CN111049159A - Power system transient stability prevention control method embedded into deep belief network - Google Patents
Power system transient stability prevention control method embedded into deep belief network Download PDFInfo
- Publication number
- CN111049159A CN111049159A CN201911330999.5A CN201911330999A CN111049159A CN 111049159 A CN111049159 A CN 111049159A CN 201911330999 A CN201911330999 A CN 201911330999A CN 111049159 A CN111049159 A CN 111049159A
- Authority
- CN
- China
- Prior art keywords
- generator
- transient stability
- output
- power
- belief network
- 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
- 230000001052 transient effect Effects 0.000 title claims abstract description 103
- 230000002265 prevention Effects 0.000 title claims abstract description 40
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000004088 simulation Methods 0.000 claims abstract description 20
- 238000004364 calculation method Methods 0.000 claims abstract description 19
- 238000011217 control strategy Methods 0.000 claims abstract description 15
- 230000001687 destabilization Effects 0.000 claims abstract description 5
- 230000003449 preventive effect Effects 0.000 claims description 19
- 238000005457 optimization Methods 0.000 claims description 12
- 238000005070 sampling Methods 0.000 claims description 11
- 230000002068 genetic effect Effects 0.000 claims description 10
- 238000010586 diagram Methods 0.000 description 12
- 230000006870 function Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000013097 stability assessment Methods 0.000 description 1
Images
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
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
- H02J3/48—Controlling the sharing of the in-phase component
-
- 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
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
- H02J3/50—Controlling the sharing of the out-of-phase component
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the technical field of power system automation technology, and aims to provide a power system transient stability prevention control method embedded into a deep belief network, which comprises the following steps of S1: determining the active output fluctuation range and the load fluctuation range of the generator, generating N generator active output samples, acquiring a large amount of initial state data, and performing time domain simulation calculation on the initial state data to generate sample data; s2: establishing a deep belief network, training the deep belief network by using sample data, fitting the active power output of the generator and the transient stability of the system, and generating a transient stability evaluator of the power system; s3: adding cost constraint, power flow constraint and stable operation constraint for the NSGA-II algorithm based on the transient stability constraint condition, and building an NSGA-II evolutionary algorithm model; s4: and acquiring the output fluctuation range and the load fluctuation range of the transient destabilization generator under the fault, and iteratively optimizing the NSGA-II evolutionary algorithm model to obtain a prevention control strategy.
Description
Technical Field
The invention belongs to the technical field of power system automation, and particularly relates to a power system transient stability prevention control method embedded into a deep belief network.
Background
With the increasing scale of the power system, the network structure is more complex, the system operation point is closer to the stability limit, and the requirement on the stability prevention and control of the power system is higher. Transient instability is often the main cause of large-scale accidents of power systems, and effective transient stability assessment and accident prevention measures of power systems are the key to solve the problems. The traditional transient stability calculation usually adopts a method of time domain simulation plus proper criterion, has the advantages of accurate calculation and high reliability, but the model contains a nonlinear differential algebraic equation, has complex calculation and long operation time, and is difficult to meet the requirement of on-line calculation. The deep learning model has the advantages of automatic feature extraction, strong abstract capability and good convergence, has a deeper network structure, and is more favorable for discovering internal rules of data, so that the deep learning model is used for transient stability evaluation of a power system. The prevention control means that before a system fails, the potential failure risk of the system is found in advance by identifying the current system state, and the system is adjusted to a state which can still stably run after the system fails by adjusting the output of a generator and changing the load. The prevention control is combined with the transient stability, the transient stability prevention control is provided, and the system is ensured to operate in a state meeting the transient stability requirement.
Disclosure of Invention
The invention aims to provide a power system transient stability prevention control method embedded with a deep belief network, which introduces the deep belief network and NSGA-II into transient stability prevention control and realizes the quick and stable acquisition of a fault transient stability prevention control optimization strategy.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows: a power system transient stability prevention control method embedded into a deep belief network comprises the following steps:
s1: determining the active output fluctuation range and the load fluctuation range of the generator, generating N generator active output samples, acquiring a large amount of initial state data, performing time domain simulation calculation on the initial state data, generating sample data, and executing S2;
s2: establishing a deep belief network, training the deep belief network by using sample data, fitting the active power output of the generator and the transient stability of the system to generate a transient stability evaluator of the power system, embedding the transient stability evaluator of the power system into a non-dominated sorting genetic algorithm NSGA-II as a transient stability constraint condition, and executing S3;
s3: adding cost constraint, power flow constraint and stable operation constraint for the NSGA-II algorithm based on the transient stability constraint condition, building an NSGA-II evolutionary algorithm model, and executing S4;
s4: and acquiring the output fluctuation range and the load fluctuation range of the transient destabilization generator under the fault, and iteratively optimizing the NSGA-II evolutionary algorithm model to obtain a prevention control strategy.
Preferably, in S1, a latin hypercube sampling algorithm is used to obtain N generator active output samples according to the generator active output fluctuation range and the load fluctuation range.
Preferably, the generating of the sample data at S1 includes the following steps,
s11: selecting M lines in the system as an expected fault set, selecting one line as a fault line in each time of time domain simulation calculation, wherein the fault type of the fault line is a three-phase short-circuit fault, and executing S12;
s12: combining M fault lines in the expected fault set and N generator active power output samples in pairs to generate M × N transient stability simulation initial conditions, and executing S13;
s13: performing time domain simulation calculation on the M x N transient stability simulation initial conditions, obtaining a power angle curve of the generator, calculating M x N transient stability coefficients TSI, and executing S14, wherein the active power output of each generator corresponds to the M TSIs;
s14: and selecting the minimum TSI of the M TSIs and the corresponding active power output of the generator to combine into sample data of a training deep belief network, and generating N sample data.
Preferably, the TSI is calculated as:
wherein ,δmaxFor the maximum power angle difference between any two generators in the system, when TSI>0, the transient stability of the system is stable, and the larger the TSI value is, the higher the transient stability of the system is; when TSI<0, system transient instability.
Preferably, in S2, the power system transient stability estimator is embedded in the non-dominated ranking genetic algorithm NSGA-II and expressed as:
wherein ,the active output of the generator is given for prevention and control; phi (P)G) The input of the model is active output of all generators of the system, the output is TSI estimated by the model, and when the estimated TSI is used as the model of the DBN, the active output is the TSI estimated by the model>0, the DBN model considers that the transient state of the system is stable under the active output condition, and otherwise, the transient state of the system is unstable, and preventive control needs to be adopted.
Preferably, the cost constraint is expressed as
wherein ,CUiThe cost is adjusted up for the output of the generator; cDiThe cost is reduced for the output of the generator; pOiThe output of the generator before control is prevented; pPiTo prevent the generator from outputting power after control; delta PUiAdjusting the output of the generator; delta PDiAnd (4) adjusting the output of the generator down.
Preferably, the power flow constraint is expressed as
wherein ,PNi and QNiInjecting power for the active and reactive of the node; pDi and QDiActive and reactive output power for the node; vi and VjIs the node voltage amplitude αijIs the node voltage phase angle difference; gij and BijReal and imaginary parts of node admittance; snIs a collection of nodes.
Preferably, the steady operation constraint is expressed as
wherein ,andthe upper limit and the lower limit of active output of the generator are set;andoutputting upper and lower limits for a reactive power source;andthe upper limit and the lower limit of the node voltage are set;andupper and lower limits for line thermal stability constraints; slIs a set of lines.
Preferably, the S4 includes the steps of,
s41: obtaining the active power output of the generator, inputting the active power output to the deep belief network, and executing S42;
s42: judging whether the system is unstable and needs to be prevented and controlled, if so, executing S43, and if not, executing S43;
s43: the NSGA-II algorithm carries out cross and genetic iterative optimization, and S43 is executed;
s44: whether the iteration times reach the maximum iteration times or not is judged, if yes, S45 is executed, and if not, S43 is executed;
s45: and outputting the preventive control strategy.
Preferably, the deep belief network is based on a Keras framework of TensorFlow, and the number of layers of the deep belief network is four, wherein the deep belief network comprises two limited Boltzmann machine layers and a full-connection layer.
In conclusion, the beneficial effects of the invention are as follows:
the invention introduces the deep belief network and NSGA-II into the transient stability prevention control, and realizes the quick and stable acquisition of the transient stability prevention control optimization strategy aiming at faults.
Drawings
FIG. 1 is a schematic diagram illustrating steps of a transient stability prevention control method for an electric power system embedded in a deep belief network according to the present invention;
FIG. 2 is a schematic diagram of the present invention showing the Latin hypercube sampling principle;
fig. 3 is a schematic diagram illustrating the step S1 of the transient stability prevention control method for the power system embedded with the deep belief network according to the present invention;
FIG. 4 is a schematic diagram of the present invention for illustrating a deep belief network;
FIG. 5 is a schematic diagram of the transient stability estimator of the present invention;
FIG. 6 is a schematic diagram of the present invention for demonstrating deep belief network embedding NSGA-II;
fig. 7 is a schematic flowchart of S4 illustrating a transient stability prevention control method for a power system embedded in a deep belief network according to the present invention;
fig. 8 is a schematic diagram illustrating the step S4 of the transient stability prevention control method for the power system embedded with the deep belief network according to the present invention;
FIG. 9 is a schematic diagram showing an IEEE39, 68, 140 node system according to an embodiment of the present invention;
FIG. 10 is a schematic diagram showing the active output of 1000 generators generated by Latin hypercube sampling according to an embodiment of the present invention;
FIG. 11 is a schematic diagram showing the active generator output after preventive control according to an embodiment of the present invention;
FIG. 12 is a schematic diagram showing the cost after preventive control according to the embodiment of the present invention;
FIG. 13 is a schematic diagram of an embodiment of the present invention for showing TSI after preventive control;
fig. 14 is a graph showing power angles before and after preventive control of a 39-node system under fault 7 according to an embodiment of the present invention;
fig. 15 is a graph showing power angles before and after preventive control of a 39-node system under a fault 10 according to an embodiment of the present invention;
fig. 16 is a graph showing power angles before and after prevention and control of a 68-node system under a fault 8 according to an embodiment of the present invention;
fig. 17 is a graph showing power angles before and after prevention and control of a 68-node system under a fault 9 according to an embodiment of the present invention;
fig. 18 is a graph showing power angles before and after preventive control of the 140-node system under fault 1 according to the embodiment of the present invention;
fig. 19 is a graph showing power angles before and after preventive control of the 140-node system under the fault 2 according to the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to fig. 1 to 19 of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Examples
Referring to fig. 1, a method for preventing and controlling transient stability of a power system embedded in a deep belief network includes the following steps:
s1: determining the active output fluctuation range and the load fluctuation range of the generator, generating N generator active output samples, acquiring a large amount of initial state data, performing time domain simulation calculation on the initial state data, generating sample data, and executing S2;
s2: establishing a deep belief network, training the deep belief network by using sample data, fitting the active power output of the generator and the transient stability of the system to generate a transient stability evaluator of the power system, embedding the transient stability evaluator of the power system into a non-dominated sorting genetic algorithm NSGA-II as a transient stability constraint condition, and executing S3;
s3: adding cost constraint, power flow constraint and stable operation constraint for the NSGA-II algorithm based on the transient stability constraint condition, building an NSGA-II evolutionary algorithm model, and executing S4;
s4: and acquiring the output fluctuation range and the load fluctuation range of the transient destabilization generator under the fault, and iteratively optimizing the NSGA-II evolutionary algorithm model to obtain a prevention control strategy.
S1 will be described in detail below.
Specifically, in S1, a latin hypercube sampling algorithm is used to obtain N generator active output samples according to the generator active output fluctuation range and the load fluctuation range.
Referring to fig. 2, the latin hypercube sampling principle is: according to the sampling number N, the sampling range is divided into N equal parts, and one point is selected in each equal part, so that the samples are distributed in the whole sample space and have certain randomness. The active power output of the generator is set to fluctuate within the range of 90% -110%, the active power output of the generator is sampled based on Latin hypercube sampling, N generator active power output samples which are uniformly distributed are generated within the active power output range of the generator, and load active power and reactive power fluctuate up and down along with the active power of the generator according to the active power balance and load power factor constant principle.
Specifically, referring to fig. 3, generating sample data at S1 includes the steps of,
s11: selecting M lines in the system as an expected fault set, selecting one line as a fault line in each time of time domain simulation calculation, and executing S12, wherein the fault type of the fault line is a three-phase short-circuit fault;
s12: combining M fault lines in the expected fault set and N generator active output samples generated by the Latin hypercube sampling pairwise to generate M x N transient stability simulation initial conditions, and executing S13;
s13: performing time domain simulation calculation on the M x N transient stability simulation initial conditions, obtaining a power angle curve of the generator, calculating M x N transient stability coefficients TSI, and executing S14, wherein the active power output of each generator corresponds to the M TSIs;
s14: and selecting the minimum TSI of the M TSIs and the corresponding active power output of the generator to combine into sample data of a training deep belief network, and generating N sample data.
Specifically, the TSI is calculated as:
wherein ,δmaxFor the maximum power angle difference between any two generators in the system, when TSI>0, the transient stability of the system is stable, and the larger the TSI value is, the higher the transient stability of the system is; when TSI<0, system transient instability.
S2 will be described in detail below.
Specifically, the deep belief network is trained by using sample data, a nonlinear relation between the active power output of the generator and the transient stability of the system is fitted, and the transient stability evaluator of the power system based on the deep belief network is generated.
Specifically, referring to fig. 4, the deep belief network is built based on a tensrflow Keras framework, the number of built deep belief network layers is 4, and the deep belief network comprises 2 limited boltzmann machine layers and a full-connection layer. The deep belief network training is divided into two stages, the first stage is pre-training, and each layer of restricted Boltzmann machine carries out greedy unsupervised learning layer by using sample data without labels. Through pre-training, the deep belief network is near the optimal solution, and the problem that the deep neural network cannot be trained due to gradient loss or gradient explosion is solved. And in the second stage, the model is integrally trained by using sample data with labels, and the weight and the bias are finely adjusted on the basis of pre-training through a random gradient descent algorithm and back propagation to achieve the optimal fitting effect.
The energy function and the joint probability function of the deep belief network are respectively as follows:
E(v,h1,h2,h3)=-vTW1h1-
and if the number of the generators is L, the number of the neurons of each layer of the deep belief network is sequentially as follows: l-100-50-1. The learning rate is set to be 0.0001, the batch processing number is 50, the pre-training frequency is 50, the activation function is ReLU, the iteration frequency of the whole training is 1000, a value range can be set firstly for selecting the hyper-parameters, then the hyper-parameters are optimized through optimization algorithms such as a particle swarm algorithm, the accuracy of the deep belief network is used as a standard for judging the performance of the hyper-parameters, the optimal hyper-parameters are found, and the deep belief network achieves the optimal fitting effect.
Specifically, a learning rate attenuation method, mean square error MSE and L2 regularization are adopted in the model training process. The learning rate attenuation method can improve the learning speed in the early stage of training and improve the evaluation accuracy in the later stage of training; the gradient of MSE loss is reduced along with the reduction of loss, the gradient is very small when the loss approaches to 0, and the MSE is more accurate than the average absolute error MAE calculation result when the training is finished; l2 regularization may prevent model overfitting. The mathematical expressions are respectively:
in the formula ,LrIs the learning rate; d is a learning rate attenuation coefficient; e is the training times; phi (x) is a DBN model; x is the number ofiIs a training set; y isiIs equal to xiA corresponding tag value; m is the number of training set samples; n is the number of DBN layers; omegaiIs a weight coefficient; λ is the regularization parameter.
It should be noted that, referring to fig. 5, the transient stability estimator of the power system is embedded in the non-dominated sorting genetic algorithm NSGA-II as a transient stability constraint condition, specifically:
referring to fig. 6, the transient stability estimator of the power system is embedded in the non-dominated sorting genetic algorithm NSGA-II, instead of solving the time domain equation, for the determination of the transient stability of the system, which can be expressed by the following formula:
in the formula ,the active output of the generator is given for prevention and control; phi (P)G) The input of the model is active output of all generators of the system, the output is TSI estimated by the model, and when the estimated TSI is used as the model of the DBN, the active output is the TSI estimated by the model>0,The DBN model considers the transient stability of the system under this active power output condition, whereas the system transient instability is considered and preventive control is required.
S3 will be specifically described below.
Specifically, NSGA-II is a multi-objective optimization algorithm, and can search for an optimization result which simultaneously meets a plurality of constraint conditions. The NSGA-II provided by the invention has four optimization targets which are respectively as follows: and the control and regulation cost, the power flow constraint, the stable operation constraint and the transient stability constraint are prevented. The NSGA-II algorithm considers four calculation results of optimization targets at the same time, and gives an optimal prevention control strategy through modes of crossing, variation and loop iteration.
In particular, the cost constraint is expressed as
wherein ,CUiThe cost is adjusted up for the output of the generator; cDiThe cost is reduced for the output of the generator; pOiThe output of the generator before control is prevented; pPiTo prevent the generator from outputting power after control; delta PUiAdjusting the output of the generator; delta PDiAnd (4) adjusting the output of the generator down.
Specifically, the power flow constraint is expressed as
wherein ,PNi and QNiInjecting power for the active and reactive of the node; pDi and QDiActive and reactive output power for the node; vi and VjIs the node voltage amplitude αijIs the node voltage phase angle difference; gij and BijReal and imaginary parts of node admittance; snIs a collection of nodes.
Specifically, the steady state operating constraint is expressed as
wherein ,andthe upper limit and the lower limit of active output of the generator are set;andoutputting upper and lower limits for a reactive power source;andthe upper limit and the lower limit of the node voltage are set;andupper and lower limits for line thermal stability constraints; slIs a set of lines.
S4 will be described in detail below.
Referring to fig. 7 and 8, S4 includes the steps of,
s41: acquiring the active output of the real-time generator, inputting the active output to the deep belief network, and executing S42;
s42: judging whether the system is unstable and needs to be prevented and controlled, if so, executing S43, and if not, executing S43;
it is worth to be noted that, the power system transient stability evaluator calculates the TSI, when the evaluated TSI is greater than 0, the DBN model considers that the system transient is stable under the active power output condition, otherwise, the system transient instability is considered, and the prevention control is needed;
s43: the NSGA-II algorithm carries out cross and genetic iterative optimization, and S43 is executed;
it is worth explaining that the power flow constraint and the stable operation constraint are judged whether to be met through a power flow calculation result, the transient stable constraint is judged whether to be met through a deep belief network evaluation result, and the prevention control adjustment cost is obtained through calculation of a difference value between the active power output of the generator and the original active power output of the generator according to a prevention control strategy;
s44: whether the iteration times reach the maximum iteration times or not is judged, if yes, S45 is executed, and if not, S43 is executed;
s45: and outputting the preventive control strategy.
It should be noted that step S45 specifically includes, after the preventive control strategy is obtained, using the power system toolbox to obtain a power angle curve of the generator under the preventive control strategy, calculating a transient stability coefficient TSI, and determining the accuracy of the preventive control strategy; and the dispatching personnel adjusts the original active power output of the generator of the system according to the active power output of the generator given by the preventive control strategy, so that the system meets the transient stability requirement.
Referring to fig. 9, a power system transient stability prevention control method embedded in a deep belief network proposed by the present disclosure is described below by taking IEEE39, 68, 140 node systems as an example.
10 lines of all lines of each system are selected as expected fault lines, and three-phase short-circuit faults in the middle of the lines are set to form an expected fault set, as shown in table 1.
TABLE 1
Referring to fig. 10, the active power output of the generator is set to fluctuate within a range of 90% to 110% for the expected fault concentration fault line and fault type. 1000 generator active output conditions are generated by using Latin hypercube sampling, and the system load active power and reactive power are enabled to integrally float up and down along with the change of the total active output of the generator according to the principle of active power balance and constant load power factor. The 1000 generator active power output situations are combined with 10 predicted fault lines with concentrated predicted faults to form 10000 predicted fault data. And (3) performing time domain simulation calculation on expected fault data by using a power system tool kit (PST), setting the fault removal time to be 0.1 second and the total simulation time to be 20 seconds, and solving the corresponding TSI. Each generator output condition corresponds to 10 TSIs, the minimum TSI in the 10 TSIs is screened, and 1000 sample data used for training the transient stability estimator are formed together with the generator output.
Specifically, 1000 sample data are divided into a training set containing 800 data and a test set containing 200 data, the transient stability evaluator is trained by using the training set, and the accuracy of the model is verified by the test set. The invention uses a Keras framework based on TensorFlow to build a deep belief network, the number of layers of the built deep belief network is 4, and the deep belief network comprises 2 limited Boltzmann machine layers and a full connection layer. And if the number of the generators is L, the number of the neurons of each layer of the deep belief network is sequentially as follows: l-100-50-1. The learning rate is set to be 0.0001, the batch processing number is 50, the pre-training frequency is 50, the activation function is ReLU, the iteration frequency of the whole training is 1000, a value range can be set firstly for selecting the hyper-parameters, then the hyper-parameters are optimized through optimization algorithms such as a particle swarm algorithm, the accuracy of the deep belief network is used as a standard for judging the performance of the hyper-parameters, the optimal hyper-parameters are found, and the deep belief network achieves the optimal fitting effect. The different test system identification accuracy pairs are shown in table 2:
TABLE 2
Referring to fig. 7, the output fluctuation range and the load fluctuation range of the transient destabilization generator under the fault are obtained, and the NSGA-II evolutionary algorithm model is subjected to iterative optimization to obtain a prevention control strategy.
Referring to fig. 11 to 13, the power system toolbox PST is used to verify the preventive control strategy, and a time domain simulation method is used to calculate the TSI before and after preventive control under each expected fault. The fault of transient instability before prevention control is changed into transient stability through prevention control, the fault of transient stability before prevention control still keeps the transient stability, and the TSI is slightly promoted, which shows that the strategy also properly increases the system transient stability margin facing to the full fault set.
Referring to fig. 14-19, the comparison of the power angle curves before and after prevention control shows that the system returns from transient instability to transient stability through predictive control, and the feasibility of the method is verified.
In the description of the present invention, it is to be understood that the terms "counterclockwise", "clockwise", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate orientations or positional relationships based on those shown in the drawings, and are used for convenience of description only, and do not indicate or imply that the referenced devices or elements must have a particular orientation, be constructed and operated in a particular orientation, and thus, are not to be considered as limiting.
Claims (10)
1. A power system transient stability prevention control method embedded into a deep belief network is characterized by comprising the following steps,
s1: determining the active output fluctuation range and the load fluctuation range of the generator, generating N generator active output samples, acquiring a large amount of initial state data, performing time domain simulation calculation on the initial state data, generating sample data, and executing S2;
s2: establishing a deep belief network, training the deep belief network by using sample data, fitting the active power output of the generator and the transient stability of the system to generate a transient stability evaluator of the power system, embedding the transient stability evaluator of the power system into a non-dominated sorting genetic algorithm NSGA-II as a transient stability constraint condition, and executing S3;
s3: adding cost constraint, power flow constraint and stable operation constraint for the NSGA-II algorithm based on the transient stability constraint condition, building an NSGA-II evolutionary algorithm model, and executing S4;
s4: and acquiring the output fluctuation range and the load fluctuation range of the transient destabilization generator under the fault, and iteratively optimizing the NSGA-II evolutionary algorithm model to obtain a prevention control strategy.
2. The method for preventing and controlling transient stability of an electric power system embedded in a deep belief network as claimed in claim 1, wherein N generator active output samples are obtained according to the generator active output fluctuation range and the load fluctuation range by using a latin hypercube sampling algorithm in S1.
3. The method for preventing transient stability of power system embedded with deep belief network as claimed in claim 1 or 2, wherein the step of S1 generating sample data comprises the steps of,
s11: selecting M lines in the system as an expected fault set, selecting one line as a fault line in each time of time domain simulation calculation, wherein the fault type of the fault line is a three-phase short-circuit fault, and executing S12;
s12: combining M fault lines in the expected fault set and N generator active power output samples in pairs to generate M × N transient stability simulation initial conditions, and executing S13;
s13: performing time domain simulation calculation on the M x N transient stability simulation initial conditions, obtaining a power angle curve of the generator, calculating M x N transient stability coefficients TSI, and executing S14, wherein the active power output of each generator corresponds to the M TSIs;
s14: and selecting the minimum TSI of the M TSIs and the corresponding active power output of the generator to combine into sample data of a training deep belief network, and generating N sample data.
4. The method for preventing and controlling transient stability of power system embedded in deep belief network as claimed in claim 3, wherein the TSI is calculated by:
wherein ,δmaxFor the maximum power angle difference between any two generators in the system, when TSI>0, the transient stability of the system is stable, and the larger the TSI value is, the higher the transient stability of the system is; when TSI<0, system transient instability.
5. The method for preventing and controlling transient stability of power system embedded with deep belief network as claimed in claim 1, wherein the step of embedding the transient stability estimator of power system into the non-dominated sorting genetic algorithm NSGA-II in S2 is represented as:
wherein ,the active output of the generator is given for prevention and control; phi (P)G) The input of the model is active output of all generators of the system, the output is TSI estimated by the model, and when the estimated TSI is used as the model of the DBN, the active output is the TSI estimated by the model>0, the DBN model considers that the transient state of the system is stable under the active output condition, and otherwise, the transient state of the system is unstable, and preventive control needs to be adopted.
6. The method according to claim 1, wherein the cost constraint is expressed as
wherein ,CUiThe cost is adjusted up for the output of the generator; cDiThe cost is reduced for the output of the generator; pOiThe output of the generator before control is prevented; pPiTo prevent the generator from outputting power after control; delta PUiAdjusting the output of the generator; delta PDiAnd (4) adjusting the output of the generator down.
7. The method according to claim 1, wherein the power system transient stability prevention control method is characterized in that the power flow constraint is expressed as
wherein ,PNi and QNiInjecting power for the active and reactive of the node; pDi and QDiActive and reactive output power for the node; vi and VjIs the node voltage amplitude αijIs the node voltage phase angle difference; gij and BijReal and imaginary parts of node admittance; snIs a collection of nodes.
8. The method according to claim 1, wherein the steady operation constraint is expressed as
wherein ,andthe upper limit and the lower limit of active output of the generator are set;andoutputting upper and lower limits for a reactive power source;andthe upper limit and the lower limit of the node voltage are set;andupper and lower limits for line thermal stability constraints; slIs a set of lines.
9. The method for controlling transient stability of an embedded deep belief network power system as claimed in any one of claims 4 to 8, wherein the S4 comprises the steps of,
s41: obtaining the active power output of the generator, inputting the active power output to the deep belief network, and executing S42;
s42: judging whether the system is unstable and needs to be prevented and controlled, if so, executing S43, and if not, executing S43;
s43: the NSGA-II algorithm carries out cross and genetic iterative optimization, and S43 is executed;
s44: whether the iteration times reach the maximum iteration times or not is judged, if yes, S45 is executed, and if not, S43 is executed;
s45: and outputting the preventive control strategy.
10. The method for preventing and controlling transient stability of the power system embedded with the deep belief network according to any one of claims 4 to 8, wherein the deep belief network is based on a Keras framework of TensorFlow, and the number of layers of the deep belief network is four, including two restricted Boltzmann machines and a full connection layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911330999.5A CN111049159B (en) | 2019-12-20 | 2019-12-20 | Power system transient stability prevention control method embedded in deep belief network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911330999.5A CN111049159B (en) | 2019-12-20 | 2019-12-20 | Power system transient stability prevention control method embedded in deep belief network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111049159A true CN111049159A (en) | 2020-04-21 |
CN111049159B CN111049159B (en) | 2023-09-29 |
Family
ID=70238189
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911330999.5A Active CN111049159B (en) | 2019-12-20 | 2019-12-20 | Power system transient stability prevention control method embedded in deep belief network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111049159B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111950765A (en) * | 2020-07-06 | 2020-11-17 | 四川大川云能科技有限公司 | Probabilistic transient stability prediction method based on stacked noise reduction self-encoder |
CN113591379A (en) * | 2021-07-27 | 2021-11-02 | 四川大学 | Electric power system transient stability prevention and emergency coordination control auxiliary decision method |
CN113904384A (en) * | 2021-11-09 | 2022-01-07 | 国网四川省电力公司电力科学研究院 | Power grid transient stability coordination control method and system based on gradient elevator |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS63181618A (en) * | 1987-01-22 | 1988-07-26 | 中部電力株式会社 | Preventive controller of power system |
CN107590564A (en) * | 2017-09-08 | 2018-01-16 | 四川大学 | Power system active power output method of adjustment based on Transient Stability Constraints |
CN109086913A (en) * | 2018-07-11 | 2018-12-25 | 山东大学 | A kind of transient stability evaluation in power system method and system based on deep learning |
CN110163540A (en) * | 2019-06-28 | 2019-08-23 | 清华大学 | Electric power system transient stability prevention and control method and system |
-
2019
- 2019-12-20 CN CN201911330999.5A patent/CN111049159B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS63181618A (en) * | 1987-01-22 | 1988-07-26 | 中部電力株式会社 | Preventive controller of power system |
CN107590564A (en) * | 2017-09-08 | 2018-01-16 | 四川大学 | Power system active power output method of adjustment based on Transient Stability Constraints |
CN109086913A (en) * | 2018-07-11 | 2018-12-25 | 山东大学 | A kind of transient stability evaluation in power system method and system based on deep learning |
CN110163540A (en) * | 2019-06-28 | 2019-08-23 | 清华大学 | Electric power system transient stability prevention and control method and system |
Non-Patent Citations (3)
Title |
---|
LIU C,ETC: "A systematic approach for dynamic security assessment and the corresponding preventive control scheme based on decision trees", 《IEEE TRANSACTIONS ON POWER SYSTEMS》 * |
杨跃,等: "基于神经网络预测校核的暂态稳定预防控制", 《电网技术》 * |
田芳等: "基于灵敏度分析和时域仿真的暂态稳定预防控制优化方法", 《电力自动化设备》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111950765A (en) * | 2020-07-06 | 2020-11-17 | 四川大川云能科技有限公司 | Probabilistic transient stability prediction method based on stacked noise reduction self-encoder |
CN111950765B (en) * | 2020-07-06 | 2024-04-19 | 四川大川云能科技有限公司 | Probabilistic transient stability prediction method based on stacked noise reduction self-encoder |
CN113591379A (en) * | 2021-07-27 | 2021-11-02 | 四川大学 | Electric power system transient stability prevention and emergency coordination control auxiliary decision method |
CN113591379B (en) * | 2021-07-27 | 2023-11-14 | 四川大学 | Auxiliary decision-making method for transient stability prevention and emergency coordination control of power system |
CN113904384A (en) * | 2021-11-09 | 2022-01-07 | 国网四川省电力公司电力科学研究院 | Power grid transient stability coordination control method and system based on gradient elevator |
Also Published As
Publication number | Publication date |
---|---|
CN111049159B (en) | 2023-09-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Liu et al. | A Bayesian learning based scheme for online dynamic security assessment and preventive control | |
CN111049159A (en) | Power system transient stability prevention control method embedded into deep belief network | |
CN111523785A (en) | Power system dynamic security assessment method based on generation countermeasure network | |
EP2703914A1 (en) | Hybrid control system and method for automatic voltage regulator | |
CN107093895B (en) | online transient state safety and stability assessment method based on automatic screening of expected fault set | |
Al Gizi et al. | Sugeno fuzzy PID tuning, by genetic-neutral for AVR in electrical power generation | |
Mestav et al. | State estimation for unobservable distribution systems via deep neural networks | |
CN116245033B (en) | Artificial intelligent driven power system analysis method and intelligent software platform | |
Zhong et al. | Bayesian learning-based multi-objective distribution power network reconfiguration | |
CN112947672B (en) | Maximum power point tracking method and device for photovoltaic cell | |
CN109066651B (en) | Method for calculating limit transmission power of wind power-load scene | |
Jiang et al. | Explicit model predictive control applications in power systems: an AGC study for an isolated industrial system | |
Qiu et al. | Analytic deep learning-based surrogate model for operational planning with dynamic TTC constraints | |
Wang et al. | Dynamic equivalent method of PMSG‐based wind farm for power system stability analysis | |
Bellizio et al. | Verifying machine learning conclusions for securing low inertia systems | |
Massaoudi et al. | Navigating the landscape of deep reinforcement learning for power system stability control: A review | |
CN114036506A (en) | Method for detecting and defending false data injection attack based on LM-BP neural network | |
CN110675276A (en) | Method and system for inversion droop control of direct current power transmission system | |
Wei et al. | A lite cellular generalized neuron network for frequency prediction of synchronous generators in a multimachine power system | |
CN111950765B (en) | Probabilistic transient stability prediction method based on stacked noise reduction self-encoder | |
Andreoiu et al. | Lyapunov's method based genetic algorithm for multi-machine PSS tuning | |
Aththanayake et al. | Performance Analysis of Regression and Artificial Neural Network Schemes for Dynamic Model Reduction of Power Systems | |
Hussaini et al. | Droop coefficient design in droop control of power converters for improved load sharing: An artificial neural network approach | |
Xu et al. | Coordinated preventive-corrective control for power system transient stability enhancement based on machine learning-assisted optimization | |
Wooding et al. | Control and management of active buildings |
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 |