CN109494766A - A kind of intelligent power generation control method of manual depth's emotion game intensified learning - Google Patents
A kind of intelligent power generation control method of manual depth's emotion game intensified learning Download PDFInfo
- Publication number
- CN109494766A CN109494766A CN201811408219.XA CN201811408219A CN109494766A CN 109494766 A CN109494766 A CN 109494766A CN 201811408219 A CN201811408219 A CN 201811408219A CN 109494766 A CN109494766 A CN 109494766A
- Authority
- CN
- China
- Prior art keywords
- emotion
- learning
- intensified learning
- power generation
- game
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/24—Arrangements for preventing or reducing oscillations of power in networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Feedback Control In General (AREA)
Abstract
The present invention proposes that artificial emotion nitrification enhancement and deeply learning algorithm are combined by a kind of intelligent power generation control method of manual depth's emotion game intensified learning, this method, in intelligent power generation control.This method has merged artificial emotion, intensified learning, deep neural network algorithm and theory of games in artificial intelligence simultaneously.Manual depth's emotion nitrification enhancement incorporates deep neural network and artificial emotion in intensified learning frame, probability updating strategy of the deep neural network to improve traditional intensified learning;Artificial emotion quantifies electricity generation system information with emotion quantization function, to the more learning rate of New Tradition intensified learning, discount factor and output action value.Mentioned method is using frequency departure and area control error as input, using the power instruction of generating set as output.The intelligent power generation control method of mentioned manual depth's emotion game intensified learning has validity, feasibility, strong robustness, rapidity.
Description
Technical field
The invention belongs to the Generation Control fields of interacted system extensive in electric system, belong to the neck of Automatic Generation Control
Domain.
Background technique
As the fast development of renewable energy and new technique are constantly applied, modern power systems are then faced with
Key challenge, such as the stability problem of active power and frequency.Therefore, Automatic Generation Control, which has begun to be widely implemented, comes
It solves these problems, the purpose of Automatic Generation Control is to make area control error in each region and system frequency deviation most
It is small.It is past in decades, people have been devoted to research Automatic Generation Control control method, for example, proportional, integral-is micro-
Divide algorithm, sliding mode control algorithm and fractional order proportional-integral-differential algorithm etc..But these controls with intrinsic optimized parameter
Algorithm processed can not meet the control requirement of system, and these control algolithms with intrinsic optimized parameter can only be at one
Optimize active power in specific region rather than all controlled areas.
In order to further improve the performance of controller, carrying out balance system using intelligent power generation control in electric system has
Function power, while the total system frequency departure of the extensive interconnected electric power system of multizone is reduced to greatest extent.And in intelligence
In Generation Control, Q study is considered as one of most important algorithm.Q study has played very outstanding table in intelligent control
Existing, under the external disturbance interference changed over time, Q study can disturb the strategy of more new system according to different real-time electric powers
Matrix and probability matrix enable to system to have the ability for constantly updating learning strategy.But Q study still remain it is some
The convergence rate of disadvantage, Q study is slow, and when Q learns to update system matrix data, system is in order to select more accurately to move
Work value needs to increase the dimension of the Q matrix and probability matrix in Q study, so that this process, which may cause data, overflows calculating
As a result machine memory causes system that dimension disaster occurs.In addition, Q study does not adapt to Internal system parameters and constantly changes.For
The performance for improving Q study introduces deep neural network on the basis of Q study, and deep neural network can be realized system
On-line tuning and forecast function, it can not only accelerate Q study convergence rate, additionally it is possible to predict subsequent time state, be
System is capable of the state of forecasting system subsequent time in a certain range by the renewal learning to neural network matrix, and according to
Prediction result, which is made, targetedly to be acted, and is able to ascend the active control ability of system in this way, and intelligent elelctrochemical power generation is better achieved
Control.
Current various intelligent controllers can make a good job of work in mission nonlinear working range, including obscure and patrol
Collect controller, artificial neural-network control device, intensified learning controller etc..Wherein, intensified learning controller can effectively increase
Add robustness and dynamic control performance of the Automatic Generation Control under various complex situations.Although intensified learning have it is powerful
Line learns updating ability, but since the control precision that discrete movement may cause system is lower, and calculate overlong time;Because
Intensified learning has the function of that fixed prize and fixed learning rate, the learning efficiency for resulting in system are lower;Meanwhile intensified learning is only
There is a single logical gate for learning, intelligence degree is lower, so introducing artificial feelings on the basis of intensified learning
Sense study is used to optimize the performance of intensified learning.Artificial emotion study can handle continuous variable, and can and logical gate into
Row combines.In recent years, the intelligent controller based on emotion learning has been widely used, it is showed in various practical applications
Brilliant control performance out, nonlinear Control, feed influence generator, power flow control including interconnected electric power system
Device, switched reluctance machines are without sensor speed control, Dynamic Voltage Regulator, through transport power flow controller, asymmetric six phase senses
Induction motor, the real time position control of servo-hydraulic rotary actuator and Unmanned Ground Vehicle navigation etc..Artificial emotion extensive chemical
Habit combines intensified learning and emotion learning, and artificial emotion study is able to solve the tradition intensified learning as caused by discrete movement
Dimension disaster problem, while can be realized more accurate control;Artificial emotion study can be exported with the very short time
Value;Dynamic rewards function and dynamic learning rate can make system more effectively update Q value matrix, to realize convergence faster
Speed and more accurate control strategy;Artificial emotion intensified learning includes three emotion learning functions, including secondary letter
Number, exponential function and linear function, it is highly integrated with action value, learning rate and reward function, therefore, artificial emotion intensified learning
More intelligent control strategy can be generated, Automatic Generation Control device is made more to adapt to the various operation sides of extensive interconnected network
Case.
Game theory is how a kind of study subject makes a kind of mathematical theory conducive to itself decision in the environment;Either
It take group interest as a kind of mathematical theory of main body decision.The Stackelberg betting model proposed in 1934 is extensive
Using it has host-guest architecture, and is adapted with the following smart grid system.In the following power grid, Decentralized Autonomous and concentration
Coordinate as main body power generation system by be the following power grid development trend.The target of the real-time supply and demand interaction of electric system is full
Realize that economic interests maximize on the basis of the sufficient equilibrium of supply and demand, i.e., cost of electricity-generating minimizes.In extensive interconnected network, pass through
Game theory optimizes the extensive Interconnected Generating System of multizone, realizes integrally-regulated optimization.
For development, the application of above-mentioned artificial emotion nitrification enhancement and deep neural network algorithm in the power system
With deficiency, the present invention is devised a kind of algorithm learnt based on artificial emotion intensified learning and deep neural network, utilizes game
By the application in smart grid, a kind of optimization and control algolithm for multi-area Interconnected Power System is proposed, and is named
For the intelligent power generation control method of manual depth's emotion game intensified learning.
Summary of the invention
The present invention provides a kind of intelligent power generation control method of manual depth's emotion game intensified learning.The invention is by people
Work emotion nitrification enhancement and deeply learning algorithm are combined, and in the extensive interconnected electric power system of multizone
Each region using based on mentioned method controller carry out game, optimal control policy is generated in gambling process, realize
To whole regulatory function.
The control purpose of intelligent power generation controller is to reduce frequency deviation f and make a reservation in automatic electricity generation control system
Dominant eigenvalues deviation delta PT.When carrying out electric system Generation Control, which needs in view of control performance standard
Index.Control performance standard index includes Control performance standard 1 (CPS1) and Control performance standard 2 (CPS2), control performance mark
Quasi- index can enhance the frequency supporting dynamics of control area in intelligent power generation controller.
CPS1 refers to target value kCPS1Calculating process be,
Wherein, ACEAVE-minIt is average value of the area control error in 1 minute, and its calculation formula is,
ACE=Δ Ptie-i-10Bi*Δf (2)
Wherein, Δ FAVE-minIt is average value of the frequency departure in 1 minute;BiIt is the frequency bias coefficient of ith zone
(unit is MW/0.1Hz);nTIndicate the number of minutes in the assessment phase;ε1It is the range of target frequencies of CPS1 index.
CPS2 refers to target value kCPS2For the amplitude of restricted area control error, CPS2 requires region control in control area
Average value of the error in 1 minute is less than threshold value (L10), as shown in formula (3),
Wherein,ACEAVE-10-minIt is area control error in 10 minutes
Average value;BsIt is the summation of all control area frequency departures in internet;ε10It is the range of target frequencies of CPS2 index value.
In power industry, CPS1 and CPS2 index is according to daily, data combination North America state monthly and annually is reliable
The rule of the property committee is assessed.In addition, formulating CPS based on CPS1/CPS2 measurement refers to target value kCPS, for assessing CPS mark
Overall Automatic Generation Control performance under quasi-.CPS refers to target value kCPSReach the condition such as formula (4) of standard,
CPS index definition be such as formula (5),
Wherein, Tu、TsAnd TnIt is the unqualified periodicity, total periodicity and unavailable periodicity of CPS standard respectively.
The input of intelligent power generation controller is area control error and CPS1 index;The output of intelligent power generation controller is hair
The power instruction of motor.
Control strategy in manual depth's emotion game intensified learning intelligent power generation control method, mainly passes through extensive chemical
It practises with artificial emotion and completing.
More commonly used algorithm is Q learning algorithm in intensified learning, it is a kind of algorithm for being not based on model, main to use
Action value is provided to environment in intelligent body.Q learning algorithm is mainly made of five elements: Q value matrix Q, probability distribution matrix P,
Reward function R, behavior aggregate A and state set S.Q value matrix Q and moment of probability distribution are used generally, based on the intelligent body of intensified learning
Realization acts a at the current state s of battle array P in the environment, and the two matrixes are then able to update with reward, and function is such as
Formula (6), (7),
Wherein, α is learning rate;γ is discount factor;β is search factor;S and s' respectively indicates current state and lower a period of time
Quarter state;(s, s' a) are the reward functions obtained by selected movement a from current state s to subsequent time state s' to R.
Unlike the intelligent body based on traditional intensified learning, the intelligent body based on artificial emotion study can be from environment
Middle acquisition different conditions create artificial emotion.In order to construct the process of artificial emotion study, invention introduces quantizers
For state θ (input) is transformed to emotion coefficient η (output), such as formula (8), (9) are shown
Wherein, θiRepresent input information;ωiIndicate emotion weight;λiIt is the conversion value of i-th of input information;fnIndicate feelings
Inductance energy;kηIndicate emotion coefficient.
In order to make intensified learning more intelligent, emotion coefficient η must be converted into the actual effect to intensified learning.Cause
This, present invention introduces three transfer functions, i.e. index transfer function, two times transfer function and multinomial transfer function, pass through this
A little transfer functions realize the process, as shown in formula (10),
Wherein ka、kb、kc、kd、keAnd kfIt is conversion coefficient, is usually set to constant.
Based on content shown in above-mentioned formula (11), according to different emotion coefficients can modify intensified learning action value,
Reward function and learning rate.The update mode of action value, reward function and learning rate such as formula (11), (12), shown in (13),
aη←aCf(η) (11)
Rη←RCf(η) (12)
αη←αCf(η) (13)
Wherein, aη、RηAnd αηIt is action value, reward function and the learning rate of manually emotion modification respectively.
Logical gate in manual depth's emotion game intensified learning intelligent power generation control method mainly passes through depth mind
It is completed through e-learning.
Deep neural network is successively trained every layer using greedy strategy in prediction training process.By pre-training
Deep neural network afterwards can carry out on-line training.
Traditional Q study is easy to cause two serious problems: (1) convergence rate is slow;(2) dimension disaster.And depth is refreshing
These problems are able to solve through network.By learning to combine with deep-neural-network Q, both of these problems are available to be had
Effect solves, specifically:
1) in order to reach higher control precision, the motion space of Q study needs discrete to be multiple action elements, this can lead
Cause dimension disaster.In order to solve this problem, deep neural network is introduced into the continuous action for generating output layer.
2) traditional Q indoctrination session is updated and learns to dynamical system under current state, therefore, the convergence of Q study
Speed is slower.Optimal Q value matrix in order to obtain, deep neural network can use the data that system had previously updated and learnt and come
Continuously approach optimal Q value matrix.
The energy of deep neural network system can be described as such as formula (14),
Wherein, WijIt is the i-th row of the jth column of weight matrix, viAnd hjIt is that i-th of visible element and j-th are implicit respectively
Unit;aiAnd bjIt is the offset of i-th of visible element and j-th of implicit unit respectively.
Joint probability distribution calculates such as formula (15),
Wherein,It is normalized function.
The activation probability calculation such as formula (16) of implicit unit,
Wherein, σ (x) is defined as S-shaped activation primitive
The activation probability such as formula (17) of visible element,
The probability such as formula (18) of action is selected,
Wherein, p is the action probability a updated;rpIt is probability coefficent.
In manual depth's emotion game intensified learning intelligent power generation control method that the present invention is mentioned, this method merges simultaneously
Deep neural network, artificial emotion study, intensified learning and theory of games.Each region in multi-area Interconnected Power System
Controller containing one based on this method, these controllers are based on manual depth's emotion nitrification enhancement in interconnection electricity
Game is carried out in Force system, and the optimal movement of system is obtained by game.
The sample of manual depth's emotion enhanced game play study plays an important role in the training process, including
Off-line training process and on-line training process.
During off-line training, since there are different interference in smart grid, it can be obtained by emulation different
Action value and original state.Then different movement and original state are utilized, can predict the different conditions for obtaining subsequent time.
Therefore, the input of input action and original state as deep neural network, the state of subsequent time is as deep neural network
Output, the deep neural network sample during off-line training about the study of manual depth's emotion enhanced game play can be obtained.
In order to improve the convergence rate of each intelligent body, single intelligent body can be trained in advance, then training interconnection together
All intelligent bodies in electric system, i.e. an intelligent body and every other intelligent body carry out game.
Assuming that interconnected electric power system has N number of region, i.e., the controller based on the study of manual depth's emotion enhanced game play has N
It is a.Wherein " step 1 ", " step 2 " ..., " step N+1 " (initializes manual depth's feelings with random action as shown below
Feel the movement of enhanced game play learning controller).
Step 1: region { 2,3 ..., (N+1) } is most preferably acted by individualized training process choosing, and region { 1 } selection is not
Same acts to train the deep neural network in the study of manual depth's emotion enhanced game play;
Step 2: region { 1 } selects most preferably to act by step 1, and region { 3,4 ..., (N+1) } passes through individualized training
Journey selects optimal movement, and region { 2 } select different movements to train the depth in the study of manual depth's emotion enhanced game play
Neural network;
Step i: region { 1,2 ..., (i-1) } selects { step 1, step 2 ..., step (i-1) }, region (i+1),
(i+2) ..., N most preferably acted by individualized training process choosing, region { i } selects different action training manual depth's feelings
Feel the deep neural network in enhanced game play study;
Step N: region { 1,2 ..., (N-1) } selects the optimal movement of { step 1, step 2 ..., step (N-1) }, area
Domain { N } selects different movements to train the deep neural network in the study of manual depth's emotion enhanced game play;
Step (N+1): region { 1,2 ..., N } selects optimal movement by { step 1, step 2 ..., step (N) }.
When all intelligent bodies are in the t times mutual game, all intelligent bodies should all pass through at time { 1,2 ..., t-1 }
Step (N+1) selects optimal movement.
The present invention proposes a kind of intelligent power generation control method of manual depth's emotion game intensified learning, can be by artificial feelings
Sense nitrification enhancement and deeply learning algorithm combine, and can not only improve the continuity of system acting, so that control
Precision is more accurate, additionally it is possible to which the convergence of lifting system accelerates convergence rate;Meanwhile the introducing of deep neural network can
It allows system in the state of forecasting system subsequent time in a certain range, the active control ability of system can be improved in this way, more
Intelligentized Generation Control is realized well;The control based on proposed method is contained in each region in multi-area Interconnected Power System
Device processed, each controller carry out game based on manual depth's emotion nitrification enhancement in interconnected electric power system, rich
During playing chess can, by the way that the data of gambling process are carried out analytical integration, obtain the optimal movement of system, can be realized system
Integrally-regulated effect;The invention can largely improve the overall performance of system, enable the system to that intelligence is better achieved
Elelctrochemical power generation control.
Detailed description of the invention
Fig. 1 is the five regional power system LOAD FREQUENCY Controlling models of the method for the present invention.
Fig. 2 is the implementation procedure of manual depth's emotion game intensified learning intelligent power generation control method of the method for the present invention.
Fig. 3 is the multiple agent game based on manual depth's emotion enhanced game play learning controller of the method for the present invention
Journey.
Specific embodiment
A kind of manual depth's emotion game intensified learning intelligent power generation control method proposed by the present invention is detailed in conjunction with attached drawing
It is described as follows:
Fig. 1 is the five regional power system LOAD FREQUENCY Controlling models of the method for the present invention.The present invention is used with lossless
Five regional power systems of winding thread study the process of intelligent power generation control, wherein each region include generator, it is governor, non-
Reheat steam turbine and load.Control area main purpose with intelligent power generation control is to realize three main targets: 1) will be
System frequency is maintained at its standard value (50Hz);2) interconnection exchange of electric power is maintained at its predetermined value;3) by the hair of each unit
Electrical power is maintained at its economic value.Here, introduce region controls the concept of error to adjust generated output, is used for frequency departure
Zero is all reduced to scheduling dominant eigenvalues deviation.
Fig. 2 is the implementation procedure of manual depth's emotion game intensified learning intelligent power generation control method of the method for the present invention.
The intelligent power generation control method of manual depth's emotion game intensified learning proposed by the present invention, by artificial emotion intensified learning and depth
Degree nitrification enhancement is combined, and is controlled for intelligent power generation.This method as shown in Figure 2 is controlled with frequency departure and region
Error is as input value, using the power instruction of generating set as output.Wherein, emotion part is by artificial emotion intensified learning side
Method is completed, and logical gate completed by deep neural network and theory of games.According to area control error calculate reward function,
Learning rate and update Q value matrix make frequency departure close to ideal value, finally, mutual in multizone by deep neural network prediction
There is the controller based on the learning method in each region in connection electric system, these controllers carry out game in systems,
The movement of system Generation Control is modified by the optimal movement of betting data analysis selection.
Fig. 3 is the multiple agent game based on manual depth's emotion enhanced game play learning controller of the method for the present invention
Journey.Since there are different interference in smart grid, different action value and original state can be obtained by emulation.Then,
Using different action value and original state, the different conditions of subsequent time can be obtained.Accordingly, it is considered to as depth nerve
The state of the action value and original state of the input of network and the subsequent time as deep neural network output, can obtain
The deep neural network sample that manual depth's emotion enhanced game play learns during off-line training.In order to improve each intelligent body
Convergence rate can train single intelligent body in advance, then together train interconnected electric power system in all intelligent bodies, i.e., one
Intelligent body and every other intelligent body carry out game.Assuming that interconnected electric power system has N number of region, that is, it is based on manual depth's emotion
N number of controller of enhanced game play study, multiple intelligent bodies of the controller based on the study of manual depth's emotion enhanced game play are won
Process is played chess as shown in figure 3, wherein " step 1 ", " step 2 " ..., " step (N+1) " is initialize random action artificial
The gambling process of depth emotion enhanced game play learning controller.In Fig. 3, Δ t is the control week in intelligent power generation control framework
Phase.When all intelligent bodies are in the t times mutual game, all intelligent bodies should pass through step (N+ in { 1, the 2 ..., t-1 } time
1) the best movement of selection.
Claims (5)
1. a kind of intelligent power generation control method of manual depth's emotion game intensified learning, which is characterized in that this method merges
Following algorithm:
A: artificial emotion nitrification enhancement;
B: deeply learning algorithm.
2. a kind of intelligent power generation control method of manual depth's emotion game intensified learning as described in claim 1, feature
It is, artificial emotion and intensified learning method are combined by the algorithm A.
3. a kind of intelligent power generation control method of manual depth's emotion game intensified learning as described in claim 1, feature
It is, the algorithm A artificial emotion intensified learning includes three emotion learning functions, including quadratic function, index letter
Several and linear function, it is highly integrated with action value, learning rate and reward function.
4. a kind of intelligent power generation control method of manual depth's emotion game intensified learning as described in claim 1, feature
It is, the algorithm B deeply learning algorithm is based on deep neural network system, using intensified learning as frame.
5. a kind of intelligent power generation control method of manual depth's emotion game intensified learning as described in claim 1, feature
Be, the deep neural network system in the algorithm B deeply learning algorithm using act with original state as input,
Using the state of subsequent time as output.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811408219.XA CN109494766A (en) | 2018-11-23 | 2018-11-23 | A kind of intelligent power generation control method of manual depth's emotion game intensified learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811408219.XA CN109494766A (en) | 2018-11-23 | 2018-11-23 | A kind of intelligent power generation control method of manual depth's emotion game intensified learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109494766A true CN109494766A (en) | 2019-03-19 |
Family
ID=65696555
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811408219.XA Pending CN109494766A (en) | 2018-11-23 | 2018-11-23 | A kind of intelligent power generation control method of manual depth's emotion game intensified learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109494766A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110850720A (en) * | 2019-11-26 | 2020-02-28 | 国网山东省电力公司电力科学研究院 | DQN algorithm-based area automatic power generation dynamic control method |
CN111582311A (en) * | 2020-04-09 | 2020-08-25 | 华南理工大学 | Method for training intelligent agent by using dynamic reward example sample based on reinforcement learning |
CN112186796A (en) * | 2020-09-04 | 2021-01-05 | 三峡大学 | Multi-microgrid non-planned island control method based on artificial emotion reinforcement learning |
CN112398142A (en) * | 2020-10-26 | 2021-02-23 | 广西大学 | Power grid frequency intelligent control method based on empirical mode decomposition |
CN112688345A (en) * | 2020-12-01 | 2021-04-20 | 国网湖北省电力有限公司宜昌供电公司 | Intelligent frequency control method considering multi-dimensional frequency control performance standard |
CN113629768A (en) * | 2021-08-16 | 2021-11-09 | 广西大学 | Difference variable parameter vector emotion depth reinforcement learning power generation control method |
-
2018
- 2018-11-23 CN CN201811408219.XA patent/CN109494766A/en active Pending
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110850720A (en) * | 2019-11-26 | 2020-02-28 | 国网山东省电力公司电力科学研究院 | DQN algorithm-based area automatic power generation dynamic control method |
CN111582311A (en) * | 2020-04-09 | 2020-08-25 | 华南理工大学 | Method for training intelligent agent by using dynamic reward example sample based on reinforcement learning |
CN111582311B (en) * | 2020-04-09 | 2023-03-28 | 华南理工大学 | Method for training intelligent agent by using dynamic reward example sample based on reinforcement learning |
CN112186796A (en) * | 2020-09-04 | 2021-01-05 | 三峡大学 | Multi-microgrid non-planned island control method based on artificial emotion reinforcement learning |
CN112398142A (en) * | 2020-10-26 | 2021-02-23 | 广西大学 | Power grid frequency intelligent control method based on empirical mode decomposition |
CN112688345A (en) * | 2020-12-01 | 2021-04-20 | 国网湖北省电力有限公司宜昌供电公司 | Intelligent frequency control method considering multi-dimensional frequency control performance standard |
CN112688345B (en) * | 2020-12-01 | 2022-06-03 | 国网湖北省电力有限公司宜昌供电公司 | Intelligent frequency control method considering multi-dimensional frequency control performance standard |
CN113629768A (en) * | 2021-08-16 | 2021-11-09 | 广西大学 | Difference variable parameter vector emotion depth reinforcement learning power generation control method |
CN113629768B (en) * | 2021-08-16 | 2023-06-20 | 广西大学 | Differential evolution variable parameter vector emotion deep reinforcement learning power generation control method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109494766A (en) | A kind of intelligent power generation control method of manual depth's emotion game intensified learning | |
Xi et al. | Smart generation control based on multi-agent reinforcement learning with the idea of the time tunnel | |
CN110535146B (en) | Electric power system reactive power optimization method based on depth determination strategy gradient reinforcement learning | |
CN107437813B (en) | Power distribution network reactive power optimization method based on cuckoo-particle swarm | |
Liu et al. | Comparison of very short-term load forecasting techniques | |
CN102479339B (en) | Method and system for forecasting short-term wind speed of wind farm based on hybrid neural network | |
CN108416366A (en) | A kind of power-system short-term load forecasting method of the weighting LS-SVM based on Meteorological Index | |
CN114217524B (en) | Power grid real-time self-adaptive decision-making method based on deep reinforcement learning | |
CN109217306A (en) | A kind of intelligent power generation control method based on the deeply study with movement from optimizing ability | |
CN113363997A (en) | Reactive voltage control method based on multi-time scale and multi-agent deep reinforcement learning | |
CN104636985A (en) | Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network | |
Lian | Reactive power optimization based on adaptive multi-objective optimization artificial immune algorithm | |
CN110866640A (en) | Power load prediction method based on deep neural network | |
CN110429652A (en) | A kind of intelligent power generation control method for expanding the adaptive Dynamic Programming of deep width | |
Rizwan et al. | Artificial intelligence based approach for short term load forecasting for selected feeders at madina saudi arabia | |
CN106532691A (en) | Adaptive dynamic programming-based frequency compound control method of single-region power system | |
CN115313403A (en) | Real-time voltage regulation and control method based on deep reinforcement learning algorithm | |
Xu et al. | Competitive search algorithm: a new method for stochastic optimization | |
CN115795992A (en) | Park energy Internet online scheduling method based on virtual deduction of operation situation | |
Li et al. | Data‐driven cooperative load frequency control method for microgrids using effective exploration‐distributed multi‐agent deep reinforcement learning | |
CN111799820B (en) | Double-layer intelligent hybrid zero-star cloud energy storage countermeasure regulation and control method for power system | |
CN117172097A (en) | Power distribution network dispatching operation method based on cloud edge cooperation and multi-agent deep learning | |
CN116722541A (en) | Power system load prediction method and device based on convolutional neural network | |
Li et al. | Multiagent deep meta reinforcement learning for sea computing-based energy management of interconnected grids considering renewable energy sources in sustainable cities | |
CN108428021A (en) | Micro-capacitance sensor Short-term Load Forecasting Model based on HSA-RRNN |
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: 20190319 |
|
RJ01 | Rejection of invention patent application after publication |