CN109901389A - A kind of new energy consumption method based on deep learning - Google Patents
A kind of new energy consumption method based on deep learning Download PDFInfo
- Publication number
- CN109901389A CN109901389A CN201910157651.4A CN201910157651A CN109901389A CN 109901389 A CN109901389 A CN 109901389A CN 201910157651 A CN201910157651 A CN 201910157651A CN 109901389 A CN109901389 A CN 109901389A
- Authority
- CN
- China
- Prior art keywords
- new energy
- deep learning
- model
- training
- data
- 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
Abstract
The invention belongs to generations of electricity by new energy to dissolve technical field, disclose a kind of new energy consumption method based on deep learning, the new energy consumption method based on deep learning is realized using the back-propagation algorithm of multi-layered perception neural networks, is trained using improved Dynamical Recurrent Neural Networks method；Optimization process include: deep learning model line under training process, another part is the on-line optimization process of deep learning optimal controller.The present invention is based on the consumption optimization algorithms of deep learning to determine optimal controller parameter online according to the content and quantity of Different Optimization target, the content and quantity of various boundary conditions, without artificial adjustment, has universal adaptability for application；The present invention uses the dynamic recurrent neural networks model training method based on adaptive updates coefficient conventional Dynamical Recurrent Neural Networks convergence rate can be overcome slow and the shortcomings that being easily trapped into local minimum, and the training time of optimal controller model is shortened under the premise of guarantee precision.
Description
Technical field
The invention belongs to generation of electricity by new energy consumption technical field more particularly to a kind of new energy consumptions based on deep learning
Method.
Background technique
Currently, the problems such as prior art commonly used in the trade is such that with energy security, ecological environment, climate change
It is increasingly serious, it accelerates development new energy and has become the common recognition method for pushing energy Transformation Development, coping with Global climate change.Wherein
Wind-power electricity generation and photovoltaic power generation have become the development of clean energy side that with fastest developing speed, technology is most mature, Commercial Prospect is best
Formula.But the characteristics of randomness of wind-power electricity generation and photovoltaic power generation active power output, intermittence, fluctuation, causes the big rule of new energy
A series of problems, such as access, scheduling, the influence to operation of power networks and consumption difficulty that mould exploitation faces.Compared with foreign countries, China
New energy is contributed, and fluctuation is stronger, and extensive new energy concentrates " three Norths " area power grid of access weakness relatively, local consumption energy
Power is low, and electric power sends that distance is remote, capacity is big outside, and flexible modulation power supply lacks in addition, generation of electricity by new energy safe operation and effectively consumption
Problem is more prominent.How to realize that the efficient consumption of new energy becomes an important factor in order of limitation new energy development.Shadow
The factor for ringing new energy digestion capability is numerous, and uncertain strong, does not there is the effective consumption method adapted to extensively also at present.It establishes
The new energy of the accurate a variety of uncertain factors of consideration dissolves mathematical model, and the consumption optimization algorithm using intelligence is to improve new energy
The effective ways of source consumption rate.
In conclusion current extensive new energy concentrates on " three Norths " area, and the region local consumption ability is low, electric power
Send that distance is remote, capacity is big outside, flexible modulation power supply lacks in addition, and generation of electricity by new energy safe operation and effectively consumption problem are more prominent
Out.Obviously, new energy consumption problems demand solves, however, the uncertain factor for influencing new energy consumption is numerous, it tends to be difficult to quasi-
Really modeling, the conventional effect for dissolving algorithm are difficult to meet actual demand.For this reason, it may be necessary to propose a kind of new energy based on deep learning
Source dissolves method, and abandonment is effectively reduced and abandons light ratio, this is for promoting domestic utilization of new energy resources rate, reducing the non-renewable energy of tradition
Source loss and discharge are of great significance.
Summary of the invention
In view of the problems of the existing technology, the present invention provides a kind of, and the new energy based on deep learning dissolves method.
The invention is realized in this way a kind of new energy based on deep learning dissolves method, it is described to be based on deep learning
New energy consumption method realized using the back-propagation algorithm of multi-layered perception neural networks, using improved Dynamic Recurrent nerve
Network method is trained；Optimization process include: deep learning model line under training process, another part is that deep learning is excellent
Change the on-line optimization process of controller.
Further, deep learning controller learning model building mistake under the line of the new energy consumption method based on deep learning
Journey includes generating training set, design object function and constraint condition, model training process.
Further, the new energy consumption method based on deep learning specifically includes:
The first step, for different types of data, generating training set, there are two types of methods: being generated based on power grid actual operating data
It is generated with using theoretical modeling mode；
Second step dissolves target data, economic indicator requirement generation optimization object function and constraint condition according to new energy,
Objective function expression formula is as follows:
The constraint condition of optimization is chosen from the new energy consumption model established according to the actual situation；
Third step optimizes controller model training.The training set data of generation is inputted into optimal controller, if knot
Fruit meets control target call, terminates study, exports deep learning controller model；If being as a result unsatisfactory for control target to want
It asks, then returns to modification model parameter or objective function, re -training, until meeting optimization aim requirement；
4th step, by Optimized model be added actual schedule system, be arranged objective function, input/access power grid real data,
Generate optimum results.
Further, the first step specifically includes:
(1) training set is generated based on power grid actual operating data: includes thermoelectricity, Hydropower Unit, power grid profile constraints, energy storage
Device, the data that electrical grid transmission restrained condition determines, according to actual test data in the period according to training data call format
Generate training set；
(2) training set is generated using theoretical modeling mode: including wind-powered electricity generation, photoelectricity uncertain data, by using foundation
Prediction model generates required training data.
Further, the third step specifically includes: the method that adaptive updates coefficient is introduced in weights learning algorithm is come
The training process of conventional Dynamical Recurrent Neural Networks is improved, steps are as follows for specific improvement:
(1) according to the neural network model of foundation, algorithm mathematics model is established:
Wherein u is input vector, and w is the weighted value of input vector, and v is input layer output, k1For input layer to hidden layer
Weight, k2For the weight of feedback layer, net is hidden layer output, w1For the weighted value of hidden layer to output layer, y is that network is defeated
Out；
(2) defining systematic error is e (k)=y*(k)-y (k), wherein y*It (k) is desired output, conventional Dynamic Recurrent nerve
The weight w of network1Adjustment algorithm are as follows:
Wherein η1For learning rate；
Weighed value adjusting algorithm after introducing adaptive updates factor alpha are as follows:
Wherein α1For w1Adaptive updates coefficient；
For w, k1, k2Improved adjustment algorithm and w1Adjustment algorithm principle it is consistent, weighed value adjusting publicity is respectively as follows:
W (k+1)=α2w(k)-η2(1-α2)e(k)w(k)×(1-z(k)z(k))net(k)u；
k1(k+1)=α3k1(k)-η3(1-α3)e(k)w1(k)×(1-z(k)z(k))v(k)；
k2(k+1)=α4k2(k)-η4(1-α4)e(k)w1(k)×(1-z(k)z(k))net(k-1)；
Wherein η2, η3, η4For learning rate, α2, α3, α4For adaptive updates coefficient.
Another object of the present invention is to provide the new of a kind of new energy consumption method described in application based on deep learning
Energy electric generation management platform.
In conclusion advantages of the present invention and good effect are as follows: new energy of the invention dissolves mathematical model, from source, net,
The influence factor for influencing new energy consumption nearly all so far is considered in terms of lotus three, there are model other methods not have
Standby fidelity；New energy proposed by the invention dissolves mathematical model, receives maximum and overall cost minimum with new energy
Target establishes objective function, it is contemplated that target call of both environment sustainable development and economy.
The present invention is based on the consumption optimization algorithms of deep learning can be different according to the content and quantity of Different Optimization target
The content and quantity of constraint condition determine optimal controller parameter online, without artificial adjustment, have for application universal
Adaptability；The dynamic recurrent neural networks model training method based on adaptive updates coefficient that the present invention uses can overcome
The shortcomings that conventional Dynamical Recurrent Neural Networks convergence rate is slow and is easily trapped into local minimum, contracts under the premise of guaranteeing precision
The short training time of optimal controller model.
The present invention is directed to the deep learning optimization method of new energy consumption, and training set is for different types of data using real
The acquisition of border electric network data and theoretical modeling generate two ways and establish, and it is raw both to have ensure that the reliability of data in turn simplified training set
At process.
Detailed description of the invention
Fig. 1 is the new energy consumption method flow diagram provided in an embodiment of the present invention based on deep learning.
Fig. 2 is provided in an embodiment of the present invention using neural network structure schematic diagram.
Fig. 3 is the realization procedure chart of optimization algorithm provided in an embodiment of the present invention.
Fig. 4 is improved recurrent neural networks model schematic diagram provided in an embodiment of the present invention.
Fig. 5 is the training result comparison diagram of inventive algorithm and conventional Dynamical Recurrent Neural Networks algorithm.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
The present invention proposes a kind of new energy consumption mathematical model for considering a variety of uncertain factors, to improve disappearing for new energy
The rate of receiving provides a kind of intelligent consumption optimization algorithm.
Application principle of the invention is explained in detail with reference to the accompanying drawing.
As shown in Figure 1, it is provided in an embodiment of the present invention based on deep learning new energy consumption method the following steps are included:
S101: it is realized using the back-propagation algorithm of multi-layered perception neural networks, using improved Dynamic Recurrent nerve net
Network method is trained；
S102: it is trained under the line of deep learning model, the on-line optimization of deep learning optimal controller.
It is provided in an embodiment of the present invention based on deep learning new energy consumption method specifically includes the following steps:
The invention proposes a kind of new energy for considering a variety of uncertain factors to dissolve mathematical model, and system modelling process is such as
Under:
Consider target call of both environment sustainable development and economy, maximum and overall cost is received with new energy
Minimum target establishes new energy consumption model, and the objective function of model is as follows:
Wherein:
Indicate fuel cost function of the unit j in moment t in the n of region；
Be in the n of region unit j in the active power output of t moment；
For the fuel cost coefficient of unit j in the n of region；
Indicate that unit j is in the operating status of t moment in the n of region；
T indicates the period (1≤t≤T) of scheduling, and general T is that 96, t indicates region index number (1≤t≤N)；
Indicate that wind-powered electricity generation is in the active power output of t moment in the n of region；
Indicate that photovoltaic is in the active power output of t moment in the n of region；
SnIt is the unit total number in the n of region；
It is the starting coal consumption of unit j in the n of region；
It is starting state of the unit j in moment t in the n of region；
It is the shutdown coal consumption of unit j in the n of region；
It is shutdown status of the unit j in moment t in the n of region；
The constraint condition for influencing new energy consumption is as follows:
1. Unit Commitment machine operating status logical constraint:
2. region account load balancing constraints:
I indicates transmission line call number；
Indicate quantity of state of i-th transmission lines in the n of region；
Indicate the transmission power on the i-th transmission lines of t moment；
Indicate region n in the electric load of t moment；
3. unit active power output constrains:
WithIndicate the minimum and maximum active power output of unit j in the n of region；
4. climbing rate constrains:
Δ T is time interval；
WithIt respectively indicates unit j in the n of region and most increases and contribute rate and minimum subtracts power output rate；
5. spinning reserve constrains:
SpAnd SnThe respectively spare and negative spinning reserve capacity of system positive rotation；
CtIndicate new energy in the credible capacity of period t；
6. interconnection transmission constraint:
-LI, max≤Li≤LI, max；
-Li,maxAnd Li,maxIt is expressed as the bound of the i-th transmission lines transmission capacity, inflow region is positive, and flows out area
Domain is negative；
7. wind-powered electricity generation, photoelectric power constraint:
It contributes for region n in the theory of t moment wind-powered electricity generation；
It contributes for region n in the theory of t moment photoelectricity；
8. energy storage device constraint condition:
1) capacity-constrained:
Emin≤E(t)≤Emax；
EminAnd EmaxThe maximin of capacity of energy storing device is respectively indicated, E (t) is capacity of the energy storage device in t moment；
2) charge-discharge electric power constrains:
Uc(t)+Ud(t)=1
Pc(t)、Pd(t) energy storage device charging, discharge power are respectively indicated；
Uc(t) and Ud(t) be t moment energy storage device charge-discharge behavior variable；
For the maximum chargeable power of energy storage device t moment；
Δ E=Emax-E(t-1)；
Power generating value is predicted for t moment wind-powered electricity generation；
It can discharge power for energy storage device t moment maximum；
WithFor energy storage device maximum charge, discharge power.
3) stored energy capacitance variation constraint:
E (t+1)=E (t)+Ec(t)-Ed(t)；
Ec(t) and Ed(t) energy storage device is respectively indicated in the charge volume and discharge capacity of t moment；
ηcAnd ηdIt respectively fills, discharge coefficient；
9. wind flame, which is combined, sends constraint outside；
1) combine and send power stability constraint outside:
KBmin≤KBt≤KBmax；
KBtIndicate the combined power that t moment bundling is sent outside；
KBminAnd KBmaxIndicate the minimax power requirement sent outside.
2) general power that bundling is sent outside is greater than the sum of wind-powered electricity generation and the performance number of photoelectricity:
WithIndicate t moment bundling send outside in wind-powered electricity generation and photoelectric power.
3) power output of fired power generating unit is greater than the activity of force out that bundling sends middle thermoelectricity outside:
Indicate the power output of fired power generating unit.
The present invention gives a kind of intelligent new energy to dissolve optimization algorithm, and the algorithm is using multi-layered perception neural networks
Back-propagation algorithm realize, used neural network structure as shown in Fig. 2, using improved Dynamical Recurrent Neural Networks method into
Row training.The realization process of optimization algorithm as shown in figure 3, specifically include the following steps:
New energy consumption strategy uses the optimization method based on deep learning.Optimization process is mainly divided into two parts, a part
Be deep learning model line under training process, another part is the on-line optimization process of deep learning optimal controller.It is described
Deep learning controller learning model building process includes generating training set, design object function and constraint condition, model training under line
Etc. processes.
1. being directed to different types of data, generating training set, there are two types of methods, and one is raw based on power grid actual operating data
At one is generated using theoretical modeling mode:
1) generate training set based on power grid actual operating data: this kind of data mainly include thermoelectricity, Hydropower Unit, and power grid is disconnected
Face constraint, energy storage device, electrical grid transmission constraint etc. states determine data, according to actual test data in certain period of time according to
Training data call format generates training set；
2) generate training set using theoretical modeling mode: this kind of data mainly include the uncertain datas such as wind-powered electricity generation, photoelectricity, are led to
Cross the training data needed for generating using the prediction model established.
2. dissolving the requirements such as target data, economic indicator according to new energy generates optimization object function and constraint condition, mesh
Scalar functions expression formula is as follows:
The constraint condition of optimization is chosen from the new energy consumption model established according to the actual situation.
3. optimizing controller model training.The training set data of generation is inputted into optimal controller, if result is full
Foot control target call, terminates study, exports deep learning controller model；If being as a result unsatisfactory for control target call,
Modification model parameter or objective function, re -training are returned to, until meeting optimization aim requirement.
It the shortcomings that in order to overcome conventional Dynamical Recurrent Neural Networks convergence rate slowly and be easily trapped into local minimum, is weighing
The method of adaptive updates coefficient is introduced in value learning algorithm to improve the training process of conventional Dynamical Recurrent Neural Networks, specifically
Steps are as follows for improvement:
1) improved recurrent neural networks model is established, as shown in Figure 4.Wherein u is input vector, and w is input vector
Weighted value, v are input layer output, k1For the weight of input layer to hidden layer, k2For the weight of feedback layer, net is that hidden layer is defeated
Out, w1For the weighted value of hidden layer to output layer, y is network output.
2) according to the neural network model of foundation, algorithm mathematics model is established:
3) defining systematic error is e (k)=y*(k)-y (k), wherein y*It (k) is desired output, conventional Dynamic Recurrent nerve
The weight w of network1Adjustment algorithm are as follows:
Wherein η1For learning rate.
Weighed value adjusting algorithm after introducing adaptive updates factor alpha are as follows:
Wherein α1For w1Adaptive updates coefficient.
For w, k1, k2Improved adjustment algorithm and w1Adjustment algorithm principle it is consistent, weighed value adjusting formula is respectively as follows:
W (k+1)=α2w(k)-η2(1-α2)e(k)w(k)×(1-z(k)z(k))net(k)u；
k1(k+1)=α3k1(k)-η3(1-α3)e(k)w1(k)×(1-z(k)z(k))v(k)；
k2(k+1)=α4k2(k)-η4(1-α4)e(k)w1(k)×(1-z(k)z(k))net(k-1)；
Wherein η2, η3, η4For learning rate, α2, α3, α4For adaptive updates coefficient.
4. actual schedule system is added in Optimized model, objective function is set, input/access power grid real data generates
Optimum results.
Embodiment:
The main purpose of the present embodiment is to verify the validity and convergent rapidity of inventive algorithm.Using sweet in experiment
Respectful regional 1~June in 2018, totally 17376 groups of scene actual powers, consumption data were trained for training set, wherein study speed
Rate is uniformly set as 1.5, and adaptive updates coefficient adjustment range is 0~2, and training the number of iterations maximum value is 1000, training essence
Degree is 10-3.Inventive algorithm and the training result of conventional Dynamical Recurrent Neural Networks are as shown in Figure 5.It can be seen from the figure that two
Kind algorithm is finally attained by convergence precision requirement, and inventive algorithm has faster convergence rate at training initial stage, and reaches instruction
The number of iterations needed for practicing precision is less.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (6)
1. a kind of new energy based on deep learning dissolves method, which is characterized in that the new energy based on deep learning disappears
Method of receiving using multi-layered perception neural networks back-propagation algorithm realize, using improved Dynamical Recurrent Neural Networks method into
Row training；Optimization process include: deep learning model line under training process, another part is deep learning optimal controller
On-line optimization process.
2. the new energy based on deep learning dissolves method as described in claim 1, which is characterized in that described to be based on depth
Deep learning controller learning model building process includes generating training set, design object function under the line of the new energy consumption method of habit
And constraint condition, model training process.
3. the new energy based on deep learning dissolves method as described in claim 1, which is characterized in that described to be based on depth
The new energy consumption method of habit specifically includes:
The first step, for different types of data, generating training set, there are two types of methods: being generated and is adopted based on power grid actual operating data
It is generated with theoretical modeling mode；
Second step dissolves target data according to new energy, economic indicator requires to generate optimization object function and constraint condition, target
Function expression is as follows:
The constraint condition of optimization is chosen from the new energy consumption model established according to the actual situation；
Third step optimizes controller model training；The training set data of generation is inputted into optimal controller, if result is full
Foot control target call, terminates study, exports deep learning controller model；If being as a result unsatisfactory for control target call,
Modification model parameter or objective function, re -training are returned to, until meeting optimization aim requirement；
Actual schedule system is added in Optimized model by the 4th step, and objective function is arranged, and input/access power grid real data generates
Optimum results.
4. the new energy based on deep learning dissolves method as claimed in claim 3, which is characterized in that the first step is specific
Include:
(1) training set is generated based on power grid actual operating data: includes thermoelectricity, Hydropower Unit, power grid profile constraints, energy storage dress
It sets, the data that electrical grid transmission restrained condition determines, it is raw according to training data call format according to actual test data in the period
At training set；
(2) training set is generated using theoretical modeling mode: including wind-powered electricity generation, photoelectricity uncertain data, by using the prediction of foundation
Model generates required training data.
5. the new energy based on deep learning dissolves method as described in claim 1, which is characterized in that the third step is specific
It include: that the method for adaptive updates coefficient is introduced in weights learning algorithm to improve the training of conventional Dynamical Recurrent Neural Networks
Process, steps are as follows for specific improvement:
(1) according to the neural network model of foundation, algorithm mathematics model is established:
Wherein u is input vector, and w is the weighted value of input vector, and v is input layer output, k1For the power of input layer to hidden layer
Value, k2For the weight of feedback layer, net is hidden layer output, w1For the weighted value of hidden layer to output layer, y is network output；
(2) defining systematic error is e (k)=y*(k)-y (k), wherein y*It (k) is desired output, conventional Dynamical Recurrent Neural Networks
Weight w1Adjustment algorithm are as follows:
Wherein η1For learning rate；
Weighed value adjusting algorithm after introducing adaptive updates factor alpha are as follows:
Wherein α1For w1Adaptive updates coefficient；
For w, k1, k2Improved adjustment algorithm and w1Adjustment algorithm principle it is consistent, weighed value adjusting publicity is respectively as follows:
W (k+1)=α2w(k)-η2(1-α2)e(k)w(k)×(1-z(k)z(k))net(k)u；
k1(k+1)=α3k1(k)-η3(1-α3)e(k)w1(k)×(1-z(k)z(k))v(k)；
k2(k+1)=α4k2(k)-η4(1-α4)e(k)w1(k)×(1-z(k)z(k))net(k-1)；
Wherein η2, η3, η4For learning rate, α2, α3, α4For adaptive updates coefficient.
6. a kind of new energy using the new energy consumption method described in Claims 1 to 5 any one based on deep learning is sent out
Fulgurite platform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910157651.4A CN109901389A (en) | 2019-03-01 | 2019-03-01 | A kind of new energy consumption method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910157651.4A CN109901389A (en) | 2019-03-01 | 2019-03-01 | A kind of new energy consumption method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109901389A true CN109901389A (en) | 2019-06-18 |
Family
ID=66946132
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910157651.4A Pending CN109901389A (en) | 2019-03-01 | 2019-03-01 | A kind of new energy consumption method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109901389A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472780A (en) * | 2019-07-30 | 2019-11-19 | 安徽匠桥电子信息有限公司 | A kind of distribution Optimization Scheduling neural network based |
| CN111047345A (en) * | 2019-10-11 | 2020-04-21 | 国网甘肃省电力公司电力科学研究院 | Electricity price data adjusting method considering new energy consumption rate and user satisfaction degree |
| CN111262275A (en) * | 2020-03-26 | 2020-06-09 | 西安电子科技大学 | Global cooperative control method for simulated space solar power station |
| CN111967179A (en) * | 2020-07-02 | 2020-11-20 | 江苏能来能源互联网研究院有限公司 | Dynamic optimization matching method for energy units of energy Internet |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105631599A (en) * | 2015-12-30 | 2016-06-01 | 国网甘肃省电力公司电力科学研究院 | Multi-target operation scheduling method of virtual power plant |
| US20170004399A1 (en) * | 2015-07-01 | 2017-01-05 | Ricoh Company, Ltd. | Learning method and apparatus, and recording medium |
| CN107832542A (en) * | 2017-11-22 | 2018-03-23 | 国网河南省电力公司电力科学研究院 | A kind of Unit Combination Optimization Scheduling based on spatial and temporal scales consumption scene |
| CN109063992A (en) * | 2018-07-18 | 2018-12-21 | 国网重庆市电力公司经济技术研究院 | Consider the power distribution network Expansion Planning method of regional complex energy resource system optimization operation |
| CN109256810A (en) * | 2018-11-14 | 2019-01-22 | 南京邮电大学 | Consider that blower is contributed and does not know the Multipurpose Optimal Method of cost |
-
2019
- 2019-03-01 CN CN201910157651.4A patent/CN109901389A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170004399A1 (en) * | 2015-07-01 | 2017-01-05 | Ricoh Company, Ltd. | Learning method and apparatus, and recording medium |
| CN105631599A (en) * | 2015-12-30 | 2016-06-01 | 国网甘肃省电力公司电力科学研究院 | Multi-target operation scheduling method of virtual power plant |
| CN107832542A (en) * | 2017-11-22 | 2018-03-23 | 国网河南省电力公司电力科学研究院 | A kind of Unit Combination Optimization Scheduling based on spatial and temporal scales consumption scene |
| CN109063992A (en) * | 2018-07-18 | 2018-12-21 | 国网重庆市电力公司经济技术研究院 | Consider the power distribution network Expansion Planning method of regional complex energy resource system optimization operation |
| CN109256810A (en) * | 2018-11-14 | 2019-01-22 | 南京邮电大学 | Consider that blower is contributed and does not know the Multipurpose Optimal Method of cost |
Non-Patent Citations (5)
| Title |
|---|
| FANGDI ZENG 等: "Annual renewable energy planning platform- Methodology and design", 《IEEE》 * |
| FANGDI ZENG 等: "Long-term wind accommodation of interconnected power grids via HVDC tie-line based on aggregate unit model", 《IEEE》 * |
| 李慧民 等著: "《冶金工业出版社》", 30 April 2018 * |
| 赵妍著: "《电子科技大学出版社》", 31 July 2016 * |
| 陈雯柏编: "《西安电子科技大学出版社》", 31 January 2016 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472780A (en) * | 2019-07-30 | 2019-11-19 | 安徽匠桥电子信息有限公司 | A kind of distribution Optimization Scheduling neural network based |
| CN111047345A (en) * | 2019-10-11 | 2020-04-21 | 国网甘肃省电力公司电力科学研究院 | Electricity price data adjusting method considering new energy consumption rate and user satisfaction degree |
| CN111047345B (en) * | 2019-10-11 | 2022-04-12 | 国网甘肃省电力公司电力科学研究院 | Electricity price data adjusting method considering new energy consumption rate and user satisfaction degree |
| CN111262275A (en) * | 2020-03-26 | 2020-06-09 | 西安电子科技大学 | Global cooperative control method for simulated space solar power station |
| CN111262275B (en) * | 2020-03-26 | 2023-02-24 | 西安电子科技大学 | Global cooperative control method for simulated space solar power station |
| CN111967179A (en) * | 2020-07-02 | 2020-11-20 | 江苏能来能源互联网研究院有限公司 | Dynamic optimization matching method for energy units of energy Internet |
| CN111967179B (en) * | 2020-07-02 | 2024-02-09 | 江苏能来能源互联网研究院有限公司 | Dynamic optimization matching method for energy units of energy internet |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109901389A (en) | A kind of new energy consumption method based on deep learning | |
| Xi et al. | Automatic generation control based on multiple-step greedy attribute and multiple-level allocation strategy | |
| CN109361237A (en) | Based on the micro-capacitance sensor capacity configuration optimizing method for improving Hybrid Particle Swarm | |
| CN106712075B (en) | A kind of peak regulation policy optimization method considering wind-electricity integration system security constraint | |
| Xi et al. | A virtual generation ecosystem control strategy for automatic generation control of interconnected microgrids | |
| CN110350512A (en) | A kind of Itellectualized uptown generation of electricity by new energy station method for optimizing scheduling and system | |
| Yin et al. | Fuzzy vector reinforcement learning algorithm for generation control of power systems considering flywheel energy storage | |
| CN113644670A (en) | Method and system for optimally configuring energy storage capacity | |
| CN116667325A (en) | Micro-grid-connected operation optimization scheduling method based on improved cuckoo algorithm | |
| Hu et al. | Optimal dispatch of combined heat and power units based on particle swarm optimization with genetic algorithm | |
| CN107359611A (en) | Consider the power distribution network equivalence method of a variety of enchancement factors | |
| CN114285075A (en) | Micro-grid energy online optimization method based on distributed deep reinforcement learning | |
| CN113723793A (en) | Method, device, equipment and medium for realizing park comprehensive energy system | |
| CN112734116A (en) | Optimal scheduling method for active power distribution network containing distributed energy storage | |
| Liu et al. | Day-ahead economic dispatch of renewable energy system considering wind and photovoltaic predicted output | |
| CN117134409A (en) | Micro-grid system considering electro-hydro-thermal complementation and multi-objective optimal configuration method thereof | |
| CN110098623A (en) | A kind of Prosumer unit control method based on intelligent load | |
| CN115796533A (en) | Virtual power plant double-layer optimization scheduling method and device considering clean energy consumption | |
| CN112217198B (en) | Photovoltaic power generation management method for multi-step graph neural network self-adaptive dynamic planning | |
| CN115411776A (en) | Thermoelectric cooperative scheduling method and device for residential comprehensive energy system | |
| CN110808579A (en) | Active power distribution network source load coordination operation method | |
| CN115021336A (en) | Optimization method for high-proportion new energy power system inertia deficiency scene | |
| Yu et al. | A fuzzy Q-learning algorithm for storage optimization in islanding microgrid | |
| Lagouir et al. | Optimal power flow management strategies of MicroGrid, using Petri Nets and fuzzy logic approaches | |
| Zhu et al. | Power system dynamic economic dispatch with multi-type renewable energy |
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: 20190618 |
|
| RJ01 | Rejection of invention patent application after publication |