CN104376389B - Master-slave mode microgrid power load prediction system and method based on load balancing - Google Patents
Master-slave mode microgrid power load prediction system and method based on load balancing Download PDFInfo
- Publication number
- CN104376389B CN104376389B CN201410759321.XA CN201410759321A CN104376389B CN 104376389 B CN104376389 B CN 104376389B CN 201410759321 A CN201410759321 A CN 201410759321A CN 104376389 B CN104376389 B CN 104376389B
- Authority
- CN
- China
- Prior art keywords
- module
- mrow
- load
- data
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims description 15
- 238000013500 data storage Methods 0.000 claims abstract description 4
- 238000004422 calculation algorithm Methods 0.000 claims description 73
- 238000012549 training Methods 0.000 claims description 69
- 238000013528 artificial neural network Methods 0.000 claims description 45
- 238000004891 communication Methods 0.000 claims description 27
- 210000002569 neuron Anatomy 0.000 claims description 22
- 230000007935 neutral effect Effects 0.000 claims description 21
- 238000012545 processing Methods 0.000 claims description 11
- 230000005611 electricity Effects 0.000 claims description 10
- 238000004393 prognosis Methods 0.000 claims description 7
- 238000010586 diagram Methods 0.000 claims description 6
- 230000001537 neural effect Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 238000013468 resource allocation Methods 0.000 claims description 4
- 238000004088 simulation Methods 0.000 claims description 4
- 238000013277 forecasting method Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 2
- 238000000151 deposition Methods 0.000 claims 1
- 230000006870 function Effects 0.000 description 3
- 210000004218 nerve net Anatomy 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000002123 temporal effect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001172 regenerating effect 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
- 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
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
A kind of master-slave mode microgrid power load prediction system based on load balancing of the present invention, master server including the mathematical computations for high complexity and large-scale data storage, it can carry out information exchange by Ethernet switch and the distributed substation of forecasting system.Forecasting system distribution substation can need some the task of extensive computation to be sent to master server, be completed to calculate by master server.Blower fan, the generated output real time data of photovoltaic in collection micro-capacitance sensor or sub- microgrid are responsible in the distributed substation of each forecasting system, and the region load data.The present invention can realize the detailed predicting of microgrid power and load, and it is that EMS and micro-capacitance sensor controller provide accurate data and supported to predict the outcome, while reducing forecast cost, improves the service efficiency of forecasting system server and device.The load balancing of data level can be realized in system, each network element internal realizes network element internal thread-level load balancing.
Description
Technical field
The present invention relates to microgrid power prediction, micro-grid load prediction field, and in particular to be a kind of based on load
Master-slave mode microgrid power load prediction system and method in a balanced way.
Background technology
Micro-capacitance sensor is a kind of new electric network composition, includes wind-powered electricity generation, photovoltaic, diesel generation, energy storage, load, control guarantor
Protect module etc..Micro-capacitance sensor can realize the autonomous system of self-contr ol, protection and management, both can be with the grid-connected fortune of external electrical network
OK, can also isolated operation.Micro-capacitance sensor can sufficiently promote the efficient utilization of distributed power source and regenerative resource.
Photovoltaic, wind-powered electricity generation industry are developed rapidly in recent years, and increasing low profile photovoltaic is generated electricity and wind-powered electricity generation is applied in micro-capacitance sensor
Field, accurate wind-powered electricity generation and photovoltaic power prediction contributes to micro-capacitance sensor scheduling controlling and safe operation.If can determine future
Load data in certain particular moment micro-capacitance sensor, also has important meaning to micro-capacitance sensor economic load dispatching and energy management.
Photovoltaic in the market, wind-powered electricity generation, load prediction system need multiple acquisition channels, each forecasting system and phase
It is mutually independent, therefore cause data redundancy.In addition, the forecasting system of each family needs exploitation multiple communication interface to be used to predict at present
System and outside collector, controller, EMS, the communication dispatched, once Real-Time Communication Interface is interrupted, forecasting system
Input data will produce larger error, also can be inaccurate to predicting the outcome.
Photovoltaic generation and wind-power electricity generation are exerted oneself often unstable by inside even from weather.Traditional photovoltaic and wind
Electric forecasting system needs the support of meteorological department's numerical weather forecast, but current numerical weather forecast temporal resolution and sky
Between resolution ratio be typically all unable to reach the requirement accurately predicted, therefore also increase microgrid energy management system control, regulation
The difficulty exerted oneself with balancing the load.
Current power prediction and load prediction use neural network algorithm, system during neural network algorithm training mostly
Generally requiring the substantial amounts of resource of cost is used to calculate neural network weight, and this is accomplished by using high performance server.In reality
In operation, the training of neutral net is usually a month even longer time just to need training once, high cost, high performance clothes
Business device does not obtain higher utilization rate.
The content of the invention
In view of the shortcomings of the prior art, the present invention seeks to be to provide a kind of master-slave mode based on load balancing
Microgrid power load prediction system and method, provide accurate power and load prediction data for micro-capacitance sensor, realize system
The load balancing of level and network element internal thread-level, improves system and the whole utilization efficiency of device.
To achieve these goals, the present invention is to realize by the following technical solutions:
A kind of master-slave mode microgrid power load prediction system based on load balancing, it includes master server, for counting
Learn and calculate and data storage;
Work(in forecasting system distribution substation, the real time data related for gathering local prediction, prediction local zone
Rate and load data, while carrying out modified weight calculating;The master server too network switch and the distributed substation of forecasting system
Carry out information exchange,
The master server is included with lower module:
Meteorological acquisition module, the free numerical weather forecast of automatic network is carried out for gathering;
Accurate weather prognosis module, is foundation according to local longitude, dimension, height above sea level, geopotential unit, while according to micro-capacitance sensor
Geography information and building schematic diagram in garden, carry out three-dimensional modeling;
Database, for storing power, load data in meteorological historical data and gathered data, and micro-capacitance sensor garden;
Remote master server communication module is used to communicate with the distributed substation of each forecasting system, while gathering the meteorology on Internet
Data;
Communication module, based on common ethernet communication mode, uses ICP/IP protocol.
The long-range training module of neutral net, for training power prediction and load forecasting model;
Equalization algorithm module, is responsible for the asset creation and resource allocation of control neural network training module;
Remotely predicting module, predicts that multiple micro-capacitance sensors or many height are micro- according to predicting the outcome for the distributed substation of forecasting system
The general power and total load data of power network, or when being broken down in the distributed substation of some forecasting system, work(is completed instead of it
Rate and load prediction.
Further, the distributed substation of the forecasting system includes:
Central processing unit, for dispatching the collaborative work between each submodule, completes basic system operation and data
Processing;
Forecasting system communication module, for the distributed substation of forecasting system and the communication before master server, is gathered simultaneously
Local environment Monitoring Data;
Memory module, the historical data for storing nearly trimestral environment monitor, load, power;
Neural network algorithm training module, the online retraining with load forecasting model is predicted for training power;
Neural network algorithm train prediction module, according to neural network algorithm training module train come forecast model enter
Row weight estimation;
Equalization algorithm module, is responsible for the asset creation and resource allocation of control neural network training module, works as local resource
When can not meet training requirement or can influence data acquisition, power load prediction, central processing unit can be asked to master server
Remote opening is trained, and master server completes to send load forecasting model file to local after training;
Local prediction module, for gathering the power in the locally related real time data of prediction, prediction local zone and bearing
Lotus data;
The communication module, memory module, neural network algorithm training module and local prediction module are handled with center
Device is connected, and the equalization algorithm module and neural network algorithm training prediction module connect with neural network algorithm training module
Connect, the local prediction module, neural network algorithm training module and forecasting system communication module are connected with memory module,
It is easy to the storage of data.
Further, the local prediction module includes photovoltaic prediction module, wind-powered electricity generation prediction module and load prediction module.
Preferably, the neural network algorithm of the long-range training module of neutral net is calculated using improved BP neural network
Method, the improved BP neural network algorithm is as follows:
Right value update Δ Wlj:
Wherein, WljRepresent l-th of neuron to the connection weight between j-th of neuron of output layer, OijRepresent nerve net
Network is exported, yjRepresent desired value, vlRepresent l-th of neuron input value, θjThreshold value is represented, L represents neuron number.
Preferably, the database uses Oracle or sybase database.
A kind of master-slave mode microgrid power load forecasting method based on load balancing, its method and step is as follows:
(1) distributed capture prediction steps;The related real time data of the local prediction of collection, passes through Neural Network Prediction
Power and load data in local zone, while carrying out modified weight calculating, export budget result;
(2) meteorological acquisition step, the free numerical weather forecast of automatic network is carried out for gathering;
(3) weather prognosis step, the local small range longitude used according to step (1), dimension, height above sea level, geopotential unit for
Foundation, carries out three-dimensional modeling;The network weather data that model is obtained using step (2) as input condition, using physical equation and
The meteorological change of thermodynamical equilibrium equation simulation, and after neural network algorithm amendment, finally give accurate meteorological letter in micro-capacitance sensor garden
Breath;
(4) general power and total load data of micro-capacitance sensor are predicted;According to predicting the outcome and predicted multiple micro- for step (1)
The general power and total load data of power network or many sub- micro-capacitance sensors;
Every prediction is all based on neural computing, and neural computing is to be based on BP neural network algorithm, tradition
Neutral net be that neural net model establishing and computing performed by serialized manner on certain server or computer, the present invention is
Utilize " load balance scheduling algorithm ", by single serial neural net model establishing and computing on certain computer or server it is parallel
Perform or partial arithmetic task submits to " remote master server " execution, the parallel method on single personal computer or server is joined
Examine Fig. 5.
The accurate weather prognosis module of remote master server can be according to local small range longitude, dimension, height above sea level, geopotential unit
For foundation, with reference to the geography information in micro-capacitance sensor garden and building schematic diagram, three-dimensional modeling is carried out.Model is with network weather
Data are changed, after neutral net amendment, finally as input condition using physical equation and thermodynamical equilibrium equation simulation meteorology
Accurate weather information in the range of every 30 × 30 square metres is obtained in micro-capacitance sensor garden, temporal resolution is less than 5 minutes.
Remote master server database uses the Oracle or sybase database industrially generally used.For storing gas
As the power in historical data and gathered data, and micro-capacitance sensor garden, load data.
Remote master server communication module is used to communicate with the distributed substation of each forecasting system, while gathering Internet
On meteorological data.Communication module uses ICP/IP protocol based on common ethernet communication mode.
The long-range training module of remote master server neutral net, which is used for training power, to be predicted and load forecasting model.It is wherein refreshing
Improved BP neural network algorithm is used through network algorithm.Wherein right value update, the calculation formula that we use is as follows:
Wherein WljRepresent l-th of neuron to the connection weight between j-th of neuron of output layer, OijRepresent nerve net
Network is exported, yjRepresent desired value, vlRepresent l-th of neuron input value, θjThreshold value is represented, L represents neuron number.
Remote master server remotely predicting module can be multiple according to the prediction that predicts the outcome of the distributed substation of forecasting system
The general power and total load data of micro-capacitance sensor or many sub- micro-capacitance sensors, or broken down in the distributed substation of some forecasting system
When, the function of power and load prediction is completed instead of it.
The method of neural network algorithm is:It is initialization weight data, each weights assignment random number, random number model first
It is trapped among between -1 to 1;Using historical data sample as the input value of neutral net, the input of each layer of neutral net is calculated respectively
And output.After the final output for obtaining neutral net, theoretical value and calculated value mean square error are asked, if error meets default bar
Part, then training terminates;If being unsatisfactory for preparatory condition, then maximum cycle is judged, if same beyond maximum cycle
Sample also terminates training, otherwise by every layer of neuron partial gradient of backwards calculation and corrects neuron weights;Until meeting mean square error
The preparatory condition of difference or maximum cycle.
Equalization algorithm module first can inspection operation system version, and load it is corresponding it is grand, initialization global variable.Algorithm
How many CPU can be counted and start to monitor each CPU occupancy.Balance module algorithm can create a task in internal memory
Manager is used for task burst, task monitors, distributing system resource and recovery system resource.If local resource can meet institute
When having task, task manager can start to perform the task after burst.If local resource is it is impossible to meet all tasks, and
The task can not be waited, and equalization algorithm module can connect master server, and the task is submitted into master server performs.
Each subtask first can initialization task counter, it is ensured that the task occupying system resources time be it is limited,
The task of data loading, data processing will be completed after the tasks carrying, and to outside I/O Request reading and writing of files, file lock will be protected
The uniqueness of file read-write is demonstrate,proved, corrupt data is prevented.
The present invention can realize the detailed predicting of microgrid power and load, predict the outcome as EMS and micro- electricity
Net controller is supported there is provided accurate data, while reducing forecast cost, improves the use of forecasting system server and device
Efficiency.The load balancing of data level can be realized in system, each network element internal realizes network element internal thread-level load balancing.
Brief description of the drawings
Describe the present invention in detail with reference to the accompanying drawings and detailed description;
The system structure diagram that it is the present invention that Fig. 1, which is,;
Fig. 2 is neural network BP training algorithm schematic flow sheet;
Fig. 3 is the distributed substation structured flowchart of forecasting system of the present invention;
Fig. 4 is present system level load balance scheduling algorithm flow chart;
Fig. 5 is network element internal thread-level load balance scheduling algorithm flow chart of the present invention.
Embodiment
To be easy to understand the technical means, the inventive features, the objects and the advantages of the present invention, with reference to
Embodiment, is expanded on further the present invention.
A kind of master-slave mode microgrid power and load forecasting method based on load balancing include:It is remote master server, pre-
Examining system distribution substation.
Remote master server includes:Meteorological acquisition module, accurate weather prognosis module, database, communication module, nerve
Network remote training module, remotely predicting module, equalization algorithm module etc..
Forecasting system distribution substation is made up of following sections:Core processor, communication module, memory module, nerve
Network algorithm training module, local prediction module, equalization algorithm module.
Remote master server meteorology acquisition module is used to gather the free numerical weather forecast for carrying out automatic network, and this weather is pre-
The resolution ratio that calls time is typically at one more than hour, and spatial dimension is about more than 6 × 6 square kilometres of region.
The accurate weather prognosis module of remote master server can be according to local small range longitude, dimension, height above sea level, geopotential unit
For foundation, with reference to the geography information in micro-capacitance sensor garden and building schematic diagram, three-dimensional modeling is carried out.Model is with network weather
Data are changed, after neutral net amendment, finally as input condition using physical equation and thermodynamical equilibrium equation simulation meteorology
Accurate weather information in the range of every 30 × 30 square metres is obtained in micro-capacitance sensor garden, temporal resolution is less than 5 minutes.
Remote master server database uses the Oracle or sybase database industrially generally used.For storing gas
As the power in historical data and gathered data, and micro-capacitance sensor garden, load data.
Remote master server communication module is used to communicate with the distributed substation of each forecasting system, while gathering Internet
On meteorological data.Communication module uses ICP/IP protocol based on common ethernet communication mode.
The long-range training module of remote master server neutral net, which is used for training power, to be predicted and load forecasting model.It is wherein refreshing
Improved BP neural network algorithm is used through network algorithm.Wherein right value update, the calculation formula that we use is as follows:
Wherein WljRepresent l-th of neuron to the connection weight between j-th of neuron of output layer, OijRepresent nerve net
Network is exported, yjRepresent desired value, vlRepresent l-th of neuron input value, θjThreshold value is represented, L represents neuron number.
Remote master server remotely predicting module can be multiple according to the prediction that predicts the outcome of the distributed substation of forecasting system
The general power and total load data of micro-capacitance sensor or many sub- micro-capacitance sensors, or broken down in the distributed substation of some forecasting system
When, the function of power and load prediction is completed instead of it.
Remote master server equalization algorithm module refers to Fig. 4
Forecasting system distribution substation core processor is used to dispatch the collaborative work between each submodule, completes basic
System operation and data processing.
Forecasting system distribution substation communication module is used for substation and the communication before master server, works as while can gather
Ground environmental monitoring data.
Forecasting system distribution substation memory module is used to storing nearly trimestral environment monitor, load, power and gone through
History data.
The algorithm and the neutral net of server that forecasting system distribution substation neural network algorithm training module is used are calculated
Method is consistent, but training data is less, is generally used for the online retraining of model.
Forecasting system distribution substation neural network algorithm training prediction module is instructed according to neural network algorithm training module
Practise the forecast model come and be weighted prediction.
Forecasting system distribution substation equalization algorithm module map 5.
Fig. 1 is the system structure diagram of the present invention, and master server is used for the mathematical computations of high complexity and large-scale
Data storage, it can carry out information exchange by Ethernet switch and the distributed substation of forecasting system.Forecasting system is distributed
Formula station can need some the task of extensive computation to be sent to master server, be completed to calculate by master server.It is each pre-
It is responsible for blower fan, the generated output real time data of photovoltaic in collection micro-capacitance sensor or sub- microgrid, and the area in examining system distribution substation
The load data in domain.
Fig. 2 is neural network BP training algorithm flow.Neural network algorithm needs to do two pieces thing:1 training forecast model, 2 meters
Predict the outcome.This algorithm is the process for training forecast model.Training pattern equivalent to find function description input with
The relation of output, after this model is found, as long as input data as requested, it is possible to predicted the outcome, here it is
" process that calculating predicts the outcome ", Fig. 2 is the process for describing training pattern, and model training is not to predict the outcome to be required for instruction every time
Experienced, a model training, which has been got well, can just be placed on there, can use several days, some months or several years, if it find that pre- every time
Survey result and when too big actual result deviation, it is necessary to re -training model, find the relation of new input and output.God
Through network using three layers of feed-forward framework.It is initialization weight data, each weights assignment random number, random number range first
Between -1 to 1.Using historical data sample as neutral net input value, calculate respectively each layer of neutral net input and
Output.After the final output for obtaining neutral net, theoretical value and calculated value mean square error are asked, if error meets preparatory condition,
Then training terminates.If being unsatisfactory for preparatory condition, then maximum cycle is judged, if beyond maximum cycle equally
Terminate training, otherwise by every layer of neuron partial gradient of backwards calculation and correct neuron weights.Until meet mean square error or
The preparatory condition of maximum cycle.
Fig. 3 is the distributed substation structural representation of forecasting system, and dotted line represents the direction of device internal control stream, solid line table
The direction of showing device internal data flow.Equalization algorithm module is responsible for the asset creation and resource point of control neural network training module
Match somebody with somebody, when local resource can not meet training requirement or can influence data acquisition, power load prediction, device can be to main service
Device request remote opening training, master server completes to send model file to local after training.
Fig. 4 is system-level load balance scheduling algorithm flow chart, and whether local resource meets needs, opened if being unsatisfactory for
The algorithm is moved, circulation is intercepted substation request by master server after network connection service is opened, once there is substation to ask main service
Device completes neural metwork training, and master server carries out the historical data that corresponding substation information and substation are searched in database
Training.After training terminates, model file can be sent to corresponding substation.
Fig. 5 is network element internal thread-level load balance scheduling algorithm flow chart.Equalization algorithm module first can inspection operation
System version, and load corresponding grand, initialization global variable.How many CPU algorithms can count and start to monitor each CPU
Occupancy.Balance module algorithm can create a task manager in internal memory is used for task burst, task monitors, distribution system
Resource of uniting and recovery system resource.If local resource can meet all tasks, task manager can start to perform burst
Task afterwards.If local resource is it is impossible to meet all tasks, and the task can not be waited, and equalization algorithm module can connect
Master server is connect, the task is submitted into master server performs.
Each subtask first can initialization task counter, it is ensured that the task occupying system resources time be it is limited,
The task of data loading, data processing will be completed after the tasks carrying, and to outside I/O Request reading and writing of files, file lock will be protected
The uniqueness of file read-write is demonstrate,proved, corrupt data is prevented.
The general principle and principal character and advantages of the present invention of the present invention has been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the simply explanation described in above-described embodiment and specification is originally
The principle of invention, without departing from the spirit and scope of the present invention, various changes and modifications of the present invention are possible, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent thereof.
Claims (6)
1. a kind of master-slave mode microgrid power load prediction system based on load balancing, it includes master server, for mathematics
Calculate and data storage;
Power in forecasting system distribution substation, the real time data related for gathering local prediction, prediction local zone and
Load data, while carrying out modified weight calculating;The master server Ethernet switch enters with the distributed substation of forecasting system
Row information is interacted,
The master server is included with lower module:
Meteorological acquisition module, the free numerical weather forecast of automatic network is carried out for gathering;
Accurate weather prognosis module, is foundation according to local longitude, dimension, height above sea level, geopotential unit, while according to micro-capacitance sensor garden
Interior geography information and building schematic diagram, carry out three-dimensional modeling;
Database, for storing power, load data in meteorological historical data and gathered data, and micro-capacitance sensor garden;Remotely
Master server communication module is used to communicate with the distributed substation of each forecasting system, while gathering the meteorological number on Internet
According to;
Communication module, based on common ethernet communication mode, uses ICP/IP protocol;
The long-range training module of neutral net, for training power prediction and load forecasting model;
Equalization algorithm module, is responsible for the asset creation and resource allocation of control neural network training module;
Remotely predicting module, according to the predict the outcome multiple micro-capacitance sensors of prediction or many sub- micro-capacitance sensors of the distributed substation of forecasting system
General power and total load data, or when being broken down in the distributed substation of some forecasting system, instead of its complete power and
Load prediction;The neural network algorithm of the long-range training module of neutral net uses improved BP neural network algorithm, described
Improved BP neural network algorithm is as follows:
Right value update Δ Wlj:
<mrow>
<msub>
<mi>&Delta;W</mi>
<mrow>
<mi>l</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mo>-</mo>
<mn>0.02</mn>
<mrow>
<mo>(</mo>
<msub>
<mi>O</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>&times;</mo>
<mrow>
<mo>(</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>l</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>L</mi>
</munderover>
<msub>
<mi>W</mi>
<mrow>
<mi>l</mi>
<mi>j</mi>
</mrow>
</msub>
<msub>
<mi>v</mi>
<mi>l</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&theta;</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>&times;</mo>
<msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mrow>
<mo>(</mo>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>l</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>L</mi>
</msubsup>
<msub>
<mi>W</mi>
<mrow>
<mi>l</mi>
<mi>j</mi>
</mrow>
</msub>
<msub>
<mi>v</mi>
<mi>l</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&theta;</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</msup>
<mo>)</mo>
</mrow>
<mrow>
<mo>-</mo>
<mn>2</mn>
</mrow>
</msup>
<mo>&times;</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mrow>
<mo>(</mo>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>l</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>L</mi>
</msubsup>
<msub>
<mi>W</mi>
<mrow>
<mi>l</mi>
<mi>j</mi>
</mrow>
</msub>
<msub>
<mi>v</mi>
<mi>l</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&theta;</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
Wherein, WljRepresent l-th of neuron to the connection weight between j-th of neuron of output layer, OijRepresent that neutral net is defeated
Go out, yjRepresent desired value, vlRepresent l-th of neuron input value, θjThreshold value is represented, L represents neuron number.
2. the master-slave mode microgrid power load prediction system according to claim 1 based on load balancing, its feature exists
In the distributed substation of the forecasting system includes:
Central processing unit, for dispatching the collaborative work between each submodule, the basic system operation of completion and data processing;
Forecasting system communication module, for the distributed substation of forecasting system and the communication before master server, while gathering local
Environmental monitoring data;
Memory module, the historical data for storing nearly trimestral environment monitor, load, power;
Neural network algorithm training module, the online retraining with load forecasting model is predicted for training power;
Neural network algorithm train prediction module, according to neural network algorithm training module train come forecast model added
Power prediction;
Equalization algorithm module, is responsible for the asset creation and resource allocation of control neural network training module, when local resource can not
When meeting training requirement or can influence data acquisition, power load prediction, central processing unit can ask long-range to master server
Training is opened, master server completes to send load forecasting model file to local after training;
Power and load number in local prediction module, the real time data related for gathering local prediction, prediction local zone
According to;
The communication module, memory module, neural network algorithm training module and local prediction module connect with central processing unit
Connect, the equalization algorithm module and neural network algorithm training prediction module are connected with neural network algorithm training module, institute
State local prediction module, neural network algorithm training module and forecasting system communication module to be connected with memory module, be easy to
The storage of data.
3. the master-slave mode microgrid power load prediction system according to claim 2 based on load balancing, its feature exists
In the local prediction module includes photovoltaic prediction module, wind-powered electricity generation prediction module and load prediction module.
4. the master-slave mode microgrid power load prediction system according to claim 1 based on load balancing, its feature exists
In the database uses Oracle or sybase database.
5. a kind of master-slave mode microgrid power load forecasting method based on load balancing, its method and step is as follows:
(1) distributed capture prediction steps;The related real time data of the local prediction of collection, it is local by Neural Network Prediction
Power and load data in region, while carrying out modified weight calculating, export budget result;
(2) meteorological acquisition step, the free numerical weather forecast of automatic network is carried out for gathering;
(3) weather prognosis step, the local small range longitude used according to step (1), dimension, height above sea level, geopotential unit for foundation,
Carry out three-dimensional modeling;The network weather data that model is obtained using step (2) are as input condition, using physical equation and thermodynamics
Equation simulation meteorology change, and after neural network algorithm amendment, finally give accurate weather information in micro-capacitance sensor garden;
(4) general power and total load data of micro-capacitance sensor are predicted;According to the multiple micro-capacitance sensors or many of prediction that predict the outcome of step (1)
The general power and total load data of individual sub- micro-capacitance sensor;
The neural network algorithm includes system-level load balance scheduling algorithm steps, is wanted when local resource can not meet training
Ask or data acquisition can be influenceed, power load prediction when, device can to master server ask remote opening load balance scheduling
Circulation is intercepted substation request by Algorithm for Training, master server after network connection service is opened, once there is substation to ask main service
Device completes neural metwork training, and master server carries out the historical data that corresponding substation information and substation are searched in database
Training, master server completes to send model file to corresponding substation after training;
The method of neural network algorithm is:It is initialization weight data, each weights assignment random number, random number range first
Between -1 to 1;Using historical data sample as neutral net input value, calculate respectively each layer of neutral net input and
Output;After the final output for obtaining neutral net, theoretical value and calculated value mean square error are asked, if error meets preparatory condition,
Then training terminates;If being unsatisfactory for preparatory condition, then maximum cycle is judged, if beyond maximum cycle equally
Terminate training, otherwise by every layer of neuron partial gradient of backwards calculation and correct neuron weights;Until meet mean square error or
The preparatory condition of maximum cycle.
6. method according to claim 5, it is characterised in that also include network element internal thread in the neural network algorithm
Level load balance scheduling algorithm steps, equalization algorithm module first can inspection operation system version, and load corresponding grand, initial
Change global variable;How many CPU algorithms can count and start to monitor each CPU occupancy;Including the meeting of balance module algorithm
Depositing middle one task manager of establishment is used for task burst, task monitors, distributing system resource and recovery system resource;If this
When ground resource can meet all tasks, task manager can start to perform the task after burst;If local resource can not
All tasks are met, and the task can not be waited, and equalization algorithm module can connect master server, and the task is submitted into main clothes
Business device is performed;
Each subtask first can initialization task counter, it is ensured that the task occupying system resources time is limited, at this
The task of data loading, data processing will be completed after tasks carrying, and to outside I/O Request reading and writing of files, file lock will ensure text
The uniqueness of part read-write, prevents corrupt data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410759321.XA CN104376389B (en) | 2014-12-10 | 2014-12-10 | Master-slave mode microgrid power load prediction system and method based on load balancing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410759321.XA CN104376389B (en) | 2014-12-10 | 2014-12-10 | Master-slave mode microgrid power load prediction system and method based on load balancing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104376389A CN104376389A (en) | 2015-02-25 |
CN104376389B true CN104376389B (en) | 2017-09-15 |
Family
ID=52555284
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410759321.XA Active CN104376389B (en) | 2014-12-10 | 2014-12-10 | Master-slave mode microgrid power load prediction system and method based on load balancing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104376389B (en) |
Families Citing this family (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104951846B (en) * | 2015-06-02 | 2018-12-28 | 国电南京自动化股份有限公司 | Micro-capacitance sensor short term power and load prediction system and error classification dynamic correcting method |
CN105159093B (en) * | 2015-10-08 | 2018-06-29 | 国电南京自动化股份有限公司 | Microgrid energy Optimal Control System and its design method based on model adaptation |
TWI559250B (en) * | 2015-10-16 | 2016-11-21 | 行政院原子能委員會核能研究所 | Real-Time Scheduling Method for Micro-grid Energy Management |
CN107315571B (en) | 2016-04-27 | 2020-07-31 | 中科寒武纪科技股份有限公司 | Device and method for executing forward operation of full-connection layer neural network |
CN106330558A (en) * | 2016-08-31 | 2017-01-11 | 哈尔滨工业大学(威海) | Controller load prediction system and method applied to software defined network |
EP3388910A1 (en) * | 2017-04-10 | 2018-10-17 | ABB Schweiz AG | Method and apparatus for monitoring the condition of subsystems within a renewable generation plant or microgrid |
CN108229678B (en) * | 2017-10-24 | 2021-04-06 | 深圳市商汤科技有限公司 | Network training method, operation control method, device, storage medium and equipment |
CN107748934B (en) * | 2017-10-26 | 2023-06-23 | 国网新疆电力公司经济技术研究院 | Distribution network electric heating load prediction method based on improved BP neural network algorithm |
CN109242755A (en) * | 2018-08-01 | 2019-01-18 | 浙江深眸科技有限公司 | Computer vision processing server framework neural network based |
CN109272140B (en) * | 2018-08-10 | 2021-09-28 | 国网河北省电力有限公司石家庄供电分公司 | Big data analysis-based power transformation equipment operation period cost prediction method |
WO2020118587A1 (en) * | 2018-12-12 | 2020-06-18 | 华北电力大学扬中智能电气研究中心 | Load prediction method and device |
CN110232476B (en) * | 2019-05-30 | 2021-12-03 | 华中科技大学 | Micro-grid load prediction method based on deep learning |
CN110458342A (en) * | 2019-07-26 | 2019-11-15 | 国网江苏省电力有限公司金湖县供电分公司 | One kind monitoring system and method based on improved NARX neural network microclimate |
CN112131178B (en) * | 2020-09-07 | 2024-05-28 | 上海电气风电集团股份有限公司 | Data processing method and device for fan load calculation, and determining method and device |
CN113837480B (en) * | 2021-09-29 | 2023-11-07 | 河北工业大学 | Impact load prediction method based on improved GRU and differential error compensation |
CN116032190B (en) * | 2023-03-29 | 2023-05-30 | 海澜智云科技有限公司 | Motor load management system using neural network |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202126703U (en) * | 2011-06-15 | 2012-01-25 | 中国电力科学研究院 | Wind power optimization scheduling plan system |
CN103345227A (en) * | 2013-07-02 | 2013-10-09 | 东南大学 | Micro grid monitoring and energy management device and method |
CN103514487A (en) * | 2013-07-15 | 2014-01-15 | 国家电网公司 | Load forecasting method of power distribution network with distributed power supply |
CN103699944A (en) * | 2013-12-31 | 2014-04-02 | 国电南京自动化股份有限公司 | Wind and photovoltaic generation power prediction system with multiple prediction modes |
CN104008426A (en) * | 2014-05-15 | 2014-08-27 | 上海交通大学 | Distributed computing environment performance predicting method based on integrated learning |
CN104065101A (en) * | 2014-06-26 | 2014-09-24 | 华电电力科学研究院 | Distributed microgrid energy management system |
-
2014
- 2014-12-10 CN CN201410759321.XA patent/CN104376389B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN202126703U (en) * | 2011-06-15 | 2012-01-25 | 中国电力科学研究院 | Wind power optimization scheduling plan system |
CN103345227A (en) * | 2013-07-02 | 2013-10-09 | 东南大学 | Micro grid monitoring and energy management device and method |
CN103514487A (en) * | 2013-07-15 | 2014-01-15 | 国家电网公司 | Load forecasting method of power distribution network with distributed power supply |
CN103699944A (en) * | 2013-12-31 | 2014-04-02 | 国电南京自动化股份有限公司 | Wind and photovoltaic generation power prediction system with multiple prediction modes |
CN104008426A (en) * | 2014-05-15 | 2014-08-27 | 上海交通大学 | Distributed computing environment performance predicting method based on integrated learning |
CN104065101A (en) * | 2014-06-26 | 2014-09-24 | 华电电力科学研究院 | Distributed microgrid energy management system |
Non-Patent Citations (1)
Title |
---|
基于BP神经网络的负载均衡技术的研究;吴长伟;《中国优秀硕士学位论文全文数据库 信息科技辑》;20130515(第5期);第37页倒数第7行-38页第1行 * |
Also Published As
Publication number | Publication date |
---|---|
CN104376389A (en) | 2015-02-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104376389B (en) | Master-slave mode microgrid power load prediction system and method based on load balancing | |
Ren et al. | A fully data-driven method based on generative adversarial networks for power system dynamic security assessment with missing data | |
Osório et al. | Short-term wind power forecasting using adaptive neuro-fuzzy inference system combined with evolutionary particle swarm optimization, wavelet transform and mutual information | |
Li et al. | Building's electricity consumption prediction using optimized artificial neural networks and principal component analysis | |
Yezioro et al. | An applied artificial intelligence approach towards assessing building performance simulation tools | |
Zheng et al. | Bio‐inspired optimization of sustainable energy systems: A review | |
CN103489038A (en) | Photovoltaic ultra-short-term power prediction method based on LM-BP neural network | |
CN105354363A (en) | Fluctuation wind speed prediction method based on extreme learning machine | |
CN104951834A (en) | LSSVM (least squares support vector machine) wind speed forecasting method based on integration of GA (genetic algorithm) and PSO (particle swarm optimization) | |
US10474177B2 (en) | System and method for stability monitoring, analysis and control of electric power systems | |
CN107092989A (en) | The Forecasting Methodology and equipment of short-term wind-electricity power | |
CN103793887B (en) | Short-term electric load on-line prediction method based on self-adaptive enhancement algorithm | |
CN103365727A (en) | Host load forecasting method in cloud computing environment | |
CN106295857A (en) | A kind of ultrashort-term wind power prediction method | |
CN103927460A (en) | Wind power plant short-term wind speed prediction method based on RBF | |
CN104102951A (en) | Short-term wind power prediction method based on EMD (Empirical Mode Decomposition) historical data preprocessing | |
CN112101626A (en) | Distributed photovoltaic power generation power prediction method and system | |
CN104915889A (en) | Method of acquiring comprehensive load model parameters in online mode based on daily load curve | |
Patel et al. | Short term load forecasting of Indian system using linear regression and artificial neural network | |
Chandrasekaran | Feasibility study on machine‐learning‐based hybrid renewable energy applications for engineering education | |
Zhu et al. | Structural Safety Monitoring of High Arch Dam Using Improved ABC‐BP Model | |
Ge et al. | A virtual data collection model of distributed PVs considering spatio-temporal coupling and affine optimization reference | |
Alam et al. | AI-based efficiency analysis technique for photovoltaic renewable energy system | |
CN104361399A (en) | Solar irradiation intensity minute-scale predication method | |
Dourhmi et al. | Improved hourly prediction of BIPV photovoltaic power building using artificial learning machine: a case study |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |