object, the technical scheme is that providing a kind of Demand-side load forecasting method based on Spark.The present invention builds load forecasting model based on the LSTM neural net methods in deep learning field, and can calculate to a nicety out electric load；Predicted based on Spark memory parallels Computational frame come parallel load, LSTM neural network algorithms are subjected to parallelization, parallelization analysis is carried out to historical load data, shortens the load prediction time, improve disposal ability of the LSTM neural network algorithms to big data.
Description
Technical field
A kind of short-term load forecasting method based on the parallel frameworks of Spark of the present invention, belongs to electric power demand side technical field
With big data technical field.
Background technology
With the continuous progressive of modern technologies and the further investigation to intelligent grid, load prediction theory and technology is very
Great development.At present, have many main flows method be applied to electric power demand side load prediction, as artificial neural network, support to
Amount machine, ARMA model etc. can face that the training time is long, is trapped in office when facing the training data of magnanimity higher-dimension
The problems such as portion's minimum, easy over-fitting.Meanwhile electric load is related to many hidden variables, such as illumination, wind-force, festivals or holidays,
These variables are typically difficult to obtain or quantify；Noise data is excluded from mass data, extraction effective information is also a difficulty
Point.With the development of intelligent grid, communication network technology and sensor technology, Demand-side have accumulated substantial amounts of basis can number
According to, while as the various new electricity consumption such as distributed power source, energy storage, electric automobile form occurs, demand side data is exponentially
Increase, traditional unit data mining pattern has been difficult to meet application request.
Existing big data processing framework can be divided into three classes, only batch processing framework Apache Hadoop, only stream process framework
Apache Storm, Apache Samza, wherein combination frame Apache Spark, Apache Flink, Hadoop and Spark
It is big data management and the parallel processing technique of current main flow.ApacheSpark is that a general large-scale data is quickly located
Manage engine, be the efficient big data analytical framework after ApacheHadoop, be class Hadoop MapReduce it is general simultaneously
Row system.Spark, which primarily focuses on, is calculated and is handled the fortune that Optimization Mechanism accelerates batch processing workload by perfect internal memory
Scanning frequency degree, it has merged a variety of computation schemas, covers streaming computing and internal memory calculating etc., and provides SQL and machine learning calculation
The support of Faku County, it can be achieved flexible integrated.Spark has been widely used in the Internet domain, such as Alibaba, Tengxun, hundred
The company such as degree and Yahoo has built the real-time big data platform based on Spark.
LSTM neutral nets are built upon a kind of new depth machine learning nerve net on recurrent neural network (RNN)
Network, a prolonged time lag is established between inputting, feeding back and preventing gradient from breaking out, if process dried layer nonlinear transformation increases
Strong model capacity-building.Using substantial amounts of training sample set, depth model can realize more forcing for high-dimension function in theory
Closely, the implicit information contained in data is excavated, there are powerful very high level conceptual feature capabilities.
The content of the invention
The purpose of the present invention is：Spark big datas platform is combined with LSTM neutral net load forecasting methods, avoided
Unit data mining pattern problem encountered.
In order to achieve the above object, the technical scheme is that to provide a kind of Demand-side load based on Spark pre-
Survey method, it is characterised in that comprise the following steps：
Step 1, gather target area sometime the history Power system load data in span and corresponding historical temperature number
According to formation raw data set；
Step 2, using Spark platforms raw data set is converted into elasticity distribution formula data set, comprised the following steps：
Step 2.1, using Spark platforms raw data set is converted into original elastic distributed data collection, i.e., original RDD
Data set；
Step 2.2, FlatMap operations are carried out to original RDD data sets, map it onto n (key, value) key assignments
It is right；
Step 2.3, each (key, value) key-value pair to step 2.2 carry out Map operations, according to Map mapping function journeys
Sequence logic re-starts mapping, forms new (key, value) key-value pair；
After step 2.4, (key, the value) key-value pair obtained to step 2.3 are grouped according to key, call pre-defined
Function pair is handled per component group, and a result key-value pair is returned per component group；
What step 2.5, (key, the value) key-value pair obtained according to key fields Connection Step 2.3 and step 2.4 obtained
As a result key-value pair, new RDD data sets are produced；
Step 2.6, the RDD data sets that step 2.5 obtains are buffered in internal memory；
Whether n (key, value) key-value pairs that step 2.7, judgment step 2.2 are formed all are disposed, if so, then
RDD data sets in internal memory are saved in file system, otherwise return to step 2.3；
The characteristic value of RDD data sets in step 3, extraction document system, and it is divided into training dataset and test data
Collection；
Step 4, based on Spark platforms, parallelization realization is carried out to LSTM neural network algorithms, utilizes training data set pair
LSTM neural network algorithms after parallelization are trained, and obtain forecast model；
Step 5, by forecast model test data set is predicted, and utilizes mean absolute percentage error and speed-up ratio
Carry out assessment models prediction effect.
Preferably, the mean absolute percentage error is MAPE, then has：
In formula, YkFor the predicted value of k-th of future position, ykFor k-th
The actual value of future position, n are the total number of future position.
Preferably, the speed-up ratio is Speedup, then have：
Speedup=t/T, in formula, t is the unit operation time, and T is cluster run time.
The present invention builds load forecasting model based on the LSTM neural net methods in deep learning field, can be accurate
Ground predicts electric load；Predicted based on Spark memory parallels Computational frame come parallel load, LSTM neural network algorithms are entered
Row parallelization, parallelization analysis is carried out to historical load data, shortens the load prediction time, improves LSTM neural network algorithms pair
The disposal ability of big data.
Brief description of the drawings
Fig. 1 is the load prediction system operation flow chart based on distributed memory Computational frame；
Fig. 2 is the model structure schematic diagram of single hidden layer LSTM neutral nets；
Fig. 3 is Spark parallel computation flow charts；
Fig. 4 is traditional load forecasting method and the LSTM load prediction results comparison diagrams based on Spark.
Embodiment
To become apparent the present invention, hereby with preferred embodiment, and accompanying drawing is coordinated to be described in detail below.
As shown in figure 1, a kind of Demand-side load forecasting method based on Spark provided by the invention comprises the following steps：
Step 1, data acquisition：Collection note in real time is carried out once to the load data of public institution of 20, certain city per hour
Record, collection period are on June 29,1 day to 2008 January in 2004, and each region is expected 39408 sample datas, for wherein
Excalation value is filled up, and ensures the completeness of initial data.
Step 2, data normalization processing：The data that step 1 collects are normalized, ensure different pieces of information model
The input data enclosed plays identical effect, and raw data set is formed after being stored in HDFS file system.The normalizing that the present invention uses
It is as follows to change formula：
In formula, x is the initial data for treating normalized, xmax、xminRespectively in initial data
Maximum and minimum value, xnormFor the data after normalized.
Step 2, using Spark platforms raw data set is converted into elasticity distribution formula data set, with reference to Fig. 3, including with
Lower step：
Step 2.1, Input stages：Raw data set is read from HDFS file system, and is converted to original elastic distribution
Data set, elasticity distribution formula data set is referred to as RDD data sets below.
Step 2.2, FlatMap stages：Original RDD data sets will be inputted it is mapped to 0 and arrives multiple output RDD data sets.Press
According to mapper logic, n (key, value) key-value pairs are mapped to.
Step 2.3, Map operational phases：According to Map mapping function programmed logics, step 2.2 is formed it is each (key,
Value) key-value pair re-starts mapping, forms new (key, value) key-value pair.The step is mainly step 2.4
The Reduce stages determine suitable key fields.
Step 2.4, Reduce stages：After (key, the value) key-value pair of step 2.3 is grouped according to key fields, call
Function is handled, one result key-value pair of every group of return.
Step 2.5, Join stages：(key, the value) key-value pair and step obtained according to key fields Connection Step 2.3
2.4 obtained result key-value pairs, produce new RDD data sets；
Step 2.6, Cache stages：The RDD data sets that step 2.5 obtains are buffered in internal memory；
Whether n (key, value) key-value pairs that step 2.7, judgment step 2.2 are formed all are disposed, if so, then
RDD data sets in internal memory are saved in file system, otherwise return to step 2.3.
The characteristic value of RDD data sets in step 3, extraction document system, and it is divided into training dataset and test data
Collection.
Step 4, based on Spark platforms, parallelization realization is carried out to LSTM neural network algorithms, utilizes training data set pair
LSTM neural network algorithms after parallelization are trained, and obtain forecast model.
LSTM neural network algorithms in the present invention use the LSTM neutral nets with an input layer, 3 hidden layers
Model, using dense functions, to perform recurrence, (such as Fig. 2 provides single hidden layer LSTM Artificial Neural Network Structures and illustrated output layer
Figure).Compared with common RNN, the hiding layer unit of LSTM neural network models has three doors：Input gate indicates whether that collection is new
Load data information be added in current concealed nodes, allow to input if value is 1, if value not allow if 0, so
Can abandons useless input information；Forget the historical load number that door indicates whether to retain current hiding node layer storage
According to, if value be 1 if retain, the historical load data that present node stores is emptied if value is by 0；Out gate indicates whether
Give present node output valve to next layer, exported if 1 present node and give next layer (next hidden layer or output
Layer), do not exported if 0 present node output valve.The specific calculation formula of LSTM units is as follows：
it=sigmoid (Wiht-1+Uixt)
ft=sigmoid (Wfht-1+Ufxt)
ot=sigmoid (Woht-1+Uoxt)
gt=tanh (Wght-1+Ugxt)
In formula, it、ft、otRepresent input gate, forget door and out gate, ctMnemon is represented, W represents recurrence connection
Weight, U represent input layer to the weight of hidden layer,The multiplication between element is represented,Addition between representative element, sigmoid
It is two kinds of activation primitives with tanh.Sigmoid activation primitive expression formulas are：Sigmoid activation primitives can be with
Input value is mapped in the range of [0,1].Tanh activation primitives are：Tanh activation primitives can be by input value
It is mapped in the range of [- 1,1].
The LSTM models that the present invention uses include an input layer, 3 hidden layers altogether, and output layer uses dense functions reality
Now full connection, each hidden layer have 100 units.Each minibatch size is 2, network structure training
10epoches.Mean_squared_error is used in training as loss function, using rmsprop as optimizer more
New algorithm.
Step 5, by forecast model test data set is predicted, and utilizes mean absolute percentage error and speed-up ratio
Carry out assessment models prediction effect.
Mean absolute percentage error is MAPE, then has：
In formula, YkFor the predicted value of k-th of future position, ykFor k-th
The actual value of future position, n are the total number of future position.In load forecast, MAPE values are smaller, and predicted load is more accurate.
Speed-up ratio is Speedup, then have：
Speedup=t/T, in formula, t is the unit operation time, and T is cluster run time.Using speed-up ratio measure algorithm simultaneously
The performance of rowization, speed-up ratio is bigger, and concurrency is better.
The present invention lifts whole short-term load forecasting stream based on distributed memory Computational frame Spark structure forecast models
The efficiency of journey.The framework proposes main memory cluster calculating, by the way that data set is cached in internal memory, reduces the I/O operation of data,
So as to improve reading and writing data speed.
The present invention use Apache Spark Distributed Architecture, can be with based on the original power consumption data of Spark platforms reading
Extensive electric power data is read from local or HDFS, improves read-write and the storage capacity of mass data, utilizes its internal memory meter
The advantage of calculation can realize more efficient, quick load prediction.LSTM neural net methods of the present invention are fully sharp
With the relation between historical data, load prediction precision and efficiency are improved, while there is stronger generalization ability, has one
Fixed social value and realize meaning.
Claims (3)
1. a kind of Demand-side load forecasting method based on Spark, it is characterised in that comprise the following steps：
Step 1, gather target area sometime the history Power system load data in span and corresponding historical temperature data, shape
Into raw data set；
Step 2, using Spark platforms raw data set is converted into elasticity distribution formula data set, comprised the following steps：
Step 2.1, using Spark platforms raw data set is converted into original elastic distributed data collection, i.e., original RDD data
Collection；
Step 2.2, FlatMap operations are carried out to original RDD data sets, map it onto n (key, value) key-value pairs；
Step 2.3, each (key, value) key-value pair to step 2.2 carry out Map operations, are patrolled according to Map mapping function programs
Collect and re-start mapping, form new (key, value) key-value pair；
After step 2.4, (key, the value) key-value pair obtained to step 2.3 are grouped according to key, pre-defined function is called
Every component group is handled, a result key-value pair is returned per component group；
The result that step 2.5, (key, the value) key-value pair obtained according to key fields Connection Step 2.3 and step 2.4 obtain
Key-value pair, produce new RDD data sets；
Step 2.6, the RDD data sets that step 2.5 obtains are buffered in internal memory；
Whether n (key, value) key-value pairs that step 2.7, judgment step 2.2 are formed all are disposed, if so, then by
The RDD data sets deposited are saved in file system, otherwise return to step 2.3；
The characteristic value of RDD data sets in step 3, extraction document system, and it is divided into training dataset and test data set；
Step 4, based on Spark platforms, parallelization realization is carried out to LSTM neural network algorithms, it is parallel using training data set pair
LSTM neural network algorithms after change are trained, and obtain forecast model；
Step 5, by forecast model test data set is predicted, and commented using mean absolute percentage error and speed-up ratio
Estimate forecast result of model.
2. a kind of Demand-side load forecasting method based on Spark as claimed in claim 1, it is characterised in that described average
Absolute percent error is MAPE, then has：
In formula, YkFor the predicted value of k-th of future position, ykFor k-th of prediction
The actual value of point, n are the total number of future position.
A kind of 3. Demand-side load forecasting method based on Spark as claimed in claim 1, it is characterised in that the acceleration
Than for Speedup, then have：
Speedup=t/T, in formula, t is the unit operation time, and T is cluster run time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711004651.8A CN107704970A (en) | 2017-10-24 | 2017-10-24 | A kind of Demand-side load forecasting method based on Spark |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711004651.8A CN107704970A (en) | 2017-10-24 | 2017-10-24 | A kind of Demand-side load forecasting method based on Spark |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107704970A true CN107704970A (en) | 2018-02-16 |
Family
ID=61182276
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711004651.8A Pending CN107704970A (en) | 2017-10-24 | 2017-10-24 | A kind of Demand-side load forecasting method based on Spark |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107704970A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846549A (en) * | 2018-05-08 | 2018-11-20 | 天津大学 | Power grid power supply and distribution based on intelligent appliance end regulates and controls method |
| CN109032671A (en) * | 2018-06-25 | 2018-12-18 | 电子科技大学 | A kind of distributed deep learning method and system based on data parallel strategy |
| CN109255505A (en) * | 2018-11-20 | 2019-01-22 | 国网辽宁省电力有限公司经济技术研究院 | A kind of short-term load forecasting method of multi-model fused neural network |
| CN109285589A (en) * | 2018-10-31 | 2019-01-29 | 重庆邮电大学 | A kind of aluminium electroloysis degree of superheat prediction technique based on Spark big data platform |
| CN111178587A (en) * | 2019-12-06 | 2020-05-19 | 广东工业大学 | Spark framework-based short-term power load rapid prediction method |
| CN111178612A (en) * | 2019-12-19 | 2020-05-19 | 绍兴大明电力设计院有限公司 | LSTM load prediction method of grid user based on big data ODPS engine |
| CN111860621A (en) * | 2020-06-30 | 2020-10-30 | 贵州民族大学 | Data-driven distributed traffic flow prediction method and system |
| CN112070280A (en) * | 2020-08-19 | 2020-12-11 | 贵州民族大学 | Real-time traffic flow parallel prediction method, system, terminal and storage medium |
| CN112418496A (en) * | 2020-11-10 | 2021-02-26 | 国网四川省电力公司经济技术研究院 | Power distribution station energy storage configuration method based on deep learning |
-
2017
- 2017-10-24 CN CN201711004651.8A patent/CN107704970A/en active Pending
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108846549A (en) * | 2018-05-08 | 2018-11-20 | 天津大学 | Power grid power supply and distribution based on intelligent appliance end regulates and controls method |
| CN109032671A (en) * | 2018-06-25 | 2018-12-18 | 电子科技大学 | A kind of distributed deep learning method and system based on data parallel strategy |
| CN109032671B (en) * | 2018-06-25 | 2022-05-03 | 电子科技大学 | Distributed deep learning method and system based on data parallel strategy |
| CN109285589A (en) * | 2018-10-31 | 2019-01-29 | 重庆邮电大学 | A kind of aluminium electroloysis degree of superheat prediction technique based on Spark big data platform |
| CN109255505B (en) * | 2018-11-20 | 2021-09-24 | 国网辽宁省电力有限公司经济技术研究院 | Short-term load prediction method of multi-model fusion neural network |
| CN109255505A (en) * | 2018-11-20 | 2019-01-22 | 国网辽宁省电力有限公司经济技术研究院 | A kind of short-term load forecasting method of multi-model fused neural network |
| CN111178587B (en) * | 2019-12-06 | 2022-11-22 | 广东工业大学 | Spark framework-based short-term power load rapid prediction method |
| CN111178587A (en) * | 2019-12-06 | 2020-05-19 | 广东工业大学 | Spark framework-based short-term power load rapid prediction method |
| CN111178612A (en) * | 2019-12-19 | 2020-05-19 | 绍兴大明电力设计院有限公司 | LSTM load prediction method of grid user based on big data ODPS engine |
| CN111178612B (en) * | 2019-12-19 | 2023-06-09 | 绍兴大明电力设计院有限公司 | LSTM load prediction method of grid user based on big data ODPS engine |
| CN111860621A (en) * | 2020-06-30 | 2020-10-30 | 贵州民族大学 | Data-driven distributed traffic flow prediction method and system |
| CN111860621B (en) * | 2020-06-30 | 2024-05-03 | 贵州民族大学 | Data-driven distributed traffic flow prediction method and system |
| CN112070280A (en) * | 2020-08-19 | 2020-12-11 | 贵州民族大学 | Real-time traffic flow parallel prediction method, system, terminal and storage medium |
| CN112070280B (en) * | 2020-08-19 | 2023-10-31 | 贵州民族大学 | Real-time traffic flow parallel prediction method, system, terminal and storage medium |
| CN112418496A (en) * | 2020-11-10 | 2021-02-26 | 国网四川省电力公司经济技术研究院 | Power distribution station energy storage configuration method based on deep learning |
| CN112418496B (en) * | 2020-11-10 | 2022-04-22 | 国网四川省电力公司经济技术研究院 | Power distribution station energy storage configuration method based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107704970A (en) | A kind of Demand-side load forecasting method based on Spark | |
| CN109034448B (en) | Trajectory prediction method based on vehicle trajectory semantic analysis and deep belief network | |
| Ke et al. | Short-term electrical load forecasting method based on stacked auto-encoding and GRU neural network | |
| CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
| Pak et al. | A hybrid model based on convolutional neural networks and long short-term memory for ozone concentration prediction | |
| CN110782093B (en) | PM fusing SSAE deep feature learning and LSTM2.5Hourly concentration prediction method and system | |
| Ju et al. | Division-based rainfall-runoff simulations with BP neural networks and Xinanjiang model | |
| Fu et al. | Prediction of particular matter concentrations by developed feed-forward neural network with rolling mechanism and gray model | |
| CN113554466B (en) | Short-term electricity consumption prediction model construction method, prediction method and device | |
| Lv et al. | A long Short-Term memory cyclic model with mutual information for hydrology forecasting: A Case study in the xixian basin | |
| CN111080032A (en) | Load prediction method based on Transformer structure | |
| CN110148296A (en) | A kind of trans-city magnitude of traffic flow unified prediction based on depth migration study | |
| CN106599520A (en) | LSTM-RNN model-based air pollutant concentration forecast method | |
| CN110212528A (en) | Reconstructing method is lacked based on the power distribution network metric data for generating confrontation and dual Semantic Aware | |
| Salam et al. | Energy consumption prediction model with deep inception residual network inspiration and LSTM | |
| Xu et al. | A graph spatial-temporal model for predicting population density of key areas | |
| CN112419711B (en) | Closed parking lot parking demand prediction method based on improved GMDH algorithm | |
| Huang et al. | Research on urban modern architectural art based on artificial intelligence and GIS image recognition system | |
| CN117314266B (en) | Novel intelligent scientific and technological talent evaluation method based on hypergraph attention mechanism | |
| CN114565187A (en) | Traffic network data prediction method based on graph space-time self-coding network | |
| CN104199870A (en) | Method for building LS-SVM prediction model based on chaotic search | |
| CN116681176A (en) | Traffic flow prediction method based on clustering and heterogeneous graph neural network | |
| Feng et al. | A dual-staged attention based conversion-gated long short term memory for multivariable time series prediction | |
| CN115640901A (en) | Small sample load prediction method based on hybrid neural network and generation countermeasure | |
| Caihong et al. | Real-time flood classification forecasting based on k-means++ clustering and neural network |
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: 20180216 |
|
| RJ01 | Rejection of invention patent application after publication |