CN107391689A - Wind power output abnormal data reconstructing method based on big data technology - Google Patents

Wind power output abnormal data reconstructing method based on big data technology Download PDF

Info

Publication number
CN107391689A
CN107391689A CN201710612234.5A CN201710612234A CN107391689A CN 107391689 A CN107391689 A CN 107391689A CN 201710612234 A CN201710612234 A CN 201710612234A CN 107391689 A CN107391689 A CN 107391689A
Authority
CN
China
Prior art keywords
mrow
msub
data
layer
blower fan
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
Application number
CN201710612234.5A
Other languages
Chinese (zh)
Inventor
宁文元
孙荣富
王靖然
王若阳
徐海翔
郭子明
张岩
钱苏晋
李亮
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
BEIJING TECHSTAR TECHNOLOGY Co Ltd
State Grid Corp of China SGCC
State Grid Jibei Electric Power Co Ltd
Original Assignee
BEIJING TECHSTAR TECHNOLOGY Co Ltd
State Grid Corp of China SGCC
State Grid Jibei Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by BEIJING TECHSTAR TECHNOLOGY Co Ltd, State Grid Corp of China SGCC, State Grid Jibei Electric Power Co Ltd filed Critical BEIJING TECHSTAR TECHNOLOGY Co Ltd
Priority to CN201710612234.5A priority Critical patent/CN107391689A/en
Publication of CN107391689A publication Critical patent/CN107391689A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/10—File systems; File servers
    • G06F16/18—File system types
    • G06F16/182—Distributed file systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/25—Integrating or interfacing systems involving database management systems
    • G06F16/258—Data format conversion from or to a database
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/28—Databases characterised by their database models, e.g. relational or object models
    • G06F16/284—Relational databases
    • 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]

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Molecular Biology (AREA)
  • Public Health (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Biophysics (AREA)
  • Water Supply & Treatment (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Power Engineering (AREA)
  • Wind Motors (AREA)

Abstract

The present invention provides a kind of wind power output abnormal data reconstructing method based on big data technology, including：The distribution of substantial amounts of wind power output initial data is stored in HDFS different nodes；Several wind power output initial data for needing to carry out data reconstruction are read from HDFS；Each distribution process task uses parallel computation mode, and circular treatment distributes to each wind power output initial data in the data block of itself, abnormal data reconstruct is carried out to it.Advantage is：Can accurately, efficiently, reduction reconstruct quickly be carried out to the exceptional data point in wind-powered electricity generation history data, so as to be advantageous to the parameters such as the annual electricity generating capacity of the performance of accurate evaluation blower fan and wind power plant and operation conditions, prediction blower fan and wind power plant.

Description

Wind power output abnormal data reconstructing method based on big data technology
Technical field
The invention belongs to abnormal data reconfiguration technique field, and in particular to a kind of wind power output based on big data technology is different Regular data reconstructing method.
Background technology
With the arrival in electric power big data epoch, there is to outer energy number in industry in application process for electric power big data According to a large amount of association analysis demands of the diversiform datas such as, weather data, and these all directly results in the increasing of electric power data type Add, so as to substantially increase the complexity of electric power big data.Power industry will surpass significantly to the demand of big data, its urgency More other basic energy resource industries.Electric power big data will pass through the links such as future electrical energy industrial production and management, play uniqueness And huge effect.In a large amount of accesses of the new energy such as power generation link, wind-light storage, the electricity of traditional relative " static state " is broken Power produces so that the metering and management of power generation become increasingly sophisticated.
Wind-powered electricity generation service data is the important component of electric power big data.Currently, wind-power electricity generation is regenerative resource exploitation Technology is most ripe in utilization, the most forms of electricity generation of exploitation scale and commercialized development prospect.With wind generating technology not Disconnected development, randomness, fluctuation and the intermittence presented by wind power, shadow of the wind-powered electricity generation large-scale grid connection to power system Ring also more and more obvious, influence of the research wind-powered electricity generation to system simultaneously proposes that correlation technique solution has turned into currently on this basis Study hotspot and important topic.
Wind-powered electricity generation history data is research wind power wave characteristic, wind power prediction, wind power to power network shadow Assessment and control strategy, the calculating of wind power curve, test and the basis of amendment are rung, therefore, to wind power plant history run The record of data is with regard to particularly important.
For wind-power electricity generation operation management side, the power curve obtained by surveying wind speed and wind power is to weigh wind-powered electricity generation The optimal scale of unit and wind power plant economic and technique level, can be used for assessing the performance and operation conditions of blower fan and wind power plant, Predict annual electricity generating capacity of blower fan and wind power plant etc.；For predicting service provider, history actual measurement wind speed and wind power data are as pre- The basic input data of examining system, has a significant impact to precision of prediction.
However, generally comprising exceptional data point in the mass data being collected into from wind power plant, such abnormity point is shown Shortage of data, out-of-limit, the features such as power swing gradient is excessive.By the analysis of operating experience and statistical result to wind power plant, The main source of exceptional data point has：Fan blade is influenceed to form abnormity point by environmental factors such as dirt and ice is embodied in wind-powered electricity generation In the power curve of field；Wind power plant caused by wind is rationed the power supply is repaired or abandons to shut down；Sensor fault dispatch control system problem；Surveyed by meter Abnormal observation etc. caused by measuring error.
Therefore, how reduction reconstruct is carried out to the exceptional data point in wind-powered electricity generation history data, is accurate evaluation blower fan The key of the annual electricity generating capacity of performance and operation conditions, prediction blower fan and wind power plant with wind power plant etc., in the prior art without phase Close solution.
The content of the invention
The defects of existing for prior art, the present invention provide a kind of wind power output abnormal data based on big data technology Reconstructing method, can effectively it solve the above problems.
The technical solution adopted by the present invention is as follows：
The present invention provides a kind of wind power output abnormal data reconstructing method based on big data technology, comprises the following steps：
Step 1, the distribution of substantial amounts of wind power output initial data is stored in HDFS different nodes；Wherein, the wind Electric output initial data includes data label, blower fan ID, power plant ID, region ID, data acquisition date, data acquisition moment, reality Border active power value, acquisition mode, collector, blower fan wind speed, blower fan wind direction, atmospheric density and fan condition；
Step 2, several wind power output initial data for needing to carry out data reconstruction are read from HDFS；Will be described some Individual wind power output original data division is several data blocks, and each data block is distributed into a corresponding distribution process Task；
Each distribution process task uses parallel computation mode, and circular treatment is distributed to each in the data block of itself Individual wind power output initial data, abnormal data reconstruct is carried out to it；
Wherein, enter in the following ways for an arbitrary wind power output initial data, corresponding distribution process task Row abnormal data reconstructs：
Step 2.1, the distribution process task recognition is to fan condition；Wherein, the fan condition includes following nine kinds State：Treat wind state, generating state, drop forelock electricity condition, planned outage state, unplanned outage state, scheduling stoppage in transit state, Stoppage in transit state of being involved in communicating interrupt state, field and stoppage in transit state of being involved outside the venue；
Step 2.2, if it is communicating interrupt state to recognize fan condition, data are not reconstructed, wait pending data It is reconstructed again after given state；
If fan condition is recognized to treat wind state, planned outage state, unplanned outage state, scheduling stoppage in transit shape Stoppage in transit state of being involved in state, field or stoppage in transit state of being involved outside the venue, then be reconstructed into 0 value by the actual active power value of blower fan；
If recognizing fan condition as generating state or drop forelock electricity condition value, step 2.3 is performed；
Step 2.3, according to blower fan ID, the BP neural network model after training corresponding with blower fan is transferred, based on the instruction BP neural network model after white silk, blower fan theory active power value is calculated；
Step 2.4, if it is generating state to recognize fan condition, the actual active power value of blower fan is reconstructed into wind Mechanism opinion active power value；
If recognizing fan condition as drop forelock electricity condition value, power of fan dispatch value is further obtained；Then, sentence Whether disconnected blower fan theory active power value is more than or equal to power of fan dispatch value, if it is, the actual active power by blower fan Value is reconstructed into power of fan dispatch value；If not, the actual active power value of blower fan is reconstructed into blower fan theory active power value.
Preferably, in step 2.3, the BP neural network model after training corresponding with blower fan obtains by the following method：
Step 2.3.1, establish BP neural network model；The topological structure of the BP neural network model include input layer, Hidden layer and output layer；Wherein, the neuron number of input layer is n, and the neuron number of hidden layer is l, the nerve of output layer First number is m；Any input layer is xi, i ∈ (1,2 ... n)；Any hidden layer neuron is hj, j ∈ (1,2 ... l)；Appoint Output layer neuron of anticipating is ok, k ∈ (1,2 ... m)；
Step 2.3.2, the basic parameter of BP neural network model is initialized, including：Learning rate μ, input layer are to implicit The weight w of layerij, hidden layer to output layer weight wjk, input layer to hidden layer biasing number aj, hidden layer to the inclined of output layer Put several bkAnd excitation function t (x)；Wherein, weight w of the input layer to hidden layerij, hidden layer to output layer weight wjk, it is defeated Enter layer to the biasing number a of hidden layerj, hidden layer to output layer biasing number bkInitialization value is the random number in (- 1,1)；
Wherein：Weight w of the input layer to hidden layerijImplication is：Arbitrary input layer xiTo arbitrary hidden layer Neuron hjBetween weight；Weight w of the hidden layer to output layerjkImplication is：Arbitrary hidden layer neuron hjTo arbitrary Output layer neuron okBetween weight；Biasing number a of the input layer to hidden layerjImplication is：Each input layer is to arbitrarily Hidden layer neuron hjBiasing number；Biasing number b of the hidden layer to output layerkImplication is：Each hidden layer neuron is to arbitrary defeated Go out a layer neuron okBiasing number
Step 2.3.3, obtain training sample data；Wherein, the training sample data are separate unit blower fan history run shape The three-dimensional data of state, including：Blower fan wind speed, blower fan wind direction and atmospheric density；
Using the training sample data, based on the basic parameter of the initialization BP neural network model in step 2.3.2, The BP neural network model is trained, the BP neural network model after being trained；
Wherein, the BP neural network model is trained using following methods：
Step 2.3.3.1, input layer include three neurons, and each input layer is respectively to be in generating state wind Blower fan wind speed, blower fan wind direction and the atmospheric density of machine；
Hidden layer neuron h is calculated using below equationjOutput valve：
Output layer neuron o is calculated using below equation againkOutput valve：
Step 2.3.3.2, it is as follows to define loss function：
Wherein：ykFor the desired output of output layer neuron, initial value is history corresponding to each training sample data Actual active power value；E is deviation；
Make ek=yk-ok, ekFor deviation corresponding to k-th of output layer neuron；
Then E can be expressed as：
The output layer neuron o that step 2.3.3.2 is calculatedkOutput valve substitute into loss function, be calculated partially Difference E；
Whether step 2.3.3.3, judgment bias value E meet to require, if satisfaction requirement, goes to step 2.3.3.10；If no Meet to require, go to step 2.3.3.4；
Step 2.3.3.4, use following formula calculate hidden layer to output layer weight adjustment amount for：
Δwjk(q+1)=(1- γ) hjek+γΔwjk(q)
Wherein：
Wherein：γ is weights inertia coeffeicent, eqAnd eq-1Respectively q and q-1 training error；Δwjk(q)For the q times instruction Hidden layer neuron h when practicingjTo output layer neuron okWeight adjustment amount；Δwjk(q+1)Hidden layer god when being trained for the q+1 times Through first hjTo output layer neuron okWeight adjustment amount；
Step 2.3.3.5, then use following formula calculate input layer to hidden layer weight adjustment amount for：
Wherein：γ is weights inertia coeffeicent；Δwij(q)Input layer x when being trained for the q timesiTo hidden layer neuron hjWeight adjustment amount；Δwij(q+1)Input layer x when being trained for the q+1 timesiTo hidden layer neuron hjWeight adjustment Amount；
Step 2.3.3.6, biasing number b is calculated using following formulakUpdated value：
bk=bk+μek
Step 2.3.3.7, biasing number a is calculated using following formulakUpdated value：
Step 2.3.3.8, therefore, using the weight adjustment amount of the hidden layer that step 2.3.3.4 is calculated to output layer, The biasing number b that weight adjustment amount, the step 2.3.3.6 of the input layer that step 2.3.3.5 is calculated to hidden layer are calculatedk Updated value and the biasing number α that are calculated of step 2.3.3.7kUpdated value optimize and revise before the BP god that once trains to obtain Corresponding parameter through network model, the BP neural network model after thus being updated；
Step 2.3.3.9, the BP neural network model after the renewal obtained based on step 2.3.3.8, return to step 2.3.3.1
Step 2.3.3.10, the BP neural network model after being trained.
Wind power output abnormal data reconstructing method provided by the invention based on big data technology has advantages below：
Can accurately, efficiently, reduction reconstruct quickly is carried out to the exceptional data point in wind-powered electricity generation history data, so as to Be advantageous to the parameter such as the performance of accurate evaluation blower fan and wind power plant and the annual electricity generating capacity of operation conditions, prediction blower fan and wind power plant.
Brief description of the drawings
Fig. 1 is that the flow of the wind power output abnormal data reconstructing method provided by the invention based on big data technology is illustrated Figure；
Fig. 2 is that level graph of a relation is dispatched in the active data management of blower fan；
Fig. 3 is distributed storage graph of a relation；
Fig. 4 is to reconstruct Organization Chart based on the overall abnormal data for merging business.
Embodiment
In order that technical problem solved by the invention, technical scheme and beneficial effect are more clearly understood, below in conjunction with Drawings and Examples, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein only to The present invention is explained, is not intended to limit the present invention.
The characteristics of being handled after large-scale wind power access it is necessary to have mass data storage and high-performance data, these features The difficulty of design magnanimity power data related platform is added, while the calculating for the intelligent algorithm being related to scheme is brought Challenge.Therefore, present invention employs the framework related core algorithm of support and data weight of distributed storage to Distributed Calculation The processing business of structure flow.
Hadoop is the software platform of an exploitation and operation processing large-scale data, is one of Appach with java languages Speech realizes open source software framework, realizes and carries out Distributed Calculation to mass data in the cluster of a large amount of computers composition. Most crucial design is exactly in Hadoop frameworks：HDFS and MapReduce.HDFS provides the storage of mass data, MapReduce Provide the calculating to data.
The flow that data are handled in Hadoop can be simply interpreted as：Obtained after the cluster processing that data pass through Haddop As a result.By the method, development can only consider to realize Map and Reduce classes, various present in other multiple programmings Problem, such as distributed storage, scheduling, load balancing, fault-tolerant processing, network service, by the MapReduce in framework It is responsible for processing with HDFS file system, developer does not have to spend too many energy.
The present invention assesses the wind-powered electricity generation unit quality of data provided with " towards the Data quality assessment model of new energy power station " Index and data mode are analysis foundation, the wind calculated with " the wind-powered electricity generation unit based on neutral net abandons wind-powered electricity generation amount assessment models " Electric unit theoretical power (horse-power) value is training result, by distributed big data parallel computation deployment way (Hadoop), to number be present Performance number reconstruct is carried out according to the active data of exception.
A kind of wind power output abnormal data reconstructing method based on big data technology provided by the invention, with reference to figure 1, including Following steps：
Step 1, the distribution of substantial amounts of wind power output initial data is stored in HDFS different nodes；Wherein, the wind Electric output initial data includes data label, blower fan ID, power plant ID, region ID, data acquisition date, data acquisition moment, reality Border active power value, acquisition mode, collector, blower fan wind speed, blower fan wind direction, atmospheric density and fan condition；
Data storage method is described in detail as follows：
What the present invention mainly studied is the data reconstruction method of the daily 96 power data values of wind park, and historical data comes from In provincial control centre, provincial control centre's understructure is subdispatch, and subdispatch lower floor is wind park, each wind park Some independent blower fans are included again.Power data per Fans includes data label, blower fan ID, power plant ID, region ID, number According to date, data moment, generating contribution value, acquisition mode, collector, collection moment, wind speed, the relevant information such as state, by This is visible, and blower fan information data amount is huge, if it is slower to be only stored in data access rate in a relation table, and with when Between growth, wind-powered electricity generation data volume constantly increases, in order to solve the requirement higher to data access speed and intelligent training algorithm Performance requirement, according to the characteristics of wind power foundation data message, the file structure on HDFS is devised according to certain rule.
The active data management of blower fan dispatches hierarchical relationship as shown in Fig. 2 as shown in Figure 2, single wind park includes more typhoons Machine, more Fans belong to some subdispatch, the provincial scheduling that subdispatch ownership is specified.It can be obtained by analysis, the active number of blower fan According to that can be storage cell according to separate unit blower fan, can so fine-grained management be carried out to data, while can be according to business demand Computing resource is maximally utilized, avoids causing the waste of computing resource.
, first should be according to a wind according to the characteristics of blower fan data message in order to which data are deployed in Hadoop platform Power plant is a tables of data, and wind power information data are splitted into the form of several tables of data, and will be per number using Sqoop instruments File format is converted into according to table to upload in HDFS.Sqoop is the data dump instrument under Hadoop, being capable of implementation relation type number According to storehouse and the data conversion of Hadoop databases, and Sqoop is established on MapReduce in itself, is had to Hadoop fine Compatibility, there is higher dump efficiency.Simultaneously in order to consider the needs of power training algorithm, say that multiple files are scattered and deposit Be placed in HDFS multiple nodes, it is specific store pattern as shown in figure 3, using Fig. 3 storage mode, will be using wind park as unit Distributed data storage in HDFS different nodes, this storage form can largely improve data access rate, and The computational efficiency of the training algorithm program of theoretical power (horse-power) value can be improved.In addition, by wind-powered electricity generation data storage in HDFS, moreover it is possible to utilize Its file backup mechanism and data equilibrating mechanism that carry, the reliability and fault-tolerance of the active data of small wind-powered electricity generation can be significantly improved, So that wind power data error rate in storage and calculating is greatly reduced.
Step 2, several wind power output initial data for needing to carry out data reconstruction are read from HDFS；Will be described some Individual wind power output original data division is several data blocks, and each data block is distributed into a corresponding distribution process Task；
Each distribution process task uses parallel computation mode, and circular treatment is distributed to each in the data block of itself Individual wind power output initial data, abnormal data reconstruct is carried out to it；
Wherein, enter in the following ways for an arbitrary wind power output initial data, corresponding distribution process task Row abnormal data reconstructs：
Step 2.1, the distribution process task recognition is to fan condition；Wherein, the fan condition includes following nine kinds State：Treat wind state, generating state, drop forelock electricity condition, planned outage state, unplanned outage state, scheduling stoppage in transit state, Stoppage in transit state of being involved in communicating interrupt state, field and stoppage in transit state of being involved outside the venue；
Step 2.2, if it is communicating interrupt state to recognize fan condition, data are not reconstructed, wait pending data It is reconstructed again after given state；
If fan condition is recognized to treat wind state, planned outage state, unplanned outage state, scheduling stoppage in transit shape Stoppage in transit state of being involved in state, field or stoppage in transit state of being involved outside the venue, then be reconstructed into 0 value by the actual active power value of blower fan；
If recognizing fan condition as generating state or drop forelock electricity condition value, step 2.3 is performed；
Step 2.3, according to blower fan ID, the BP neural network model after training corresponding with blower fan is transferred, based on the instruction BP neural network model after white silk, blower fan theory active power value is calculated；
In this step, the BP neural network model after training corresponding with blower fan obtains by the following method：
Step 2.3.1, establish BP neural network model；The topological structure of the BP neural network model include input layer, Hidden layer and output layer；Wherein, the neuron number of input layer is n, and the neuron number of hidden layer is l, the nerve of output layer First number is m；Any input layer is xi, i ∈ (1,2 ... n)；Any hidden layer neuron is hj, j ∈ (1,2 ... l)；Appoint Output layer neuron of anticipating is ok, k ∈ (1,2 ... m)；
Step 2.3.2, the basic parameter of BP neural network model is initialized, including：Learning rate μ, input layer are to implicit The weight w of layerij, hidden layer to output layer weight wjk, input layer to hidden layer biasing number aj, hidden layer to the inclined of output layer Put several bkAnd excitation function t (x)；Wherein, weight w of the input layer to hidden layerij, hidden layer to output layer weight wjk, it is defeated Enter layer to the biasing number a of hidden layerj, hidden layer to output layer biasing number bkInitialization value is the random number in (- 1,1)；
Wherein：Weight w of the input layer to hidden layerijImplication is：Arbitrary input layer xiTo arbitrary hidden layer Neuron hjBetween weight；Weight w of the hidden layer to output layerjkImplication is：Arbitrary hidden layer neuron hjTo arbitrary Output layer neuron okBetween weight；Biasing number a of the input layer to hidden layerjImplication is：Each input layer is to arbitrarily Hidden layer neuron hjBiasing number；Biasing number b of the hidden layer to output layerkImplication is：Each hidden layer neuron is to arbitrary defeated Go out a layer neuron okBiasing number
Step 2.3.3, obtain training sample data；Wherein, the training sample data are separate unit blower fan history run shape The three-dimensional data of state, including：Blower fan wind speed, blower fan wind direction and atmospheric density；
Using the training sample data, based on the basic parameter of the initialization BP neural network model in step 2.3.2, The BP neural network model is trained, the BP neural network model after being trained；
Wherein, the BP neural network model is trained using following methods：
Step 2.3.3.1, input layer include three neurons, and each input layer is respectively to be in generating state wind Blower fan wind speed, blower fan wind direction and the atmospheric density of machine；
Hidden layer neuron h is calculated using below equationjOutput valve：
Output layer neuron o is calculated using below equation againkOutput valve：
Step 2.3.3.2, it is as follows to define loss function：
Wherein：ykFor the desired output of output layer neuron, initial value is history corresponding to each training sample data Actual active power value；E is deviation；
Make ek=yk-ok, ekFor deviation corresponding to k-th of output layer neuron；
Then E can be expressed as：
The output layer neuron o that step 2.3.3.2 is calculatedkOutput valve substitute into loss function, be calculated partially Difference E；
Whether step 2.3.3.3, judgment bias value E meet to require, if satisfaction requirement, goes to step 2.3.3.10；If no Meet to require, go to step 2.3.3.4；
Step 2.3.3.4, use following formula calculate hidden layer to output layer weight adjustment amount for：
Δwjk(q+1)=(1- γ) hjek+γΔwjk(q)
Wherein：
Wherein：γ is weights inertia coeffeicent, eqAnd eq-1Respectively q and q-1 training error；Δwjk(q)For the q times instruction Hidden layer neuron h when practicingjTo output layer neuron okWeight adjustment amount；Δwjk(q+1)Hidden layer god when being trained for the q+1 times Through first hjTo output layer neuron okWeight adjustment amount；
Step 2.3.3.5, then use following formula calculate input layer to hidden layer weight adjustment amount for：
Wherein：γ is weights inertia coeffeicent；Δwij(q)Input layer x when being trained for the q timesiTo hidden layer neuron hjWeight adjustment amount；Δwij(q+1)Input layer x when being trained for the q+1 timesiTo hidden layer neuron hjWeight adjustment Amount；
Step 2.3.3.6, biasing number b is calculated using following formulakUpdated value：
bk=bk+μek
Step 2.3.3.7, biasing number a is calculated using following formulakUpdated value：
Step 2.3.3.8, therefore, using the weight adjustment amount of the hidden layer that step 2.3.3.4 is calculated to output layer, The biasing number b that weight adjustment amount, the step 2.3.3.6 of the input layer that step 2.3.3.5 is calculated to hidden layer are calculatedk Updated value and the biasing number a that are calculated of step 2.3.3.7kUpdated value optimize and revise before the BP god that once trains to obtain Corresponding parameter through network model, the BP neural network model after thus being updated；
Step 2.3.3.9, the BP neural network model after the renewal obtained based on step 2.3.3.8, return to step 2.3.3.1
Step 2.3.3.10, the BP neural network model after being trained.
Step 2.4, if it is generating state to recognize fan condition, the actual active power value of blower fan is reconstructed into wind Mechanism opinion active power value；
If recognizing fan condition as drop forelock electricity condition value, power of fan dispatch value is further obtained；Then, sentence Whether disconnected blower fan theory active power value is more than or equal to power of fan dispatch value, if it is, the actual active power by blower fan Value is reconstructed into power of fan dispatch value；If not, the actual active power value of blower fan is reconstructed into blower fan theory active power value.
By above-mentioned flow processing, the abnormal conditions such as missing, out-of-limit, doomed dead, base present in active data can be reconstructed In high confidence level and the Data quality assessment model of high correlation analysis, it can avoid what is brought because lacking multifactor verification missing Judge and misjudge active data problem by accident, the basic data of high-quality is provided for further data processing and inversion.
By the way of step 2, restructural goes out the active power value of separate unit blower fan.In practical application, (closed by Reduce And business), with reference to business demand, the processing of station level calculation result can be carried out with merging, herein module integrated wind-powered electricity generation place There are the theoretical model data of blower fan, calculate the theoretical power (horse-power) under reconstruct wind speed corresponds to.Similarly, if desired region and province adjust number of levels According to merging treatment, the merging that Reduce carries out all areas result can be continued through, can be by Reduce to meter at all levels Result is calculated to be combined.When need not merge upwards step by step, the training result progress that can also directly invoke separate unit blower fan is defeated Go out.Its implementation process is as shown in Figure 4.
More abnormal data in a large amount of blower foundation data being collected into from wind power plant generally be present, such as lack, be dead The problems such as several, out-of-limit, such abnormity point are assessed electric network influencing wind power wave characteristic, wind power prediction, wind power And control strategy, the calculating of wind power curve, test and amendment etc. research are negatively affected.Using provided by the invention Wind power output abnormal data reconstructing method based on big data technology, it is preferential to carry out fan condition examination, can be targetedly Selection needs the blower fan for calculating theoretical power to carry out mass data calculating, avoids and repeats invalid computation work；Big data The introducing of technology, the defects of can overcoming intelligent algorithm execution cycle is longer when training magnanimity big data, it can be distributed Formula multithreading is calculated, and improves the reliability of algorithm operation efficiency and operation result.
The present invention can complete the reconstruct work of magnanimity blower foundation Information abnormity data from Practical angle efficiently and accurately Make, reconstruct confidence level is higher, improves the integrality of data, is advantageous to the recycling of data.
Described above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art For member, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications also should Depending on protection scope of the present invention.

Claims (2)

1. a kind of wind power output abnormal data reconstructing method based on big data technology, it is characterised in that comprise the following steps：
Step 1, the distribution of substantial amounts of wind power output initial data is stored in HDFS different nodes；Wherein, the wind-powered electricity generation goes out Power initial data includes data label, blower fan ID, power plant ID, region ID, the data acquisition date, the data acquisition moment, actually had Work(performance number, acquisition mode, collector, blower fan wind speed, blower fan wind direction, atmospheric density and fan condition；
Step 2, several wind power output initial data for needing to carry out data reconstruction are read from HDFS；Will several described wind Electric output original data division is several data blocks, and each data block is distributed into a corresponding distribution process and appointed Business；
Each distribution process task uses parallel computation mode, and circular treatment distributes to each wind in the data block of itself Electric output initial data, abnormal data reconstruct is carried out to it；
Wherein, carry out in the following ways different for an arbitrary wind power output initial data, corresponding distribution process task Regular data reconstructs：
Step 2.1, the distribution process task recognition is to fan condition；Wherein, the fan condition includes following nine kinds of states： Treat wind state, generating state, drop forelock electricity condition, planned outage state, unplanned outage state, scheduling stoppage in transit state, communication Stoppage in transit state of being involved in interrupt status, field and stoppage in transit state of being involved outside the venue；
Step 2.2, if it is communicating interrupt state to recognize fan condition, data are not reconstructed, wait pending data to give It is reconstructed again after state；
If fan condition is recognized to treat wind state, planned outage state, unplanned outage state, scheduling stoppage in transit state, field Interior involvement stoppage in transit state or stoppage in transit state of being involved outside the venue, then be reconstructed into 0 value by the actual active power value of blower fan；
If recognizing fan condition as generating state or drop forelock electricity condition value, step 2.3 is performed；
Step 2.3, according to blower fan ID, the BP neural network model after training corresponding with blower fan is transferred, after the training BP neural network model, blower fan theory active power value is calculated；
Step 2.4, if it is generating state to recognize fan condition, the actual active power value of blower fan is reconstructed into blower fan reason By active power value；
If recognizing fan condition as drop forelock electricity condition value, power of fan dispatch value is further obtained；Then, wind is judged Whether mechanism opinion active power value is more than or equal to power of fan dispatch value, if it is, the actual active power value weight by blower fan Structure is power of fan dispatch value；If not, the actual active power value of blower fan is reconstructed into blower fan theory active power value.
2. the wind power output abnormal data reconstructing method according to claim 1 based on big data technology, it is characterised in that In step 2.3, the BP neural network model after training corresponding with blower fan obtains by the following method：
Step 2.3.1, establish BP neural network model；The topological structure of the BP neural network model includes input layer, implied Layer and output layer；Wherein, the neuron number of input layer is n, and the neuron number of hidden layer is l, and the neuron of output layer is individual Number is m；Any input layer is xi, i ∈ (1,2 ... n)；Any hidden layer neuron is hj, j ∈ (1,2 ... l)；It is any defeated It is o to go out layer neuronk, k ∈ (1,2 ... m)；
Step 2.3.2, the basic parameter of BP neural network model is initialized, including：Learning rate μ, input layer to hidden layer Weight wij, hidden layer to output layer weight wjk, input layer to hidden layer biasing number aj, hidden layer to output layer biasing number bkAnd excitation function t (x)；Wherein, weight w of the input layer to hidden layerij, hidden layer to output layer weight wjk, input layer To the biasing number a of hidden layerj, hidden layer to output layer biasing number bkInitialization value is the random number in (- 1,1)；
Wherein：Weight w of the input layer to hidden layerijImplication is：Arbitrary input layer xiTo arbitrary hidden layer nerve First hjBetween weight；Weight w of the hidden layer to output layerjkImplication is：Arbitrary hidden layer neuron hjTo arbitrary output Layer neuron okBetween weight；Biasing number a of the input layer to hidden layerjImplication is：Each input layer implies to arbitrary Layer neuron hjBiasing number；Biasing number b of the hidden layer to output layerkImplication is：Each hidden layer neuron is to arbitrary output layer Neuron okBiasing number
Step 2.3.3, obtain training sample data；Wherein, the training sample data are separate unit blower fan history run state Three-dimensional data, including：Blower fan wind speed, blower fan wind direction and atmospheric density；
Using the training sample data, based on the basic parameter of the initialization BP neural network model in step 2.3.2, to institute State BP neural network model to be trained, the BP neural network model after being trained；
Wherein, the BP neural network model is trained using following methods：
Step 2.3.3.1, input layer include three neurons, and each input layer is respectively in generating state blower fan Blower fan wind speed, blower fan wind direction and atmospheric density；
Hidden layer neuron h is calculated using below equationjOutput valve：
<mrow> <msub> <mi>h</mi> <mi>j</mi> </msub> <mo>=</mo> <mi>t</mi> <mrow> <mo>(</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>n</mi> </munderover> <msub> <mi>w</mi> <mrow> <mi>i</mi> <mi>j</mi> </mrow> </msub> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>+</mo> <msub> <mi>a</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> </mrow>
Output layer neuron o is calculated using below equation againkOutput valve：
<mrow> <msub> <mi>o</mi> <mi>k</mi> </msub> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>j</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>l</mi> </munderover> <msub> <mi>h</mi> <mi>j</mi> </msub> <msub> <mi>w</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <mo>+</mo> <msub> <mi>b</mi> <mi>k</mi> </msub> </mrow>
Step 2.3.3.2, it is as follows to define loss function：
<mrow> <mi>E</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </munderover> <msup> <mrow> <mo>(</mo> <msub> <mi>y</mi> <mi>k</mi> </msub> <mo>-</mo> <msub> <mi>o</mi> <mi>k</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow>
Wherein：ykFor the desired output of output layer neuron, initial value, which is that history is actual corresponding to each training sample data, to be had Work(performance number；E is deviation；
Make ek=yk-ok, ekFor deviation corresponding to k-th of output layer neuron；
Then E can be expressed as：
<mrow> <mi>E</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mn>2</mn> </mfrac> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </munderover> <msup> <msub> <mi>e</mi> <mi>k</mi> </msub> <mn>2</mn> </msup> </mrow>
The output layer neuron o that step 2.3.3.2 is calculatedkOutput valve substitute into loss function, deviation E is calculated；
Whether step 2.3.3.3, judgment bias value E meet to require, if satisfaction requirement, goes to step 2.3.3.10；If it is unsatisfactory for It is required that go to step 2.3.3.4；
Step 2.3.3.4, use following formula calculate hidden layer to output layer weight adjustment amount for：
Δwjk(q+1)=(1- γ) hjek+γΔwjk(q)
Wherein：
<mrow> <mi>&amp;gamma;</mi> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mn>1</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mfrac> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mi>q</mi> </msub> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>q</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mfrac> <mo>&lt;</mo> <mn>1</mn> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>1</mn> <mo>-</mo> <mfrac> <mrow> <mn>5</mn> <mo>|</mo> <msub> <mi>e</mi> <mi>q</mi> </msub> <mo>|</mo> </mrow> <mrow> <mn>6</mn> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>q</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mfrac> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mn>1</mn> <mo>&amp;le;</mo> <mfrac> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mi>q</mi> </msub> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>q</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mfrac> <mo>&amp;le;</mo> <mn>2</mn> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo>,</mo> </mrow> </mtd> <mtd> <mrow> <mn>1.2</mn> <mo>&lt;</mo> <mfrac> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mi>q</mi> </msub> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <msub> <mi>e</mi> <mrow> <mi>q</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> <mo>|</mo> </mrow> </mfrac> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>
Wherein：γ is weights inertia coeffeicent, eqAnd eq-1Respectively q and q-1 training error；Δwjk(q)When being trained for the q times Hidden layer neuron hjTo output layer neuron okWeight adjustment amount；Δwjk(q+1)Hidden layer neuron when being trained for the q+1 times hjTo output layer neuron okWeight adjustment amount；
Step 2.3.3.5, then use following formula calculate input layer to hidden layer weight adjustment amount for：
<mrow> <msub> <mi>&amp;Delta;w</mi> <mrow> <mi>i</mi> <mi>j</mi> <mrow> <mo>(</mo> <mi>q</mi> <mo>+</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> </msub> <mo>=</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>&amp;gamma;</mi> <mo>)</mo> </mrow> <msub> <mi>&amp;mu;h</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <msub> <mi>h</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <msub> <mi>x</mi> <mi>i</mi> </msub> <msubsup> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </msubsup> <msub> <mi>w</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <msub> <mi>e</mi> <mi>k</mi> </msub> <mo>+</mo> <msub> <mi>&amp;gamma;&amp;Delta;w</mi> <mrow> <mi>i</mi> <mi>j</mi> <mrow> <mo>(</mo> <mi>q</mi> <mo>)</mo> </mrow> </mrow> </msub> </mrow>
Wherein：γ is weights inertia coeffeicent；Δwij(q)Input layer x when being trained for the q timesiTo hidden layer neuron hj's Weight adjustment amount；Δwij(q+1)Input layer x when being trained for the q+1 timesiTo hidden layer neuron hjWeight adjustment amount；
Step 2.3.3.6, biasing number b is calculated using following formulakUpdated value：
bk=bk+μek
Step 2.3.3.7, biasing number a is calculated using following formulakUpdated value：
<mrow> <msub> <mi>a</mi> <mi>k</mi> </msub> <mo>=</mo> <msub> <mi>a</mi> <mi>k</mi> </msub> <mo>+</mo> <msub> <mi>&amp;mu;h</mi> <mi>j</mi> </msub> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <msub> <mi>h</mi> <mi>j</mi> </msub> <mo>)</mo> </mrow> <msubsup> <mo>&amp;Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>m</mi> </msubsup> <msub> <mi>w</mi> <mrow> <mi>j</mi> <mi>k</mi> </mrow> </msub> <msub> <mi>e</mi> <mi>k</mi> </msub> </mrow>
Step 2.3.3.8, therefore, using the weight adjustment amount of the hidden layer that step 2.3.3.4 is calculated to output layer, step 2.3.3.5 the biasing number b that weight adjustment amount, the step 2.3.3.6 of the input layer being calculated to hidden layer are calculatedkMore The biasing number a that new value and step 2.3.3.7 are calculatedkUpdated value optimize and revise before once train obtained BP nerve nets The corresponding parameter of network model, the BP neural network model after thus being updated；
Step 2.3.3.9, the BP neural network model after the renewal obtained based on step 2.3.3.8, return to step 2.3.3.1,
Step 2.3.3.10, the BP neural network model after being trained.
CN201710612234.5A 2017-07-25 2017-07-25 Wind power output abnormal data reconstructing method based on big data technology Pending CN107391689A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710612234.5A CN107391689A (en) 2017-07-25 2017-07-25 Wind power output abnormal data reconstructing method based on big data technology

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710612234.5A CN107391689A (en) 2017-07-25 2017-07-25 Wind power output abnormal data reconstructing method based on big data technology

Publications (1)

Publication Number Publication Date
CN107391689A true CN107391689A (en) 2017-11-24

Family

ID=60337629

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710612234.5A Pending CN107391689A (en) 2017-07-25 2017-07-25 Wind power output abnormal data reconstructing method based on big data technology

Country Status (1)

Country Link
CN (1) CN107391689A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108335228A (en) * 2017-12-26 2018-07-27 南京海兴电网技术有限公司 A kind of workload equilibrium worksheet processing method in the power distribution network based on improved BP
CN109695546A (en) * 2018-12-29 2019-04-30 中国大唐集团新能源科学技术研究院有限公司 A kind of universal fan monitoring method

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103001249A (en) * 2012-11-15 2013-03-27 中国电力科学研究院 Method for predicting short-term power of wind farm on the basis of BP (back propagation) neural network
CN105736253A (en) * 2016-02-04 2016-07-06 华电电力科学研究院 Method for judging state of wind turbine generator on basis of air speed and power and calculating energy utilization ratio
CN106779377A (en) * 2016-12-05 2017-05-31 王忠平 Wind electricity digestion process evaluation methods

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103001249A (en) * 2012-11-15 2013-03-27 中国电力科学研究院 Method for predicting short-term power of wind farm on the basis of BP (back propagation) neural network
CN105736253A (en) * 2016-02-04 2016-07-06 华电电力科学研究院 Method for judging state of wind turbine generator on basis of air speed and power and calculating energy utilization ratio
CN106779377A (en) * 2016-12-05 2017-05-31 王忠平 Wind electricity digestion process evaluation methods

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
刘成等: "基于多任务学习的自然图像分类研究", 《计算机应用研究》 *
张永宏等: "BP神经网络在图像字符识别中的改进和应用", 《南京信息工程大学学报自然科学版》 *
韩自奋: "风电风功率预测研究", 《中国博士学位论文全文数据库工程科技Ⅱ辑》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108335228A (en) * 2017-12-26 2018-07-27 南京海兴电网技术有限公司 A kind of workload equilibrium worksheet processing method in the power distribution network based on improved BP
CN108335228B (en) * 2017-12-26 2022-04-08 南京海兴电网技术有限公司 Workload balance order dispatching method in power distribution network based on improved BP neural network
CN109695546A (en) * 2018-12-29 2019-04-30 中国大唐集团新能源科学技术研究院有限公司 A kind of universal fan monitoring method

Similar Documents

Publication Publication Date Title
Boza et al. Artificial intelligence to support the integration of variable renewable energy sources to the power system
CN102479339B (en) Method and system for forecasting short-term wind speed of wind farm based on hybrid neural network
Andryushkevich et al. Composition and application of power system digital twins based on ontological modeling
Zhou et al. Multi-energy net load forecasting for integrated local energy systems with heterogeneous prosumers
WO2021077646A1 (en) Comprehensive multi-precision index evaluation method and system for electric power prediction models
CN104376389B (en) Master-slave mode microgrid power load prediction system and method based on load balancing
CN107947165A (en) A kind of power distribution network flexibility evaluation method towards regulatory demand
CN109242143A (en) A kind of neural network wind power forecasting method and system
CN106649085A (en) Cloud computing-based software test system
CN102509027B (en) Wind powder combined predication method based on cross entropy theory
CN107123982A (en) A kind of distribution network reliability analysis method of economic benefit based on equipment alteration
CN103699944A (en) Wind and photovoltaic generation power prediction system with multiple prediction modes
CN105046374A (en) Power interval predication method based on nucleus limit learning machine model
CN108027908A (en) Analogy method and system
CN103399496A (en) Massive real-time data load simulation testing cloud platform for smart power grid, and testing method of cloud platform
CN104992248A (en) Microgrid photovoltaic power station generating capacity combined forecasting method
CN105790266A (en) Microgrid parallel multi-target robust optimization scheduling integrated control method
CN108596242A (en) Power grid meteorology load forecasting method based on wavelet neural network and support vector machines
CN104102951A (en) Short-term wind power prediction method based on EMD historical data preprocessing
CN112633632A (en) Integrated short-term wind power cluster power prediction method based on signal decomposition technology
CN107391689A (en) Wind power output abnormal data reconstructing method based on big data technology
CN105678415A (en) Method for predicting net load of distributed power supply power distribution network
Aslan Archimedes optimization algorithm based approaches for solving energy demand estimation problem: a case study of Turkey
Alam et al. AI-based efficiency analysis technique for photovoltaic renewable energy system
Roberts Structured innovation approach for application to the wave energy sector

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: 20171124

RJ01 Rejection of invention patent application after publication