CN108493933A - A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms - Google Patents
A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms Download PDFInfo
- Publication number
- CN108493933A CN108493933A CN201810331744.XA CN201810331744A CN108493933A CN 108493933 A CN108493933 A CN 108493933A CN 201810331744 A CN201810331744 A CN 201810331744A CN 108493933 A CN108493933 A CN 108493933A
- Authority
- CN
- China
- Prior art keywords
- depth
- tree
- electric load
- depth decision
- 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
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms, it includes：Power consumer load profile is acquired using harvester and influences the factor data of electric load variation；The power consumer part throttle characteristics data and influence electric load changing factor data that are acquired are pre-processed, training data sample set and test set are established；The parameter of depth decision-tree model, i.e., the initial forest model parameter in depth decision Tree algorithms are set；Depth decision-tree model is trained using training data sample set；The depth decision-tree model after training is tested using test set, determines depth decision-tree model depth；After being trained and being tested the depth decision-tree model completed, the observation Value Data for needing to carry out part throttle characteristics excavation, output load Predicting Performance Characteristics result are inputted into model；Reach the target that depth excavation is carried out to power system load characteristic, to instruct electric power enterprise sacurity dispatching and stable operation.
Description
Technical field
The invention belongs to analysis of Power Load Characteristic fields, more particularly to a kind of electric power based on depth decision Tree algorithms
Part throttle characteristics method for digging.
Background technology
Electric system should provide safe and reliable, standardized electric energy to all types of user, and the moment meets power consumer i.e.
The electrical demand of load.As social economy's fast development, industrial structure upgrading, global climate environmental change and the people give birth to
Running water is flat to be continuously improved, and larger variation has occurred in Characteristics of Electric Load than before.This to electric system ensure power balance, when
It carves safe and stable operation and produces impact.In order to cope with such a situation, the excavation of depth is carried out to Characteristics of Electric Load,
The changing rule for holding Characteristics of Electric Load under the new situation is a measure for effectively solving the problems, such as this.
The conventional method of analysis of Power Load Characteristic is often statistical analysis method, can not in depth to Characteristics of Electric Load into
Capable in depth mining analysis.Rough power system load property sort and simple analysis result can only be obtained.This method consumption
When it is laborious, the variation of fast-changing Characteristics of Electric Load can not be timely responded to.Therefore a method rapidly and efficiently is found to come pair
It is a urgent problem to be solved that power system load characteristic, which excavate,.
Invention content
The technical problem to be solved by the present invention is to：A kind of Characteristics of Electric Load excavation based on depth decision Tree algorithms is provided
Method extraction to basic Characteristics of Electric Load and is successively handled by using depth decision Tree algorithms, to reach to electric power
System load characteristic carries out the target of depth excavation, to instruct electric power enterprise sacurity dispatching and stable operation.
The technical scheme is that：
A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms, it includes the following steps：
Step S110, it acquires power consumer load profile using harvester and influences the factor of electric load variation
Data；
Step S120, the power consumer part throttle characteristics data and influence electric load changing factor data that are acquired are carried out
Pretreatment, establishes training data sample set and test set；
Step S130, the parameter of depth decision-tree model is set, i.e., the initial forest model ginseng in depth decision Tree algorithms
Number；
Step S140, depth decision-tree model is trained using training data sample set；
Step S150, the depth decision-tree model after training is tested using test set, determines depth decision tree mould
Moldeed depth degree；
Step S160, it after being trained and being tested the depth decision-tree model completed, inputs and is born into model
The observation Value Data that lotus characteristic is excavated, output load Predicting Performance Characteristics result.
The harvester includes data acquisition analysis system (SCADA), Wide Area Measurement System (WAMS) and failure record
Wave monitors system (FRMS).
The power consumer load profile includes：The part throttle characteristics and frequency domain power curve frequency spectrum in timesharing domain；It is described
Influencing the factor data that electric load changes includes：It is the max. daily temperature of power consumer their location, Daily minimum temperature, per day
Temperature, rainfall, air humidity and date property.
The parameter of the depth decision-tree model includes：Decision tree generating mode, the number of decision tree, random attribute
Number, decision tree depth capacity, leaf node at least record number and leaf node at least records percentage.
The method being trained to depth decision-tree model using training data sample set includes：Including electric load
Data characteristic extracts mining process：Further feature extraction is carried out using decision tree to the electric load characteristic of existing collection,
Obtain more representative characteristic attribute；Using depth decision tree structure, to extracting the more representative feature category obtained
Property carry out depth Characteristics of Electric Load excavate.
It is described that the depth decision-tree model after training is tested using test set, determine depth decision-tree model depth
Method be：For the depth decision-tree model currently formed, the ability of existing model is tested using test set, depth
Decision Tree algorithms training process continues to increase depth, until the ability of model no longer improves.
Advantageous effect of the present invention：
The present invention to carry out Characteristics of Electric Load the excavation of depth by using depth decision Tree algorithms.First by existing
Some Power System Intelligent harvesters are on magnanimity power consumer load profile and influence many of electric load variation
Factor data is acquired.Secondly collected Power system load data and other factors data are pre-processed, establishes instruction
Practice data set.Established training data set pair depth decision Tree algorithms are utilized to be trained again.Finally completed using training
Depth decision Tree algorithms to power consumer part throttle characteristics carry out intelligent excavating.In addition, used depth decision Tree algorithms are one
The newer machine learning Classification Algorithms in Data Mining of kind, which has without a large amount of setting hyper parameters, voluntarily determining model is deep
The advantages of spending, you can to effectively carrying out intelligent excavating to Characteristics of Electric Load, acquired results can serve power grid enterprises' scheduling, fortune
The various aspects such as row.To be conducive to improve the economic benefit of electric power enterprise.
Advantage and effect of the present invention：
(1) the Characteristics of Electric Load method for digging based on depth decision Tree algorithms that the present invention designs, it is contemplated that novel
Characteristics of Electric Load index, rather than foundation of the previous simple time domain specification index as analysis Characteristics of Electric Load.
(2) the Characteristics of Electric Load method for digging based on depth decision Tree algorithms that the present invention designs, using depth decision
Tree algorithm, this is a kind of newer machine learning data mining algorithm, relative to other data mining algorithms, such as artificial neural network
Network (Artificial Neural Network, ANN), fuzzy neural network (Fuzzy Neural Network, FNN) algorithm
Method for solving, have many advantages, such as that training speed is fast, can with parallel training and excavate, more characterization ability.
Description of the drawings
Fig. 1 is the method flow diagram of the Characteristics of Electric Load method for digging based on depth decision Tree algorithms of the present invention.
Specific implementation mode
The present invention is based on the Characteristics of Electric Load method for digging of depth decision Tree algorithms to include the following steps：
Step S110, by existing Power System Intelligent harvester to magnanimity power consumer load profile and
The factors data for influencing electric load variation are acquired.Above-mentioned Power System Intelligent harvester includes：Data are adopted
The intelligent acquisitions device such as collection and monitoring system (SCADA), Wide Area Measurement System (WAMS) and fault recording and monitoring system (FRMS).
Power consumer load profile includes：The part throttle characteristics in timesharing domain, frequency domain power curve frequency spectrum.Influence electric load variation
Factor data includes：Max. daily temperature, Daily minimum temperature, mean daily temperature, rainfall, the air of power consumer their location are wet
Degree and date property etc..
Step S120, on the power consumer part throttle characteristics data acquired and the factors number for influencing electric load variation
According to being pre-processed, training data sample set and test set are established.Pretreatment is mainly the normalized of numeric data, is eliminated
Dimension impact between index.It is described in detail below：
Wherein xminFor the maximum value of sample data, xmaxFor the minimum value of sample data, x*To return to sample data
One result changed；X is sample data.
For categorical datas such as date properties, need numeric data is manually set and correspond therewith, then proceed to according to
The normalized mode of numeric data pre-processes data.
Training data sample set and test set respectively account for the 80% and 20% of sample total.
Step S130 needs that depth decision-tree model parameter is voluntarily arranged according to user, i.e., in depth decision Tree algorithms
Initial forest model parameter.The parameter of depth decision Tree algorithms includes：Decision tree generating mode, the number of decision tree, random category
Property number, decision tree depth capacity, leaf node at least records number, leaf node at least records percentage.Design parameter is as follows：
Algorithm types：ID3 algorithms, CART algorithms, C4.5 algorithms
The number of decision tree：Acquiescence 500, ranging from (0,1000]；
Random attribute number：Single tree selects optimal characteristics attribute, random feature number every time when generating.
It is attribute sum that alternative type, which has logN, N/3, sqrtN, tetra- types of N, wherein N,；
Set depth capacity：The depth capacity of single tree, and range [1, ∞), -1 indicates growth completely；
Leaf node at least records number：The minimum number of (optional) leaf segment point data.Minimum number is 2；
Leaf node at least records percentage：(optional) leaf node data amount check accounts for the minimum scale of father node, range [0,
100], -1 no limitation, acquiescence -1 are indicated.
Step S140 is trained depth decision-tree model using training data sample set, initially forms comprising electric power
Load data feature extraction mining process and the depth decision-tree model successively handled with certain depth.It specifically describes such as
Under：
(1) Power system load data feature extraction mining process：
Value window is set, sliding value is carried out to electric load sequence data, each small fragment obtained by value is defeated
Enter the forest model being made of more decision trees to be handled, obtain the Power system load data characteristic fragment with enhancing characteristic,
And by all characteristic fragment assemblies, one enhancing that characterization ability is had more than raw power load data of output is vectorial.
(2) the depth decision-tree model successively handled with certain depth：
The depth decision-tree model input enhancing vector completed to training, each layer of decision tree set will generate enhancing vector
As next layer of input.It successively handles, information is successively transmitted, until obtaining final result.
Step S150 tests the depth decision-tree model currently formed using test set, automatically determines depth and determine
Plan tree-model depth.Continue to increase model depth if accuracy rate increases, otherwise stops the depth for increasing depth decision tree.
Step S160 after obtaining the depth decision-tree model of training completion, is inputted into model and is needed to carry out part throttle characteristics
The observation Value Data of excavation, final output part throttle characteristics Result.
It can be obtained by for the part throttle characteristics corresponding to new electric load observation, by this by above step
The accurate assurance of part throttle characteristics can serve the various aspects such as power grid enterprises' scheduling, operation.To be conducive to improve electric power enterprise
Economic benefit.
Claims (6)
1. a kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms, it includes the following steps：
Step S110, using harvester acquire power consumer load profile and influence electric load variation because of prime number
According to；
Step S120, the power consumer part throttle characteristics data and influence electric load changing factor data that are acquired are located in advance
Reason, establishes training data sample set and test set；
Step S130, the parameter of depth decision-tree model, i.e., the initial forest model parameter in depth decision Tree algorithms are set；
Step S140, depth decision-tree model is trained using training data sample set；
Step S150, the depth decision-tree model after training is tested using test set, determines that depth decision-tree model is deep
Degree；
Step S160, it after being trained and being tested the depth decision-tree model completed, is inputted into model and needs to carry out load spy
Property excavate observation Value Data, output load Predicting Performance Characteristics result.
2. a kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms according to claim 1, feature
It is：The harvester includes data acquisition analysis system SCADA, Wide Area Measurement System WAMS and failure wave-recording monitoring system
Unite FRMS.
3. a kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms according to claim 1, feature
It is：The power consumer load profile includes：The part throttle characteristics and frequency domain power curve frequency spectrum in timesharing domain；The influence
Electric load variation factor data include：The max. daily temperature of power consumer their location, Daily minimum temperature, per day temperature
Degree, rainfall, air humidity and date property.
4. a kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms according to claim 1, feature
It is：The parameter of the depth decision-tree model includes：Decision tree generating mode, random attribute number, is determined at the number of decision tree
Plan tree depth capacity, leaf node at least record number and leaf node at least records percentage.
5. a kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms according to claim 1, feature
It is：The method being trained to depth decision-tree model using training data sample set includes：Including electric load number
According to feature extraction mining process：Further feature extraction is carried out using decision tree to the electric load characteristic of existing collection, is obtained
Obtain more representative characteristic attribute；Using depth decision tree structure, to extracting the more representative characteristic attribute obtained
The Characteristics of Electric Load for carrying out depth excavates.
6. a kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms according to claim 1, feature
It is：It is described that the depth decision-tree model after training is tested using test set, determine depth decision-tree model depth
Method is：For the depth decision-tree model currently formed, the ability of existing model is tested using test set, depth is determined
Plan tree algorithm training process continues to increase depth, until the ability of model no longer improves.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810331744.XA CN108493933A (en) | 2018-04-13 | 2018-04-13 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810331744.XA CN108493933A (en) | 2018-04-13 | 2018-04-13 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108493933A true CN108493933A (en) | 2018-09-04 |
Family
ID=63315870
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810331744.XA Pending CN108493933A (en) | 2018-04-13 | 2018-04-13 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108493933A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109214578A (en) * | 2018-09-19 | 2019-01-15 | 平安科技（深圳）有限公司 | Electronic device, the building electro-load forecast method based on decision-tree model and storage medium |
| CN111404150A (en) * | 2020-03-30 | 2020-07-10 | 广西电网有限责任公司电力科学研究院 | Transient stability assessment method and system suitable for large power grid operation |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184402A (en) * | 2015-08-31 | 2015-12-23 | 国家电网公司 | Personalized user short-term load forecasting algorithm based on decision-making tree |
| CN106096748A (en) * | 2016-04-28 | 2016-11-09 | 武汉宝钢华中贸易有限公司 | Entrucking forecast model in man-hour based on cluster analysis and decision Tree algorithms |
| CN108011367A (en) * | 2017-12-04 | 2018-05-08 | 贵州电网有限责任公司电力科学研究院 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
-
2018
- 2018-04-13 CN CN201810331744.XA patent/CN108493933A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184402A (en) * | 2015-08-31 | 2015-12-23 | 国家电网公司 | Personalized user short-term load forecasting algorithm based on decision-making tree |
| CN106096748A (en) * | 2016-04-28 | 2016-11-09 | 武汉宝钢华中贸易有限公司 | Entrucking forecast model in man-hour based on cluster analysis and decision Tree algorithms |
| CN108011367A (en) * | 2017-12-04 | 2018-05-08 | 贵州电网有限责任公司电力科学研究院 | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms |
Non-Patent Citations (1)
| Title |
|---|
| 郑凯文 等: "基于迭代决策树(GBDT) 短期负荷预测研究", 《贵州电力技术》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109214578A (en) * | 2018-09-19 | 2019-01-15 | 平安科技（深圳）有限公司 | Electronic device, the building electro-load forecast method based on decision-tree model and storage medium |
| CN111404150A (en) * | 2020-03-30 | 2020-07-10 | 广西电网有限责任公司电力科学研究院 | Transient stability assessment method and system suitable for large power grid operation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108011367A (en) | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms | |
| CN110097297B (en) | Multi-dimensional electricity stealing situation intelligent sensing method, system, equipment and medium | |
| CN111738462B (en) | Fault first-aid repair active service early warning method for electric power metering device | |
| CN104008143B (en) | Vocational ability index system establishment method based on data mining | |
| CN107633265A (en) | For optimizing the data processing method and device of credit evaluation model | |
| CN107169145A (en) | A kind of method of user's stealing menace level detection based on clustering algorithm | |
| CN106372748A (en) | Hard-rock tunnel boring machine boring efficiency prediction method | |
| CN112149873B (en) | Low-voltage station line loss reasonable interval prediction method based on deep learning | |
| CN105404944A (en) | Big data analysis method for warning of heavy-load and overload of electric power system | |
| CN111210170B (en) | Environment-friendly management and control monitoring and evaluation method based on 90% electricity distribution characteristic index | |
| CN109190890A (en) | A kind of user behavior analysis method based on custom power consumption data | |
| CN106446016A (en) | Parallel association rule mining-based method for predicting running reliability of power distribution network | |
| CN105654196A (en) | Adaptive load prediction selection method based on electric power big data | |
| CN107909208A (en) | Damage method drops in a kind of taiwan area distribution | |
| CN108345908A (en) | Sorting technique, sorting device and the storage medium of electric network data | |
| CN102750367A (en) | Big data checking system and method thereof on cloud platform | |
| CN110348683A (en) | The main genetic analysis method, apparatus equipment of electrical energy power quality disturbance event and storage medium | |
| CN104850868A (en) | Customer segmentation method based on k-means and neural network cluster | |
| CN107437135A (en) | A kind of novel energy-storing selection method | |
| CN113516192A (en) | Method, system, device and storage medium for identifying user electricity consumption transaction | |
| CN107871183A (en) | Permafrost Area highway distress Forecasting Methodology based on uncertain Clouds theory | |
| CN108493933A (en) | A kind of Characteristics of Electric Load method for digging based on depth decision Tree algorithms | |
| CN107608938B (en) | Factor screening method for binary classification based on enhanced regression tree algorithm | |
| CN113435733A (en) | Water resource risk assessment and prediction system | |
| CN114676931B (en) | Electric quantity prediction system based on data center technology |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180904 |
|
| WD01 | Invention patent application deemed withdrawn after publication |