CN109784692A - A kind of fast and safely constraint economic load dispatching method based on deep learning - Google Patents
A kind of fast and safely constraint economic load dispatching method based on deep learning Download PDFInfo
- Publication number
- CN109784692A CN109784692A CN201811631297.6A CN201811631297A CN109784692A CN 109784692 A CN109784692 A CN 109784692A CN 201811631297 A CN201811631297 A CN 201811631297A CN 109784692 A CN109784692 A CN 109784692A
- Authority
- CN
- China
- Prior art keywords
- noise reduction
- constraint
- economic dispatch
- deep learning
- security constrained
- 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.)
- Granted
Links
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a kind of fast and safely constraint economic load dispatching method based on deep learning, key step are as follows: 1) determine the deep neural network for being suitable for security constrained economic dispatch model.2) storehouse noise reduction autocoder SDAE is trained.3) the security constrained economic dispatch model based on deep learning is established.4) k=1 is enabled, by Operation of Electric Systems condition entry into deep neural network, obtains the action constaint set J of security constrained economic dispatch model(1).5) by constraint set J(1).It is input in security constrained economic dispatch model, obtains security constrained economic dispatch scheme.6) N-1 inspection is carried out to security constrained economic dispatch scheme, if there is new constraint J(new), then k=k+1 is enabled, constraint set is updated to J(k)=J(k)∪J(new), and return step 5.If without new constraint, output safety constrains economic dispatch program.It the composite can be widely applied to the security constrained economic dispatch analysis of electric system various industries.
Description
Technical field
The present invention relates to electric power system and its automation field, specifically a kind of fast and safely constraint based on deep learning
Economic load dispatching method.
Background technique
Security constrained economic dispatch is the important tool of power grid security economical operation.But consider the route constraint of N-1 failure
The scale that security constrained economic dispatch model can be greatly increased becomes a large-scale mathematical programming problem.This will
Scaling concern is brought to the solution of security constrained economic dispatch, is difficult to solve using business LP solver.However, root
According to only having sub-fraction to be active constraint in all constraints of engineering experience, other is all redundant constaint.
Therefore, industry generally carries out unconfined economic load dispatching optimization problem and is solved at present, then iteratively increases
It is tied in model and is solved, this method is generally possible to obtain last optimal solution and does not need too many number of iterations.In addition,
In external certain electric structures, some most possible active constraints are often also added in advance by model according to artificial experience
In.
And has a large amount of scholars in academia at present and propose that scene screening technique accelerates safety about to reduce redundant constaint
The solution of beam economic load dispatching, such method mainly passes through the multiple small-scale optimization problems of solution or another complexity of foundation solution is convex
Plan optimization problem reduces security constraint number, and model solution precision and redundant constaint delete the balance between number still
There are also the spaces of promotion.
Summary of the invention
Present invention aim to address problems of the prior art.
To realize the present invention purpose and the technical solution adopted is that such, it is a kind of based on deep learning fast and safely about
Beam economic load dispatching method, mainly comprises the steps that
1) deep neural network for being suitable for security constrained economic dispatch model, i.e. storehouse noise reduction autocoder are determined
SDAE。
Determine that the key step for the deep neural network for being suitable for security constrained economic dispatch is as follows:
1.1) storehouse noise reduction autocoder SDAE is established.The storehouse noise reduction autocoder SDAE by n noise reduction from
The dynamic layer-by-layer storehouse of encoder DAE forms.
Wherein, the input layer of first of noise reduction autocoder DAE is denoted as Yl-1, middle layer is denoted as Yl, output layer is denoted as Zl。
Middle layer YlIt is as follows:
In formula,Presentation code function.R is activation primitive.θ is coding parameter. WlIt is compiled automatically for first of noise reduction
The weight of the coding function of code device DAE.blFor the biasing of the coding function of first of noise reduction autocoder DAE.
Wherein, activation primitive R is as follows:
In formula, x is the input of neuron.
Output layer ZlIt is as follows:
In formula,Indicate decoding functions.θ ' is decoding parametric.Wl' be first of noise reduction autocoder DAE solution
The weight of code function.bl' for first of noise reduction autocoder DAE decoding functions biasing.
1.2) Operation of Electric Systems condition is obtained, Operation of Electric Systems condition includes load power PDWith system topological knot
Structure.
1.3) formula 4 is utilized, sample is pre-processed.
Wherein, vmeanAnd vstdIt is the mean value and standard deviation of vector V respectively.V is the data for needing to be normalized,
Including outputting and inputting for training sample, test samples and test sample.
1.4) by pretreated load power PDIt inputs in vertical storehouse noise reduction autocoder SDAE, to export hair
Motor power output PG。
Generator output PGIt is as follows:
In formula,For the coding function of n-th of noise reduction autocoder DAE.
2) the deep learning strategy for being suitable for security constrained economic dispatch is determined, thus to storehouse noise reduction autocoder
SDAE is trained.
The key step being trained to storehouse noise reduction autocoder SDAE is as follows:
2.1) unsupervised pre-training is carried out to storehouse noise reduction autocoder SDAE, selects one group of coding parameter θ and decoding
Parameter θ ', so that calculating parameter M is reached minimum.
Calculating parameter M is as follows:
2.2) supervision fine tuning, i.e. selection coding parameter θ have been carried out to storehouse noise reduction autocoder SDAE, has made calculating parameter
L reaches minimum.
Calculating parameter L is as follows:
2.3) coding parameter θ is updated using RMSprop learning algorithm, it may be assumed that
Wherein,It is objective function O to variable θ0In tthUpdate.⊙ is Hadamard multiplier.ρ is gradient
Index of bunching.Update for objective function O to variable θ.For variable θoIn the t-1 times iteration.R is gradient.rtFor
Preceding tthThe gradient of secondary iteration accumulation.rt-1For the gradient of preceding t-1 iteration accumulation.ε is constant.
3) based on the deep neural network after training, the security constrained economic dispatch model based on deep learning is established.
Security constrained economic dispatch model based on deep learning is as follows:
H1And H2For coefficient matrix.
The constraint of security constrained economic dispatch model based on deep learning is respectively as shown in formula 10 to formula 12.
eGPG=eDPD。 (10)
In formula, eGAnd eDExpression is all 1 vector.
In formula,For the branch power of the branch ij in c articles of line fault, wherein c=0 indicates no route event
Barrier.For the branch power lower limit of the branch ij in c articles of line fault.For the branch ij in c articles of line fault
The branch power upper limit.
PG∈χG (12)
In formula, χGFor generator output set.
4) k=1 is enabled, by Operation of Electric Systems condition entry into deep neural network, obtains the peace based on deep learning
The action constaint set J of staff cultivation economic load dispatching model(1)。
5) by constraint set J(1).It is input in the security constrained economic dispatch model based on deep learning, obtains security constraint
Economic dispatch program.
6) N-1 inspection is carried out to security constrained economic dispatch scheme, if there is new constraint J(new), then k=k+1, constraint set are enabled
It is updated to J(k)=J(k)∪J(new), and return step 5.If without new constraint, output safety constrains economic dispatch program.
The solution have the advantages that unquestionable.The present invention is based on history datas, automatic using storehouse noise reduction
Encoder (SDAE) extracts this non-linear relation, and new system operation conditions are carried out with " prediction " of action constaint set.
Depth learning technology is embedded into existing Scheduling Framework by the present invention, and computation burden is transferred to off-line training, online can be straight
It connects by all action constaint sets of system operation conditions efficient mapping, improves security constrained economic dispatch solution efficiency.
Active constraint discrimination method proposed by the present invention based on SDAE deep neural network, can rapidly and accurately by
System operation conditions obtain the action constaint set of security constrained economic dispatch model, embody SDAE model and run item to system
The powerful approximation capability of non-linear relation between part and generator output.It is proposed by the present invention based on deep learning fast and safely about
Beam economic load dispatching method, the insert depth learning art in the case where tradition optimizes Computational frame, mentioned method are not influencing computational accuracy
In constringent situation, calculating speed can be effectively improved.To solve the problems, such as that security constrained economic dispatch provides one kind
New Century Planned Textbook.
It the composite can be widely applied to the security constrained economic dispatch analysis of electric system various industries.
Detailed description of the invention
Fig. 1 is the security constrained economic dispatch model based on deep learning;
Fig. 2 is deep neural network DNN structure chart.
Specific embodiment
Below with reference to embodiment, the invention will be further described, but should not be construed the above-mentioned subject area of the present invention only
It is limited to following embodiments.Without departing from the idea case in the present invention described above, according to ordinary skill knowledge and used
With means, various replacements and change are made, should all include within the scope of the present invention.
Embodiment 1:
Referring to Fig. 1 and Fig. 2, a kind of fast and safely constraint economic load dispatching method based on deep learning mainly includes following
Step:
1) deep neural network (DNN) for being suitable for security constrained economic dispatch model is determined, i.e. storehouse noise reduction is compiled automatically
Code device SDAE.
Determine that the key step for the deep neural network for being suitable for security constrained economic dispatch is as follows:
1.1) storehouse noise reduction autocoder SDAE (Stacked Denoising Autoencoders) is established.The heap
Stack noise reduction autocoder SDAE is formed by the n layer-by-layer storehouse of noise reduction autocoder DAE.
Wherein, the input layer of first of noise reduction autocoder DAE is denoted as Yl-1, middle layer is denoted as Yl, output layer is denoted as Zl。
Middle layer YlIt is as follows:
In formula,Presentation code function.R is activation primitive.θ is coding parameter, θ={ W, b }.WlIt is first
The weight of the coding function of noise reduction autocoder DAE.blFor the l noise reduction autocoder DAE coding function it is inclined
It sets.W is the weight of coding function.B is the biasing of coding function.
Wherein, activation primitive R is as follows:
In formula, x indicates the input of the neuron of the input layer of noise reduction autocoder DAE, middle layer or output layer.
As shown by the equation, ReLU is activated the piecewise linearity form of ReLU (Rectified Linear Unit) activation primitive
Function can effectively avoid gradient extinction tests.Therefore, ReLU function is selected as the activation primitive of SDAE in the present invention.
The output Z of output layerlIt is as follows:
In formula,Indicate decoding functions.θ ' is decoding parametric.θ '={ W', b'}.Wl' compiled automatically for first of noise reduction
The weight of the decoding functions of code device DAE.bl' for first of noise reduction autocoder DAE decoding functions biasing.W' is decoding
The weight of function.B' is the biasing of decoding functions.
1.2) Operation of Electric Systems condition, including load power P are obtainedD, system topology etc..
1.3) formula 4 is utilized, sample is pre-processed.
Wherein, vmeanAnd vstdIt is the mean value and standard deviation of vector V respectively.V is the data for needing to be normalized,
Including outputting and inputting for training sample, test samples and test sample.In the present embodiment, training sample is to drop for storehouse
The load power of the autocoder SDAE that makes an uproar training, test samples are negative to examine for storehouse noise reduction autocoder SDAE
Lotus power, test sample are the load power tested for storehouse noise reduction autocoder SDAE.
1.4) by pretreated load power PDIt inputs in vertical storehouse noise reduction autocoder SDAE, to export hair
Motor power output PG。
Generator output PGIt is as follows:
In formula,For the coding function of n-th of noise reduction autocoder DAE.For the 1st noise reduction autocoder
The coding function of DAE.
2) the deep learning strategy for being suitable for security constrained economic dispatch is determined, thus to storehouse noise reduction autocoder
SDAE is trained.The target of deep neural network study is non-thread between data to extract in order to obtain optimum code parameter θ
Property feature.
The key step being trained to storehouse noise reduction autocoder SDAE is as follows:
2.1) unsupervised pre-training is carried out to storehouse noise reduction autocoder SDAE, selects one group of coding parameter θ and decoding
Parameter θ ', so that calculating parameter M is reached minimum.
Calculating parameter M is as follows:
2.2) supervision fine tuning, i.e. selection coding parameter θ have been carried out to storehouse noise reduction autocoder SDAE, has made calculating parameter
L reaches minimum.
Calculating parameter L is as follows:
2.3) coding parameter θ is updated using RMSprop learning algorithm, it may be assumed that
Wherein,It is objective function O to variable θ0In tthUpdate.⊙ is Hadamard multiplier.η is study
Rate.ρ is gradient index of bunching.Update for objective function O to variable θ.For variable θoIn the t-1 times iteration.r
For gradient.rtFor preceding tthThe gradient of secondary iteration accumulation.rt-1For the gradient of preceding t-1 iteration accumulation.ε is constant.In this implementation
In example, ε=10-8。
The present embodiment is used as deep learning algorithm using RMSprop (root mean square propagation).
Training sample is divided into multiple batches and learnt by RMSprop learning algorithm.Each batch is successively trained and updates depth
Parameter in neural network.In addition, RMSprop learning algorithm makes each parameter by the rolling average using gradient square
Corresponding learning rate is different.The algorithm can reduce trained pressure and avoid Local Minimum.
3) based on the deep neural network after training, the security constrained economic dispatch model based on deep learning is established.
Security constrained economic dispatch model based on deep learning is as follows:
H1And H2For coefficient matrix.Coefficient matrix is mainly by constraining J(k)It determines.
The constraint of security constrained economic dispatch model based on deep learning is respectively as shown in formula 10 to formula 12.
eGPG=eDPD。 (10)
In formula, eGAnd eDExpression is all 1 vector, i.e. unit vector.
In formula,For the branch power of the branch ij in c articles of line fault, wherein c=0 indicates no route event
Barrier.J(k)Indicate constraint set.For the branch power lower limit of the branch ij in c articles of line fault.For in the c bars line
The branch power upper limit of branch ij when the failure of road.
PG∈χG (12)
In formula, χGFor generator output set.
4) k=1 is enabled, by Operation of Electric Systems condition entry into deep neural network, obtains the peace based on deep learning
The action constaint set J of staff cultivation economic load dispatching model(1)。
Service condition includes load power and topological structure etc..The present embodiment only represents service condition with load power.
5) by constraint set J(1).It is input in the security constrained economic dispatch model based on deep learning, obtains security constraint
Economic dispatch program meets constraint set J(k)'s
6) N-1 inspection is carried out to security constrained economic dispatch scheme, if there is new constraint J(new), then k=k+1, constraint set are enabled
It is updated to J(k)=J(k)∪J(new), and return step 5.If without new constraint, output safety constrains economic dispatch program.
It is to determine a kind of criterion of power system security, also known as single failure safety criterion that N-1, which examines principle,.According to this
One criterion, any independent component (generator, transmission line of electricity, transformer etc.) in N number of element of electric system break down and
After being removed, it should not cause to cause user to have a power failure because All other routes overload is tripped, not destroy the stability of system, do not occur
The accidents such as collapse of voltage.
Embodiment 2:
A kind of experiment that fast and safely constrains economic load dispatching method of the verifying based on deep learning, mainly includes following step
It is rapid:
1) training sample is obtained and is pre-processed
It is emulated in the present embodiment using IEEE-118-Washington system.Normal state is pressed to load active-power P D
Distribution is sampled to represent different service conditions, wherein load mean value is the value in IEEE118 modular system, and standard deviation is
The 10% of mean value.By large-scale optimizatoin device --- Gurobi solver is to the corresponding security constrained economic dispatch of each sample mode
Model is solved to obtain optimal power generation machine power output PG.Then, using the active power of the load bus of all sample modes as
Training sample inputs X, and generator output exports Y as training sample.Finally, being located in advance by formula (4) to training sample
Reason.
(2) it is initialized suitable for the deep neural network of security constrained economic dispatch
According to training sample capacity, training sample is divided into 100 batches;According to electric system scale to be solved and again
Miscellaneous degree, set the number of plies l of SDAE optimal load flow model as 6, every layer of neuron number be respectively 118,200,200,200,
200,19, other hyper parameters are as shown in table 1.
1 deep neural network hyper parameter of table
(3) it is suitable for the unsupervised pre-training of deep neural network of security constrained economic dispatch
Firstly, inputting P using training sampleD, according to formula (6), construct the objective function M (Z of first layer DAE training1,
PD);Then according to formula (2), formula (8), the optimized parameter W of RMSProp learning algorithm iterative solution first layer DAE is used1、
b1、W′1、 b′1.Later, it is exported by the middle layer that formula (1) and formula (3) obtain first layer DAE, as the defeated of second layer DAE
Enter, constructs the loss function M (Z of second layer DAE in the same way2,Z1), and undated parameter in the same way, with such
It pushes away, optimum code parameter θ={ W, the b } of every layer of DAE is successively solved the bottom of to top, and using the optimum code parameter as the next stage
The initial value for thering is supervision to finely tune.
(4) supervision finely tunes suitable for the deep neural network of security constrained economic dispatch
Firstly, inputting P using training sampleDWith output PG, according to formula (6), construct the mesh of SDAE Training process
Scalar functions L (PG,PD);Then still according to formula (2), (8), all optimal volumes of SDAE are iteratively solved using RMSProp learning algorithm
Code parameter θ={ W, b };So far, the deep neural network training suitable for security constrained economic dispatch is completed.
(5) security constrained economic dispatch quickly calculates
By in the SDAE optimal models of training completion, which can directly be reflected by formula (5) in test sample input step (4)
All test samples are projected in the optimal power generation power output for considering N-1 safety;Carrying out N-1 according to generator output examines determination to act as
Use constraint set;Action constaint set is added in security constrained economic dispatch model, the progress as shown in formula (9)-(12) is excellent
Change and calculates;Optimum results being substituted into N-1 inspection again and checked whether newly-increased constraint, terminated if calculated without if, otherwise
These new constraints, which will be added into security constrained economic dispatch model, re-starts optimization.
Specific simulation result is as follows:
I) security constrained economic dispatch calculates control methods
Security constrained economic dispatch control methods includes M0-M1 in emulation:
M0: industry method.
M1: the fast and safely constraint economic load dispatching method based on deep learning.
The simulation result of the present embodiment all Intel (R) Core (TM) i5-7200U CPU@2.50GHz 2.71GHz,
It is tested under the hardware environment of 16GB RAM.
II) deep learning active constraint identification precision is analyzed
The present embodiment is randomly selected to verify the overall accuracy that deep neural network recognizes active constraint by example
2000 test samples, after tested, a total of 64 in 2000 samples, 356,000 constraints, wherein action constaint set is
112,498, the judging nicety rate whether the mentioned method constraint set of the present invention works reaches 96.6%.As it can be seen that SDAE relies on it
Deep layer stack architecture and coding and decoding process have efficiently extracted the non-linear spy between system operation conditions and optimal power generation power output
Sign, realizes the high-precision fast mapping by system operation conditions to action constaint set.Therefore, depth mind proposed by the present invention
There is degree of precision through the pre- discrimination method of network active constraint.
III) probability optimal load flow on-line Algorithm calculated performance is analyzed
Table 2 lists M0 and M1 method calculates the cycle-index of security constrained economic dispatch, due to deep neural network pair
The high-precision identification of active constraint, the method for the present invention 95.3% disposable convergence in 2000 test samples.Table 3
List the time of the calculating by M1 method under certain scene that M0 solves four loop iterations of security constrained economic dispatch problem.
Seen from table 3, security constrained economic dispatch problem time-consuming is calculated 13.2 seconds by M0 method, and M1 method only needs 4.0 seconds, calculating speed
It is 3.3 times of M0 method.As it can be seen that calculating speed can be effectively improved by calculating security constrained economic dispatch problem using the method for the present invention
Degree, and its computational accuracy and convergence are unaffected.It is otherwise noted that example of the present invention considers single moment, such as
Scene of the fruit for 24 moment combined dispatchings of large-scale real system and consideration, the fast and safely constraint based on deep learning
The advantage of economic load dispatching method method will be apparent from.
The loop iteration number of 2 M0-M1 of table calculating security constrained economic dispatch
Table 3M0-M1 calculates the time comparison of security constrained economic dispatch
Method | M0 | M1 |
Time (second) | 13.2 | 4.0 |
From the experimental results: the fast and safely constraint economic load dispatching method proposed by the invention based on deep learning,
It can be with the complex nonlinear relationship between very high precision approximation system service condition and optimal Generator power output, to realize
Efficient identifications of the system operation conditions to active constraint have the characteristics that precision speed of not suffering a loss is effectively promoted.In addition,
For large scale system and multi-period scheduling problem, applicability of the invention be will be more significant.
It, can be according in conclusion the invention proposes a kind of fast and safely constraint economic load dispatching method based on deep learning
Existing computing resource and history data are held in the palm, the calculating pressure of security constrained economic dispatch problem is transferred to linear instruction
Practice.By the way that deep neural network to be embedded in the Computational frame of existing security constrained economic dispatch problem, meter can not be lost
Precision and convergence are calculated, the number of iterations is simultaneously effective reduced, improves computational efficiency.It is analyzed by Simulation Example, demonstrates this
Invent the validity of proposed method.Therefore, the effectively optimizing that the present invention can constrain economic load dispatching for power system security provides skill
Art support.
Claims (4)
1. a kind of fast and safely constraint economic load dispatching method based on deep learning, which is characterized in that mainly comprise the steps that
1) deep neural network for being suitable for security constrained economic dispatch model, i.e., the described storehouse noise reduction autocoder are determined
SDAE;
2) determine be suitable for security constrained economic dispatch deep learning strategy, thus to storehouse noise reduction autocoder SDAE into
Row training.
3) based on the deep neural network after training, the security constrained economic dispatch model based on deep learning is established;
4) k=1 is enabled, by Operation of Electric Systems condition entry into deep neural network, obtains the safety based on deep learning about
The action constaint set J of beam economic load dispatching model(1);
5) by constraint set J(1)It is input in the security constrained economic dispatch model based on deep learning, obtains security constraint economy
Scheduling scheme;
6) N-1 inspection is carried out to security constrained economic dispatch scheme, if there is new constraint J(new), then k=k+1 is enabled, constraint set updates
For J(k)=J(k)∪J(new), and return step 5;If without new constraint, output safety constrains economic dispatch program.
2. a kind of fast and safely constraint economic load dispatching method based on deep learning according to claim 1, feature exist
In: determine that the key step for the deep neural network for being suitable for security constrained economic dispatch is as follows:
1) storehouse noise reduction autocoder SDAE is established;The storehouse noise reduction autocoder SDAE is by n noise reduction autocoding
The layer-by-layer storehouse of device DAE forms;
Wherein, the input layer of first of noise reduction autocoder DAE is denoted as Yl-1, middle layer is denoted as Yl, output layer is denoted as Zl;
Middle layer YlIt is as follows:
In formula,Presentation code function;R is activation primitive;θ is coding parameter;WlFor first of noise reduction autocoder
The weight of the coding function of DAE;blFor the biasing of the coding function of first of noise reduction autocoder DAE;
Wherein, activation primitive R is as follows:
In formula, x is the input of neuron;
Output layer ZlIt is as follows:
In formula,Indicate decoding functions;θ ' is decoding parametric;Wl' be first of noise reduction autocoder DAE decoding functions
Weight;bl' for first of noise reduction autocoder DAE decoding functions biasing;
2) Operation of Electric Systems condition is obtained;The Operation of Electric Systems condition includes load power PDWith electric system topology knot
Structure;
3) formula 4 is utilized, sample is pre-processed;
Wherein, vmeanAnd vstdIt is the mean value and standard deviation of vector V respectively, V is the data for needing to be normalized, including instruction
Practice outputting and inputting for sample, test samples and test sample;
4) by pretreated load power PDIt inputs in vertical storehouse noise reduction autocoder SDAE, so that output generator is contributed
PG;
Generator output PGIt is as follows:
In formula,For the coding function of n-th of noise reduction autocoder DAE.
3. a kind of fast and safely constraint economic load dispatching method based on deep learning according to claim 1 or 2, feature
It is, the key step being trained to storehouse noise reduction autocoder SDAE is as follows:
1) unsupervised pre-training is carried out to storehouse noise reduction autocoder SDAE, select one group of coding parameter θ and decoding parameter θ ',
Calculating parameter M is set to reach minimum;
Calculating parameter M is as follows:
2) supervision fine tuning, i.e. selection coding parameter θ have been carried out to storehouse noise reduction autocoder SDAE, has reached calculating parameter L
It is minimum;
Calculating parameter L is as follows:
3) coding parameter θ is updated using RMSprop learning algorithm, it may be assumed that
Wherein,It is objective function O to variable θoIn tthIteration update;⊙ is Hadamard multiplier;η is learning rate;
ρ is gradient index of bunching;It is objective function O to variableθUpdate;For variable θoIn the t-1 times iteration;R is ladder
Degree;rtFor preceding tthThe gradient of secondary iteration accumulation;rt-1For the gradient of preceding t-1 iteration accumulation;ε is constant.
4. a kind of fast and safely constraint economic load dispatching method based on deep learning according to claim 1 or 2, feature
It is, the security constrained economic dispatch model based on deep learning is as follows:
H1And H2For coefficient matrix;
The constraint of security constrained economic dispatch model based on deep learning is respectively as shown in formula 10 to formula 12;
eGPG=eDPD; (10)
In formula, eGAnd eDExpression is all 1 vector;
In formula,For the branch power of the branch ij in c articles of line fault, wherein c=0 indicates no line fault;
For the branch power lower limit of the branch ij in c articles of line fault;For the branch of the branch ij in c articles of line fault
The upper limit of the power;
PG∈χG (12)
In formula, χGFor generator output set.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811631297.6A CN109784692B (en) | 2018-12-29 | 2018-12-29 | Rapid safety constraint economic dispatching method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811631297.6A CN109784692B (en) | 2018-12-29 | 2018-12-29 | Rapid safety constraint economic dispatching method based on deep learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109784692A true CN109784692A (en) | 2019-05-21 |
CN109784692B CN109784692B (en) | 2020-11-24 |
Family
ID=66497987
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811631297.6A Active CN109784692B (en) | 2018-12-29 | 2018-12-29 | Rapid safety constraint economic dispatching method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109784692B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110556823A (en) * | 2019-08-15 | 2019-12-10 | 中国南方电网有限责任公司 | Rapid calculation method and system for safety constraint unit combination based on model dimension reduction |
CN110929989A (en) * | 2019-10-29 | 2020-03-27 | 重庆大学 | N-1 safety checking method with uncertainty based on deep learning |
CN110991731A (en) * | 2019-11-28 | 2020-04-10 | 中国南方电网有限责任公司 | Electric power real-time market constraint self-identification clearing method and system based on deep learning |
CN110991741A (en) * | 2019-12-02 | 2020-04-10 | 中国南方电网有限责任公司 | Section constraint probability early warning method and system based on deep learning |
CN111242800A (en) * | 2019-12-20 | 2020-06-05 | 重庆大学 | Electric power market clearing system suitable for capacity constraint is out of limit |
CN111612213A (en) * | 2020-04-10 | 2020-09-01 | 中国南方电网有限责任公司 | Section constraint intelligent early warning method and system based on deep learning |
CN112926504A (en) * | 2021-03-23 | 2021-06-08 | 重庆商务职业学院 | Acoustic emission signal denoising method based on noise reduction self-encoder |
CN113761788A (en) * | 2021-07-19 | 2021-12-07 | 清华大学 | SCOPF rapid calculation method and device based on deep learning |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130060397A1 (en) * | 2011-09-07 | 2013-03-07 | General Electric Company | Management of power distribution constraints |
US20150134132A1 (en) * | 2013-11-14 | 2015-05-14 | Abb Technology Ag | Method and Apparatus for Security Constrained Economic Dispatch in Hybrid Power Systems |
CN107491867A (en) * | 2017-08-07 | 2017-12-19 | 国电南瑞科技股份有限公司 | It is a kind of for the multicycle send out defeated change repair schedule Security Checking and appraisal procedure |
CN108054757A (en) * | 2017-12-22 | 2018-05-18 | 清华大学 | A kind of embedded idle and voltage N-1 Close loop security check methods |
CN108304623A (en) * | 2018-01-15 | 2018-07-20 | 重庆大学 | A kind of Probabilistic Load Flow on-line calculation method based on storehouse noise reduction autocoder |
WO2018180386A1 (en) * | 2017-03-30 | 2018-10-04 | 国立研究開発法人産業技術総合研究所 | Ultrasound imaging diagnosis assistance method and system |
-
2018
- 2018-12-29 CN CN201811631297.6A patent/CN109784692B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130060397A1 (en) * | 2011-09-07 | 2013-03-07 | General Electric Company | Management of power distribution constraints |
US20150134132A1 (en) * | 2013-11-14 | 2015-05-14 | Abb Technology Ag | Method and Apparatus for Security Constrained Economic Dispatch in Hybrid Power Systems |
WO2018180386A1 (en) * | 2017-03-30 | 2018-10-04 | 国立研究開発法人産業技術総合研究所 | Ultrasound imaging diagnosis assistance method and system |
CN107491867A (en) * | 2017-08-07 | 2017-12-19 | 国电南瑞科技股份有限公司 | It is a kind of for the multicycle send out defeated change repair schedule Security Checking and appraisal procedure |
CN108054757A (en) * | 2017-12-22 | 2018-05-18 | 清华大学 | A kind of embedded idle and voltage N-1 Close loop security check methods |
CN108304623A (en) * | 2018-01-15 | 2018-07-20 | 重庆大学 | A kind of Probabilistic Load Flow on-line calculation method based on storehouse noise reduction autocoder |
Non-Patent Citations (2)
Title |
---|
余娟: "含电转气的电_气互联系统可靠性评估", 《中国电机工程学报》 * |
白立鹏: "考虑含有新能源的电力系统安全约束经济调度", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110556823A (en) * | 2019-08-15 | 2019-12-10 | 中国南方电网有限责任公司 | Rapid calculation method and system for safety constraint unit combination based on model dimension reduction |
CN110556823B (en) * | 2019-08-15 | 2023-11-14 | 中国南方电网有限责任公司 | Model dimension reduction based safety constraint unit combination rapid calculation method and system |
CN110929989A (en) * | 2019-10-29 | 2020-03-27 | 重庆大学 | N-1 safety checking method with uncertainty based on deep learning |
CN110991731A (en) * | 2019-11-28 | 2020-04-10 | 中国南方电网有限责任公司 | Electric power real-time market constraint self-identification clearing method and system based on deep learning |
CN110991741A (en) * | 2019-12-02 | 2020-04-10 | 中国南方电网有限责任公司 | Section constraint probability early warning method and system based on deep learning |
CN110991741B (en) * | 2019-12-02 | 2022-07-12 | 中国南方电网有限责任公司 | Section constraint probability early warning method and system based on deep learning |
CN111242800A (en) * | 2019-12-20 | 2020-06-05 | 重庆大学 | Electric power market clearing system suitable for capacity constraint is out of limit |
CN111242800B (en) * | 2019-12-20 | 2024-04-19 | 重庆大学 | Electric power market clear system suitable for capacity constraint out-of-limit |
CN111612213A (en) * | 2020-04-10 | 2020-09-01 | 中国南方电网有限责任公司 | Section constraint intelligent early warning method and system based on deep learning |
CN111612213B (en) * | 2020-04-10 | 2023-10-10 | 中国南方电网有限责任公司 | Section constraint intelligent early warning method and system based on deep learning |
CN112926504A (en) * | 2021-03-23 | 2021-06-08 | 重庆商务职业学院 | Acoustic emission signal denoising method based on noise reduction self-encoder |
CN113761788A (en) * | 2021-07-19 | 2021-12-07 | 清华大学 | SCOPF rapid calculation method and device based on deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN109784692B (en) | 2020-11-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109784692A (en) | A kind of fast and safely constraint economic load dispatching method based on deep learning | |
CN111179121B (en) | Power grid emergency control method based on expert system and deep reverse reinforcement learning | |
WO2021175058A1 (en) | Neural network architecture search method and apparatus, device and medium | |
Dehghani et al. | Adaptive network reliability analysis: Methodology and applications to power grid | |
CN103488869A (en) | Wind power generation short-term load forecast method of least squares support vector machine | |
CN106786560A (en) | A kind of power system stability characteristic automatic extraction method and device | |
CN106972504A (en) | Interval idle work optimization method based on genetic algorithm | |
CN110417011A (en) | A kind of online dynamic secure estimation method based on mutual information Yu iteration random forest | |
CN116245033B (en) | Artificial intelligent driven power system analysis method and intelligent software platform | |
Goudarzi et al. | Intelligent analysis of wind turbine power curve models | |
US11436494B1 (en) | Optimal power flow computation method based on multi-task deep learning | |
CN107292502A (en) | A kind of distribution network reliability evaluation method | |
CN109412161A (en) | A kind of Probabilistic Load calculation method and system | |
CN110991737A (en) | Ultra-short-term wind power prediction method based on deep belief network | |
US20230307911A1 (en) | Method and system for optimizing power grid emergency load-shedding based on proliferation and reduction evolution | |
Bo et al. | A DNN-based reliability evaluation method for multi-state series-parallel systems considering semi-Markov process | |
CN105809286A (en) | Representative data reconstruction-based incremental SVR (support vector regression) load prediction method | |
CN117421571A (en) | Topology real-time identification method and system based on power distribution network | |
Vohra et al. | End-to-end learning with multiple modalities for system-optimised renewables nowcasting | |
CN106202805B (en) | A kind of switchgear method for evaluating reliability | |
CN116307287B (en) | Prediction method, system and prediction terminal for effective period of photovoltaic power generation | |
CN112000708A (en) | Abnormal data processing method and system based on regulation and control adapted data fusion | |
CN115526393B (en) | Manufacturing cost prediction method based on screening of key influencing factors of power transformation engineering | |
CN107730002A (en) | A kind of communication network shutdown remote control parameter intelligent fuzzy comparison method | |
Huy et al. | Short-term load forecasting in power system using CNN-LSTM neural network |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |