CN101719667A - Distributed type management and transfer method of load - Google Patents
Distributed type management and transfer method of load Download PDFInfo
- Publication number
- CN101719667A CN101719667A CN200910188525A CN200910188525A CN101719667A CN 101719667 A CN101719667 A CN 101719667A CN 200910188525 A CN200910188525 A CN 200910188525A CN 200910188525 A CN200910188525 A CN 200910188525A CN 101719667 A CN101719667 A CN 101719667A
- Authority
- CN
- China
- Prior art keywords
- load
- neural network
- terminal
- distribution network
- transfer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012546 transfer Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000013528 artificial neural network Methods 0.000 claims abstract description 39
- 238000012549 training Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000004891 communication Methods 0.000 abstract description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 230000001537 neural effect Effects 0.000 description 15
- 230000006870 function Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 206010033799 Paralysis Diseases 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Images
Classifications
-
- Y04S10/54—
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
本发明涉及电力供配电技术领域，其公开了一种分布式管理与转移负荷的方法，包括以下步骤：(a)构造配网终端神经网络模块的输入向量和输出向量；(b)初始化神经网络；(c)用样本对神经网络进行训练；(d)用训练后的神经网络模块处理配网负荷管理与负荷转移。本发明的有益效果是：本发明通过采用神经网络模块来处理配电网络负荷管理与负荷转移，既能保证控制中心故障时能实现配电网的紧急负荷管理与转移；同时通过与负荷管理终端的通信，实现了分布式的负荷管理与转移，从而可应用于配电网络规模大的环境中。
The invention relates to the technical field of electric power supply and distribution, and discloses a method for distributed management and load transfer, comprising the following steps: (a) constructing the input vector and output vector of the distribution network terminal neural network module; (b) initializing the neural network module network; (c) use samples to train the neural network; (d) use the trained neural network module to handle distribution network load management and load transfer. The beneficial effects of the present invention are: the present invention handles distribution network load management and load transfer by adopting a neural network module, which can not only ensure that the emergency load management and transfer of the distribution network can be realized when the control center fails; The communication realizes the distributed load management and transfer, so it can be applied in the environment of large-scale power distribution network.
Description
Technical field
The present invention relates to the power supply and distribution of electric power technical field, the method for particularly a kind of distributed management and transfer load.
Background technology
Along with the expansion of scale of power and the increase of electric network composition complexity, the load management data that main website handles sharply increase, and the complexity of load transfer also increases thereupon, and trend is calculated and the realization difficulty and the processing time of network reconfiguration algorithm all increase thereupon.The present distribution network load management and the shortcoming of transfer load mode are that load management and load transfer must depend on the distribution main website, dependence to power distribution network communication is strong, when communication system breaks down or control centre's fault, then cause the The whole control system paralysis inevitably, can't realize the load management and the load transfer of power distribution network.
Summary of the invention
In order to solve the problems of the prior art, the invention provides the method for a kind of distributed management and transfer load, solve in the prior art and must depend on the distribution main website for overcoming existing load management and load transfer, can't realize the urgent load management and the load transfer of power distribution network during the control centre fault, the excessive problem of distribution network scale of distribution main website management simultaneously reaches distributed management control purpose.
The present invention solves the technical scheme that the prior art problem adopted: design and make the method for a kind of distributed management and transfer load, may further comprise the steps: (a) input vector and the output vector of structure distribution network terminal neural network module; (b) initialization neural net; (c) with sample neural net is trained; (d) handle distribution load management and load transfer with the neural network module after the training.
The present invention further improves: described step (a) further comprises: (a1) by the flow through input vector of load and placement configurations distribution network terminal neural network module of distribution network terminal; (a2) output vector of constructing the distribution network terminal neural network module by the switch and the circuit breaker deciliter state at distribution network terminal place.
The present invention further improves: in the described step (b), utilize electric power expertise initialization neural net; The described electric power expertise of utilizing is for utilizing the processing experience of electric power expert to load management and load transfer, determines the initial value that is connected weights in intermediate layer and input layer and output layer and intermediate layer.
The present invention further improves: described step (c) further comprises: (c1) calculate intermediate layer, each neuron output of output layer; (c2) calculation expectation output and actual output error; (c3) back transfer error is adjusted the weights that are connected in intermediate layer and input layer and output layer and intermediate layer; (c4) neural net is trained, if error then finishes training when meeting the demands, otherwise repetition training.
The present invention further improves: described step (d) further comprises: (d1) distribution network terminal detects fault current, receives the corresponding information of associated terminal; (d2) receive the corresponding information of associated terminal, draw the load that cuts off the switch that fault need disconnect and need to shift by exception handles; (d3) with the switch load information that receives and cut off the switch that fault need disconnect and the switch that needs to shift is flowed through information on load as the input of terminal neural network procedure, terminal is drawn the operation logic of block switch and interconnection switch and circuit breaker by corresponding neural network procedure output.
The present invention further improves: the method for described distributed management and transfer load also comprises step: (e) neural network module shifts after loading by the load management terminal cut-out again.
The present invention further improves: described step (e) further comprises: in the time of (e1) can't shifting fault feeder and all load, distribution network terminal draws the peak load that can shift; (e2) according to need transfer load amount with can shift the poor of maximal workload, according to the load of load importance and current load management terminal, cut-out is loaded again; (e3) the need transfer load amount behind the cut-out load is handled by described step (d).
The invention has the beneficial effects as follows: the present invention handles distribution network load management and load transfer by adopting neural network module, can realize the urgent load management and the transfer of power distribution network in the time of can guaranteeing control centre's fault; Simultaneously by with the communicating by letter of load management terminal, realized distributed load management and transfer, thereby can be applicable in the sweeping environment of distribution network.
Description of drawings
Fig. 1 is the method flow schematic diagram of distributed management of the present invention and transfer load.
Fig. 2 is distribution network terminal management and transfer load flow chart among the present invention.
Fig. 3 is that the distribution network terminal management is trained flow chart with the neural network procedure of transfer load among the present invention.
Fig. 4 is a distribution network terminal excision load management terminal load flow chart among the present invention.
Embodiment
The invention will be further described below in conjunction with accompanying drawing.
As Fig. 1, the method for a kind of distributed management and transfer load may further comprise the steps: the input vector and the output vector of a structure distribution network terminal neural network module; B initialization neural net; C trains neural net with sample; D handles distribution load management and load transfer with the neural network module after training.
Described step a further comprises: a1 is by distribution network terminal the flow through load and the input vector of placement configurations distribution network terminal neural network module; A2 is by the output vector of the switch and the circuit breaker deciliter state structure distribution network terminal neural network module at distribution network terminal place.
Among the described step b, utilize electric power expertise initialization neural net; The described electric power expertise of utilizing is for utilizing the processing experience of electric power expert to load management and load transfer, determines the initial value that is connected weights in intermediate layer and input layer and output layer and intermediate layer.
Described step c further comprises: c1 calculates intermediate layer, each neuron output of output layer; Output of c2 calculation expectation and actual output error; C3 back transfer error is adjusted the weights that are connected in intermediate layer and input layer and output layer and intermediate layer; C4 trains neural net, if error then finishes training when meeting the demands, otherwise repetition training.
Described steps d further comprises: the d1 distribution network terminal detects fault current, receives the corresponding information of associated terminal; D2 receives the corresponding information of associated terminal, draws the load that cuts off the switch that fault need disconnect and need to shift by exception handles; D3 is with the switch load information that receives and cut off the switch that fault need disconnect and the switch that needs to shift is flowed through information on load as the input of terminal neural network procedure, and terminal is drawn the operation logic of block switch and interconnection switch and circuit breaker by corresponding neural network procedure output.
The method of described distributed management and transfer load also comprises step: the e neural network module shifts after loading by the load management terminal cut-out again.
Described step e further comprises: when e1 can't shift fault feeder and all loads, distribution network terminal drew the peak load that can shift; E2 is according to need transfer load amount and can shift the poor of maximal workload, and according to the load of load importance and current load management terminal, cut-out is loaded again; Need transfer load amount behind the e3 cut-out load is handled by described steps d.
In one embodiment of the invention, in the system circuit breaker on the distribution, block switch and interconnection switch are used as node and are numbered, tectonic network is described matrix.Wherein block switch place distribution network terminal and circuit breaker place distribution network terminal and corresponding interconnection switch place distribution network terminal, by load and the placement configurations nerve network input parameter of flowing through, the switch at each a distribution network terminal place and circuit breaker deciliter state is exported as neural net.
Fig. 3 shows the neural network procedure training method of distribution management of the present invention and transfer load.
The present invention can adopt the BP neural net, based on the error propagation algorithm, must draw the output layer error by learning sample and teacher's sample, adjusts neural net by every layer error propagation, changes output error, so that output error is less than the assigned error index.Might be absorbed in local minimum during neural network learning, need limit the study number of times, when being absorbed in local minimum, can jump out circulation.
During the neural network training program, middle layer node number and the learning rate set in the program all can influence network convergence speed, promptly frequency of training are impacted.If the middle layer node number very little, network may can not train at all or network performance very poor; If the middle layer node number is too many, though the systematic error of network is reduced, net training time is prolonged, on the other hand, training is absorbed in local minimum point easily and can not get optimum point, also occurs easily " over-fitting ".Learning rate is value between 0 and 1, and value makes unstable networks more easily, and the less meeting of value makes the training time long.
During the neural network procedure initialization, need come neural net is got initial weight,, avoid being absorbed in the local extremum of non-requirement to reduce the training time of network by expertise.Expertise comes from the processing experience of existing electric power expert to distribution management and transfer load.
The learning sample of neural metwork training and teacher's sample come from the applied distribution network of distribution network terminal.
The BP neural net comprises input layer, intermediate layer and output layer, is a kind of error back propagation network.Its basic thought is a least square method, adopts the gradient search technology, so that the mean square error minimum of the real output value of network and desired value.Distribution network terminal in the input layer input distribution network flow through fault current and position, the intermediate layer draws distribution management and transfer load processing experience by the electric power expert with the initial weights that are connected of input layer, the node in intermediate layer is the weighted sum of input layer output, and the excitation function of node adopts S (Sigmoid) type function.
The input of the node of output layer is the weighted sum of middle layer node output, and output layer draws the processing experience of distribution fault by the electric power expert with the initial weights that are connected in intermediate layer.With the output result of output layer is that the switch at each distribution network terminal place and the desired output of circuit breaker deciliter state and teacher's sample are made comparisons, as do not meet and then change backpropagation over to, error signal is returned along original connecting path, by revising the neuronic weight coefficient of each layer, make switch that the output result who obtains on the output layer node is each distribution network terminal place and the error signal minimum between circuit breaker deciliter state and the desired output.
Get the study target function during neural network learning:
In the formula, y
d(t) be the desirable system output of current time; Y (t) is the actual output of current neural net.
Right for each sample data, begin to calculate the output valve of each node from the input node, and then bring into use backpropagation to calculate the partial derivative of all implicit nodes from output node by propagated forward, the broad sense learning rules are:
In the formula: η is a learning rate
Calculate the output layer generalized error during study earlier, calculate feedback error and adjust the output layer weight coefficient by the output layer generalized error then, adjust the input layer weight coefficient by feedback error again.
After a sample was finished the adjustment of network weight coefficient, it was right to send into another sample mode again, carries out similar study, up to the training study of finishing all samples.
Distribution network terminal management and transfer load method when Fig. 2 shows distribution network fault of the present invention.When distribution network broke down, distribution network terminal detected fault current, and received the corresponding information of associated terminal.After receiving the corresponding information of associated terminal, draw the load that cuts off the switch that fault need disconnect and need to shift by exception handles.With the switch load information that receives and cut off the switch that fault need disconnect and the switch that needs to shift is flowed through information on load as the input of terminal neural network procedure, terminal is drawn the operation logic of block switch and interconnection switch and circuit breaker by corresponding neural network procedure output.All load if can't shift fault feeder, then need to draw by corresponding neural network procedure output again the operation logic of block switch and interconnection switch and circuit breaker by the cut-out load.
Fig. 4 shows distribution network terminal excision load management terminal load flow process of the present invention.In the time of can't shifting fault feeder and all load, distribution network terminal draws the peak load that can shift, according to need transfer load amount with can shift the poor of maximal workload, according to the load of load importance and current load management terminal, the cut-out load management terminal is loaded again.Cut-out load back is drawn the operation logic of block switch and interconnection switch and circuit breaker by corresponding neural network procedure output with the need transfer load amount of this moment.
Above content be in conjunction with concrete preferred implementation to further describing that the present invention did, can not assert that concrete enforcement of the present invention is confined to these explanations.For the general technical staff of the technical field of the invention, without departing from the inventive concept of the premise, can also make some simple deduction or replace, all should be considered as belonging to protection scope of the present invention.
Claims (7)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910188525A CN101719667A (en) | 2009-12-01 | 2009-12-01 | Distributed type management and transfer method of load |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910188525A CN101719667A (en) | 2009-12-01 | 2009-12-01 | Distributed type management and transfer method of load |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101719667A true CN101719667A (en) | 2010-06-02 |
Family
ID=42434200
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200910188525A Pending CN101719667A (en) | 2009-12-01 | 2009-12-01 | Distributed type management and transfer method of load |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101719667A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102044913A (en) * | 2010-11-25 | 2011-05-04 | 深圳市科陆电子科技股份有限公司 | Method for managing load management terminal in power distribution terminal |
| CN102135760A (en) * | 2010-12-16 | 2011-07-27 | 天津工业大学 | Neural network energy coordinated controller for microgrid |
| CN103001217A (en) * | 2012-11-23 | 2013-03-27 | 山东电力集团公司 | Rapid distribution network load transfer method based on load balance |
| CN103490919A (en) * | 2013-09-02 | 2014-01-01 | 用友软件股份有限公司 | Fault management system and fault management method |
| CN117937498A (en) * | 2024-03-25 | 2024-04-26 | 天津中电华利电器科技集团有限公司 | Substation operation load optimization control method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19954950A1 (en) * | 1999-11-16 | 2001-05-17 | Abb Research Ltd | Detecting short circuit in distribution network, involves drawing conclusion regarding existence of short circuit by combined evaluation of threshold and uncertainty monitoring |
| CN101154807A (en) * | 2007-10-11 | 2008-04-02 | 天津大学 | Adaptive line selection method for single-phase-to-ground fault in distribution network based on transient zero-sequence current |
-
2009
- 2009-12-01 CN CN200910188525A patent/CN101719667A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19954950A1 (en) * | 1999-11-16 | 2001-05-17 | Abb Research Ltd | Detecting short circuit in distribution network, involves drawing conclusion regarding existence of short circuit by combined evaluation of threshold and uncertainty monitoring |
| CN101154807A (en) * | 2007-10-11 | 2008-04-02 | 天津大学 | Adaptive line selection method for single-phase-to-ground fault in distribution network based on transient zero-sequence current |
Non-Patent Citations (3)
| Title |
|---|
| 孙雅明等: "可靠性知识与最优评估配电网故障恢复（二）最优评估", 《电力系统自动化》 * |
| 廖犬发: "一种配电网故障区间诊断系统的研究", 《中国优秀博硕士学位论文全文数据库（硕士）工程科技II辑》 * |
| 毛鹏等: "基于分层分布式神经网络系统的高压架空输电线路故障测距的模型研究", 《天津电力技术》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102044913A (en) * | 2010-11-25 | 2011-05-04 | 深圳市科陆电子科技股份有限公司 | Method for managing load management terminal in power distribution terminal |
| CN102135760A (en) * | 2010-12-16 | 2011-07-27 | 天津工业大学 | Neural network energy coordinated controller for microgrid |
| CN103001217A (en) * | 2012-11-23 | 2013-03-27 | 山东电力集团公司 | Rapid distribution network load transfer method based on load balance |
| CN103490919A (en) * | 2013-09-02 | 2014-01-01 | 用友软件股份有限公司 | Fault management system and fault management method |
| CN117937498A (en) * | 2024-03-25 | 2024-04-26 | 天津中电华利电器科技集团有限公司 | Substation operation load optimization control method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114217524B (en) | Power grid real-time self-adaptive decision-making method based on deep reinforcement learning | |
| CN101763034B (en) | Method for realizing self-adapting network reconfiguration at distribution network terminal | |
| CN104124688B (en) | A kind of heuristic reconstruction method of power distribution network based on minimum spanning tree | |
| CN101719667A (en) | Distributed type management and transfer method of load | |
| CN111191918A (en) | A service routing planning method and device for a smart grid communication network | |
| CN101702537A (en) | Method of Self-Adaptive Handling of Faults in Distribution Network Terminals | |
| CN107958287A (en) | Towards the confrontation transfer learning method and system of big data analysis transboundary | |
| CN108037659B (en) | Time-varying coupling complex dynamic network synchronization method based on event driving | |
| CN113179175B (en) | Real-time bandwidth prediction method and device for power communication network service | |
| CN114384931B (en) | Multi-target optimal control method and equipment for unmanned aerial vehicle based on strategy gradient | |
| CN106487003A (en) | A kind of method of main Distribution Network Failure recovery and optimization scheduling | |
| CN113872198A (en) | A fault recovery method for active distribution network based on reinforcement learning method | |
| CN117014355A (en) | TSSDN dynamic route decision method based on DDPG deep reinforcement learning algorithm | |
| Hou et al. | Adaptive neural network tracking control for switched uncertain non‐linear systems with actuator failures and time‐varying delays | |
| CN106339817A (en) | Multi-region dynamic economy scheduling method and system | |
| CN111105025A (en) | Congestion management method for urban high-voltage distribution network based on data-driven heuristic optimization | |
| CN112668774B (en) | Dynamic resource optimizing configuration method in bridge network post-disaster repair process | |
| CN114154688A (en) | Short-term power prediction method for photovoltaic power station | |
| CN109102067A (en) | The method of increase and decrease certainly, computer equipment and the storage medium of neural network node | |
| CN115660324B (en) | Power grid multi-section transgression control method and system based on graph reinforcement learning | |
| CN117395687A (en) | Star-ground fusion network node resource deployment method, system and electronic equipment | |
| CN104991447B (en) | A kind of fired power generating unit Changing load-acceleration Forecasting Methodology based on wavelet neural network | |
| CN117057623A (en) | Comprehensive power grid safety optimization scheduling method, device and storage medium | |
| CN110968973A (en) | Simulation method and control method for transformer model, electronic device and storage medium | |
| CN114362188B (en) | Multi-terminal intelligent soft switching voltage control method based on deep reinforcement learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20100602 |