CN115329911A - Safety correction method for UPFC-containing power system based on SAE two-classification model - Google Patents
Safety correction method for UPFC-containing power system based on SAE two-classification model Download PDFInfo
- Publication number
- CN115329911A CN115329911A CN202211265207.2A CN202211265207A CN115329911A CN 115329911 A CN115329911 A CN 115329911A CN 202211265207 A CN202211265207 A CN 202211265207A CN 115329911 A CN115329911 A CN 115329911A
- Authority
- CN
- China
- Prior art keywords
- correction
- node
- model
- upfc
- safety
- 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
- 238000012937 correction Methods 0.000 title claims abstract description 195
- 238000000034 method Methods 0.000 title claims abstract description 82
- 238000013145 classification model Methods 0.000 title claims abstract description 54
- 238000005457 optimization Methods 0.000 claims abstract description 56
- 238000004364 calculation method Methods 0.000 claims abstract description 41
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 37
- 230000006870 function Effects 0.000 claims description 29
- 230000009471 action Effects 0.000 claims description 24
- 230000004913 activation Effects 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 8
- 210000002569 neuron Anatomy 0.000 claims description 4
- DGAQECJNVWCQMB-PUAWFVPOSA-M Ilexoside XXIX Chemical compound C[C@@H]1CC[C@@]2(CC[C@@]3(C(=CC[C@H]4[C@]3(CC[C@@H]5[C@@]4(CC[C@@H](C5(C)C)OS(=O)(=O)[O-])C)C)[C@@H]2[C@]1(C)O)C)C(=O)O[C@H]6[C@@H]([C@H]([C@@H]([C@H](O6)CO)O)O)O.[Na+] DGAQECJNVWCQMB-PUAWFVPOSA-M 0.000 claims description 3
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000012512 characterization method Methods 0.000 claims description 3
- 230000021615 conjugation Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000001902 propagating effect Effects 0.000 claims description 3
- 229920006395 saturated elastomer Polymers 0.000 claims description 3
- 229910052708 sodium Inorganic materials 0.000 claims description 3
- 239000011734 sodium Substances 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 230000009467 reduction Effects 0.000 abstract description 4
- 238000004458 analytical method Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 8
- 239000000243 solution Substances 0.000 description 7
- 238000012360 testing method Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Separation By Low-Temperature Treatments (AREA)
Abstract
The invention discloses a safety correction method of a power system containing UPFC based on SAE two-classification model, which comprises the steps of firstly constructing a safety correction optimization model of the power system containing UPFC, then solving the constructed safety correction optimization model of the power system containing UPFC, and obtaining the adjustment range of a safety correction node when the power system containing UPFC works; the invention can effectively get rid of the dependence of the current safety correction calculation on the model, can solve the problems of difficult feature extraction and overlong calculation time, and can realize feature dimension reduction through unsupervised learning of a large number of unlabeled samples without manual feature extraction by constructing the SAE two-classification model for judging the correction state of the node; compared with a DNN model, the SAE two-classification model has higher classification accuracy, is more accurate in judgment of the node correction state, and is suitable for wide popularization and use.
Description
Technical Field
The invention relates to the technical field of power system safety control, in particular to a safety correction method for a power system containing UPFC based on an SAE two-classification model.
Background
The safe operation of the power system is always the focus of power workers, and aiming at the line overload working condition after the fault in the system, the safety correction control can improve the power flow distribution by adjusting the output of the generator and the load shedding, and eliminate the line out-of-limit; the UPFC is a flexible alternating current transmission device which is put into operation at present and has the most comprehensive functions, and can quickly adjust the power flow distribution during safety correction, so that the overload of a line is eliminated, and the power flow distribution of a system is balanced.
The influence of reactive power is ignored in the current sensitivity method with higher calculation efficiency, and the obtained out-of-limit correction and adjustment scheme of the system containing the UPFC can cause secondary out-of-limit of bus voltage when the line is overloaded, so that new potential safety hazards are brought to the system, and the method cannot be applied to practical application. Therefore, the safety correction problem of the existing UPFC system still needs to adopt a model-based optimization method, namely, a mathematical analysis method is used for solving an optimization model to obtain corresponding adjustment measures.
The traditional model-based safety correction method cannot meet the actual engineering requirements, and comprises three reasons: (1) The calculated fluctuation is large and the timeliness is poor aiming at different line out-of-limit working conditions; (2) The method has the advantages that no solution is possible to exist under certain line out-of-limit working conditions, and inherent defects exist; (3) The load shedding action is more and the economical efficiency is poor for ensuring the system safety; therefore, a safety correction method for the power system with the UPFC based on the SAE two-classification model needs to be designed.
Disclosure of Invention
The invention aims to solve the safety correction problem after the UPFC is accessed into the power grid, and provides a safety correction method of a power system containing the UPFC based on an SAE (system architecture analysis) two-classification model, wherein the SAE two-classification model for judging the correction state of a node is constructed, the SAE two-classification model has a powerful automatic feature extraction function, and the feature dimension reduction can be realized through unsupervised learning of a large number of unlabelled samples without manual feature extraction; compared with the DNN model, the SAE two-classification model has higher classification accuracy and more accurate judgment on the node correction state.
In order to achieve the purpose, the invention adopts the technical scheme that:
step (A), constructing a safety correction optimization model of the power system containing the UPFC;
step (B), solving the constructed security correction optimization model of the power system containing the UPFC, obtaining the adjustment range of security correction nodes when the power system containing the UPFC works, and recording effective out-of-limit data under different fault conditions, power flow data at fault moments and corresponding security correction optimization results to prepare a sample set;
step (C), an SAE two-classification model for judging the correction state of the node is established and trained, the input of the SAE two-classification model is the load flow data of the fault moment in the sample set and the corresponding safety correction optimization result made in step (B), and the output is the action state of each node in the safety correction optimization result;
step (D), based on the adjustment range of the safety correction node, establishing and training a DNN regression model for calculating a safety correction value, wherein the input of the NN regression model is effective out-of-limit data in the sample set manufactured in the step (B) and the action state of each node in the safety correction optimization result obtained in the step (C), and the action state is output as a system node safety correction adjustment value;
and (E) finishing the safety correction operation of the power system containing the UPFC by using the output system node safety correction adjustment quantity.
Preferably, in step (a), a power system safety correction optimization model containing the UPFC is constructed, as shown in formula (1),
wherein, the first and the second end of the pipe are connected with each other,
representing the number of system nodes;
is an integer variable 0-1 and represents whether the node i participates in the adjustment;
and
respectively representing active and reactive correction values of the node i; introducing weight coefficients
The method is used for improving the priority of node correction state calculation and meeting the safety correction demand; introducing node correction state weights
And correction amount weight
The method is used for guaranteeing that the generator nodes are adjusted preferentially in the correction process and avoiding the action of the load nodes, and the weight values of the generator nodes are all smaller than those of the load nodes.
Preferably, the step (B) is to solve the constructed safety correction optimization model of the power system containing the UPFC, obtain the adjustment range of the safety correction node when the power system containing the UPFC operates, record effective out-of-limit data under different fault conditions, power flow data at the fault moment and corresponding safety correction optimization results to prepare a sample, wherein the specific steps are as follows,
step (B1), solving the constructed safety correction optimization model of the power system containing the UPFC, as shown in formulas (2) and (3),
wherein, superscript denotes conjugation;
and
respectively obtaining the voltage of the head end and the tail end of the serial side of the branch where the UPFC is located;
representing the virtual node voltage introduced by the UPFC serial side;
and
and
and
respectively representing the conductance and sodium conductivity of the circuit on the serial side and the parallel side of the UPFC;
and
and
and
respectively representing initial values of the node generator and the load;
and
respectively representing the real part and the imaginary part of the element at the position i-j in the admittance matrix;
representing the phase angle difference between the node i and the node j;
step (B2), obtaining the adjustment range of the safety correction node when the power system containing the UPFC works and operates, as shown in formulas (4), (5) and (6),
wherein the content of the first and second substances,
and
respectively representing the upper limit values of the capacities of the serial and parallel sides of the UPFC;
；
；
and
and
and
respectively representing the upper limit and the lower limit of the adjustable active power and the adjustable reactive power of the node i;
characterize the active power of the line i-j, and
and
upper and lower limits thereof, respectively;
and (B3) recording effective out-of-limit data under different fault conditions, tidal current data at fault time and corresponding safety correction optimization results to prepare samples, wherein the sample generation method is to randomly adjust loads to enable the system to have heavy load and N-1 and N-2 fault conditions, obtain the effective out-of-limit data of each branch and the tidal current data at the fault time, the effective out-of-limit data is that the out-of-limit volume exceeds 1% -30% of the self-stability limit value of the line, then, the out-of-limit conditions of the line are optimized and solved by a method for solving an optimization model, the safety correction optimization results of each node under the corresponding fault are obtained, and the safety correction optimization results comprise correction states and correction values.
Preferably, step (C) is to build and train an SAE two-class model for determining the correction status of the node, and the SAE two-class model inputs the load flow data at the fault time in the sample set produced in step (B) and the corresponding safety correction optimization result, and outputs the operation status of each node in the safety correction optimization result, wherein the specific steps are as follows,
step (C1), an SAE two-classification model for judging the node correction state is established, wherein the SAE two-classification model comprises an activation function and a classifier, the activation function selects a Sigmoid function, the top classifier adopts a Softmax regression layer, and the activation function is used for performing layer-by-layer feature extraction on high-dimensional input data and inputting the high-dimensional input data into the top classifier to finish classification;
step (C2), training the established SAE two-classification model, wherein the training of the SAE two-classification model comprises two stages of pre-training and fine-tuning, and the concrete steps are as follows,
step (C21), in the pre-training stage, using a label-free sample to perform unsupervised training, initializing each layer parameter of the SAE two-classification model, and converting high-dimensional input into low-dimensional hidden layer feature expression;
and (C22) in the fine adjustment stage, performing supervised training by using the samples with the labels, and performing fine adjustment on parameters of the whole network.
Preferably, step (D) is to establish and train a DNN regression model for calculating a safety correction amount based on the safety correction node adjustment range, and the input of the NN regression model is to produce effective out-of-limit data in the sample set in step (B), the operation state of each node in the safety correction optimization result obtained in step (C), and output as the safety correction adjustment amount of the system node,
step (D1), a DNN regression model for calculating a safety correction value is established, wherein the DNN regression model shares six layers and comprises four hidden layers, the number of neurons of the hidden layers is 1024, 512, 512 and 512, and the activating functions are non-saturated and non-linear ReLU functions;
and (D2) training the constructed DNN regression model, wherein the DNN regression model is obtained by sending input data into a network according to batches, calculating layer by layer forwards until an output layer, calculating the value of a loss function, reversely propagating and updating the parameters of the network to achieve the optimal characterization capability of the model, the training process is off-line training, and the correction value of the DNN regression model is determined by using the learning capability of the DNN model aiming at action nodes.
Preferably, the labeled sample in the step (C2) is generated by solving a safety correction model under each fault condition, wherein the labeled sample is characterized by system load flow data under the fault condition, and the labeled sample data is sample data indicating whether a node acts or not; the label-free sample only needs system load flow data under fault conditions, does not need labels of whether nodes act or not, and can be generated through load flow calculation under each fault condition.
The invention has the beneficial effects that:
(1) Aiming at the complex mapping relation of sample data, a two-stage deep learning framework is constructed, and two-stage input and output layers are designed to separate learning targets, so that the learning pressure of a DNN regression model is effectively relieved, and the learning precision of the DNN regression model is improved; and test results show that the data driving model has higher precision, the calculation time is stable and is far shorter than that of other methods, the requirements of the system safety correction on accuracy and rapidity are met, a decision maker is facilitated to rapidly make a correction strategy, and the system stability is improved.
(2) The invention can improve the accuracy of the node correction state judgment model by adopting the non-supervised learning of the non-labeled samples in the training process of the SAE two-classification model, and constructs the SAE two-classification model for judging the node correction state aiming at the problem of overhigh input characteristic dimension of the node adjustment state classification model, the SAE two-classification model has a powerful automatic characteristic extraction function, and the feature dimension reduction can be realized by the non-supervised learning of a large number of non-labeled samples without manually extracting the features; the test result shows that compared with the DNN model, the SAE two-classification model has higher classification accuracy and more accurate judgment on the node correction state.
(3) The invention constructs the DNN regression model for calculating the safety correction value, the whole method of the DNN regression model is purely data-driven, the problems of no solution in iteration and long calculation time based on the model method are solved, the data-driven model is not influenced by the system scale and the fault type during on-line calculation, the applicability and the practicability of safety correction calculation are expanded, and the method has engineering application value in large-system fault correction.
Drawings
FIG. 1 is a schematic diagram of a UPFC equivalent injection power model of the present invention;
FIG. 2 is a schematic diagram of the SAE two-class model structure of the present invention;
FIG. 3 is a two-stage deep learning framework diagram of the present invention;
FIG. 4 is a schematic overall algorithm flow diagram of the present invention;
FIG. 5 is a schematic diagram of the calculation time of the safety correction under different algorithms of the present invention.
Detailed Description
The invention will be further described with reference to the accompanying drawings.
As shown in fig. 1-5, the safety correction method of power system with UPFC based on SAE classification model of the present invention includes the following steps,
step (A), constructing a safety correction optimization model of the power system containing the UPFC, as shown in a formula (1),
wherein, the first and the second end of the pipe are connected with each other,
representing the number of system nodes;
is an integer variable 0-1 and represents whether the node i participates in the adjustment or not;
and
respectively representing active and reactive correction values of a node i; introducing weight coefficients
The method is used for improving the priority of node correction state calculation and meeting the safety correction demand; introducing node correction state weights
And a correction amount weight
The method is used for ensuring that the generator nodes are preferentially adjusted in the correction process and avoiding the action of the load nodes, and the weighted values of the generator nodes are all smaller than the load nodes.
Step (B), the constructed security correction optimization model of the power system containing the UPFC is solved, the adjustment range of the security correction node when the power system containing the UPFC works is obtained, effective out-of-limit data under different fault conditions, power flow data at fault moments and corresponding security correction optimization results are recorded to prepare a sample, wherein the specific steps are as follows,
step (B1), solving the constructed safety correction optimization model of the power system containing the UPFC, as shown in formulas (2) and (3),
wherein, superscript denotes conjugation;
and
respectively the voltage of the head end and the tail end of the serial side of the branch where the UPFC is located;
representing the virtual node voltage introduced by the UPFC serial side;
and
and
and
respectively representing the conductance and sodium of the line on the serial and parallel sides of the UPFC;
and
and
and
respectively representing initial values of the node generator and the load;
and
respectively representing the real part and the imaginary part of the element at the position i-j in the admittance matrix;
representing the phase angle difference between the node i and the node j;
step (B2), obtaining the adjustment range of the safety correction node when the power system containing the UPFC is operated, as shown in formulas (4), (5) and (6),
wherein the content of the first and second substances,
and
respectively representing the upper limit values of the capacity of the serial side and the parallel side of the UPFC;
；
；
and
and
and
respectively representing the upper limit and the lower limit of the adjustable active power and the adjustable reactive power of the node i;
characterizing line i-j active powerAnd is made of
And
upper and lower limits thereof, respectively;
and (B3) recording effective out-of-limit data under different fault conditions, tidal current data at fault time and corresponding safety correction optimization results to prepare samples, wherein the sample generation method is to randomly adjust loads to enable the system to have heavy load and N-1 and N-2 fault conditions, obtain the effective out-of-limit data of each branch and the tidal current data at the fault time, the effective out-of-limit data is that the out-of-limit volume exceeds 1% -30% of the self-stability limit value of the line, then, the out-of-limit conditions of the line are optimized and solved by a method for solving an optimization model, the safety correction optimization results of each node under the corresponding fault are obtained, and the safety correction optimization results comprise correction states and correction values.
Step (C), an SAE two-classification model for judging the correction state of the node is established and trained, the input of the SAE two-classification model is the load flow data of the fault moment in the sample set and the corresponding safety correction optimization result made in step (B), and the output is the action state of each node in the safety correction optimization result, wherein the specific steps are as follows,
step (C1), an SAE two-classification model for judging the node correction state is established, wherein the SAE two-classification model comprises an activation function and a classifier, the activation function selects a Sigmoid function, the classifier at the top layer adopts a Softmax regression layer, and the activation function is used for carrying out layer-by-layer feature extraction on high-dimensional input data and then inputting the high-dimensional input data into the classifier at the top layer to finish classification;
the number of hidden layer neurons of the SAE two-classification model is 256, 128, 64, 128 and 256.
Step (C2), training the established SAE two-classification model, wherein the training of the SAE two-classification model comprises two stages of pre-training and fine-tuning, and the concrete steps are as follows,
the node correction state can be judged based on the classification capability of an SAE two-classification model, nodes without action states are screened and eliminated, and an action node set is determined;
step (C21), in the pre-training stage, using a label-free sample to perform unsupervised training, initializing each layer parameter of the SAE two-classification model, and converting high-dimensional input into low-dimensional hidden layer feature expression;
and (C22) in the fine adjustment stage, supervised training is carried out by using the samples with the labels, and the parameters of the whole network are fine adjusted.
The labeled sample in the step (C2) is characterized by system load flow data under fault conditions, and can be generated by solving a safety correction model under each fault condition by using the sample data labeled as whether a node acts or not; the label-free sample only needs system load flow data under fault conditions, does not need labels of whether nodes act or not, and can be generated through load flow calculation under each fault condition.
Step (D), based on the adjustment range of the safety correction node, establishing and training a DNN regression model for calculating a safety correction value, wherein the input of the NN regression model is effective out-of-limit data in a sample set manufactured in the step (B), the action state of each node in the safety correction optimization result obtained in the step (C) is output as a safety correction adjustment quantity of the system node, and the specific steps are as follows,
step (D1), a DNN regression model for calculating a safety correction value is established, wherein the DNN regression model shares six layers and comprises four hidden layers, the number of neurons in the hidden layers is 1024, 512, 512 and 512, and the activating functions are non-saturated and non-linear ReLU functions;
and (D2) training the constructed DNN regression model, wherein the DNN regression model is obtained by sending input data into a network according to batches, calculating layer by layer forwards until an output layer, calculating the value of a loss function, reversely propagating and updating the parameters of the network to achieve the optimal characterization capability of the model, the training process is off-line training, and the correction value of the DNN regression model is determined by using the learning capability of the DNN model aiming at action nodes.
And (E) finishing the safety correction operation of the power system containing the UPFC by using the output system node safety correction adjustment quantity.
To better illustrate the use effect of the present invention, a specific embodiment of the present invention is described below, and the use effect of the present invention is verified.
In order to verify the outstanding power flow control capability of the UPFC in the system, check the correctness and the validity of the safety correction method and test the application performance of the safety correction method, the method provided by the invention is compared with the three existing safety correction methods for testing, and the existing methods are as follows; the method comprises the steps that firstly, a UPFC-free power system safety correction optimization method based on a model is adopted; the existing method II is a UPFC power system safety correction optimization method based on a model; the existing method III is a rapid and safe correction method for a UPFC-containing power system based on data-model hybrid drive, wherein a data drive model is used for judging a node adjustment range, and model drive carries out correction value optimization calculation based on the range;
the method comprises the following steps of (1) respectively representing the prior method I, the prior method II, the prior method III and the method of the invention by M1-M4:
1. the working condition setting is used for analyzing the practicability of the rapid safety correction method provided by the invention in various situations, as shown in the table 1,
TABLE 1 analysis of different out-of-limit conditions
In table 1, working conditions 1-5 indicate that the system has a single branch out-of-limit condition, and working conditions 6 and 7 indicate that the system has multiple branches out-of-limit conditions of different degrees; when the branch where the working condition 4 and the working condition 5 are located is out of limit, the existing model-based optimization calculation method cannot provide a correction result.
2. By analyzing the calculation efficiency, the method can solve the problems that the current model-based safety correction method is high in calculation complexity and easy to generate no solution, and as shown in fig. 5, the calculation time of different safety correction calculation methods under various working conditions is given.
As can be seen from FIG. 5, the model-based security correction method has a long calculation time, and the calculation time is obviously increased along with the increase of the out-of-limit degree of the branch; comparing the calculation efficiency of M1 and M2 under different working conditions, it can be seen that the strong regulation and control capability of UPFC can improve the efficiency of system safety correction to a certain extent, but the effect is not very obvious; the calculation time of M3 is greatly reduced compared with the traditional model-based method due to the addition of the DNN model, but the correction amount of the calculation equipment still depends on the optimization model, so the calculation time still fluctuates under different working conditions; on the contrary, the test result of the method provided by the invention shows that the M4 curve is stable all the time, the time for network offline training is mainly influenced by the difference of the fault conditions and even the increase of the system scale, the influence on the online calculation time is small, and the method provided by the invention has great advantages in the aspect of the correction efficiency.
In addition, the results of the working condition 4 and the working condition 5 are compared, and although the traditional model-based method has no solution, the deep learning models in the M3 and the M4 can quickly judge the correction value of the node under a certain solution-free working condition through the learning capacity of the deep learning models, and a decision maker can quickly make a correction strategy based on the result to make up the defect that the model method calculates the solution-free working condition; meanwhile, the calculation time of M4 is improved by 5-6 times compared with that of M3, and the calculation efficiency is obviously improved compared with that of the current safety correction method.
3. And (3) node classification model precision test, namely evaluating a node action classification model based on SAE and comparing the node action classification model with a DNN classification model. 8000 samples are randomly drawn from the generated samples to be used as training set label samples, 4000 samples are used as test set samples, and 10000 samples are used as non-label samples. The training results of the classification models of some nodes are extracted for verification, and the results are shown in table 2.
TABLE 2 Classification model evaluation results
As can be seen from Table 2, the classification result of the SAE two-classification model is more accurate than that of the DNN model under the condition of less label samples. The SAE two-classification model is pre-trained through unsupervised training, and fine adjustment is carried out through supervised training in the second stage, so that the SAE two-classification model can automatically extract features under the condition that the label sample is fixed, dimension reduction of data is achieved, and meanwhile the problems that the label sample is difficult to obtain and the number is small in an actual calculation example are solved.
4. And (4) calculating the precision analysis of the correction value, namely, single-branch out-of-limit correction analysis and multi-branch out-of-limit correction analysis.
(1) The single branch off-limit correction analysis shows that the working conditions 1-3 are random branch off-limit working conditions with different degrees, the calculation performance of the method provided by the invention under different working conditions is respectively tested and compared with other three methods, and the method is shown in the table 3.
TABLE 3 working Condition 1 System safety correction results under different algorithms
It can be seen from table 3 that when a certain line of the system is out of limit, the number of action nodes of M2 is obviously reduced compared with that of M1, which indicates that the participation of UPFC in regulation and control can greatly reduce the number of control devices involved in the correction process and the required correction amount, and further verify the ability of the UPFC to improve the system economy; meanwhile, the calculation result of M4 under the working condition is the same as that of M2 and M3, and the data driving model has better precision.
As shown in Table 4, the safety correction results under condition 2 are given for comparison;
TABLE 4 working conditions under different algorithms 2 System safety correction results
As can be seen from table 4, in the correction strategy given by M4, the generator active power correction amount is slightly different from that of other methods, and the result of solving the objective function 4632.01 of M4 is found to be slightly larger than 4628.98 of M2 by substituting the correction value into the solved objective function, so that the strategy obtained by M4 is not optimal, but is very close to the optimal solution; compared with the load shedding amount of different methods, the load shedding amount of the system can be seen easily, the correction strategy given by M4 only needs to adjust the output of the generator, so that the load shedding action is effectively avoided, the safety correction priority adjustment of the generator is more met, the load shedding requirement is avoided, and the economic benefit of the system is improved.
As shown in Table 5, the system safety correction results under the working conditions 3 under different algorithms are given;
TABLE 5 working Condition 3 System safety correction results under different algorithms
In table 5, although the calculation result of M4 is slightly different from that of M2, the objective function result 5422.95 is slightly smaller than 5423.61 of M2, and it can be seen that the action node and the correction amount theoretically obtained by the optimization type calculation method based on the conventional model cannot be guaranteed to be an optimal solution, and the optimal solution itself also has a certain non-optimal problem.
(2) Multi-branch off-limit correction analysis
In order to verify that the method is also applicable to the situation that a plurality of branches are out of limit, the load is randomly adjusted, the system branches are overloaded, different load levels can be set on the basis of a plurality of tidal current sections respectively, working conditions such as N-1 faults, N-2 faults and the like are met, and samples of the plurality of branches which are out of limit are obtained for retraining. For the multi-branch out-of-limit problems of the working condition 6 and the working condition 7, as shown in tables 6 and 7, calculation results of different correction methods are given.
TABLE 6 working conditions under different algorithms 6 System safety correction results
TABLE 7 working conditions under different algorithms 7 System safety correction results
It can be seen from tables 6 and 7 that, comparing the load shedding amount results of M1 and M2, the load shedding situation can be effectively alleviated or even avoided due to the addition of UPFC in the M2 system, and the stability and economic benefit of the system are improved. Meanwhile, as can be seen from comparison of the objective functions of the methods in table 4, the objective function values of M3 and M4 are the same and are both smaller than M1 and M2, which shows that the method provided by the invention better meets the requirements of safety correction than the traditional model-based optimization methods, and the provided correction strategy is more optimal. Furthermore, although the calculation results of the M4 and the M3 correction strategy are the same, as can be seen from the calculation times of the working conditions 6 and 7 in fig. 5, the calculation efficiency of the M4 is 3 times higher than that of the M3, and the online application time of the M4 is hardly fluctuated along with the increase of the system faults or the increase of the system scale.
In conclusion, compared with the conventional safety correction calculation method, the method provided by the invention not only has accuracy and applicability, can accurately solve different out-of-limit working conditions, but also has higher calculation efficiency, so that the quick safety correction method has great advantages in the aspect of guaranteeing the stability of a large system in actual engineering; simulation results show that the method provided by the invention has higher precision, and the calculation time is stable and much shorter than other methods, thereby not only meeting the requirements of system safety correction on accuracy and rapidity, but also being beneficial to a decision maker to quickly make a correction strategy, improving the system stability, effectively ensuring the safe and stable operation of a UPFC-containing system, and having important significance for fully exerting the UPFC power flow control capability.
The foregoing illustrates and describes the principles, general features, and advantages of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (6)
1. A safety correction method of a power system containing UPFC based on an SAE two-classification model is characterized by comprising the following steps: comprises the following steps of (a) carrying out,
step (A), constructing a safety correction optimization model of the power system containing the UPFC;
step (B), solving the constructed safety correction optimization model of the power system containing the UPFC, obtaining the adjustment range of a safety correction node when the power system containing the UPFC works, and recording effective out-of-limit data under different fault conditions, power flow data at the fault moment and corresponding safety correction optimization results to prepare a sample set;
step (C), an SAE two-class model for judging the correction state of the node is established and trained, the input of the SAE two-class model is the power flow data of the fault moment in the sample set and the corresponding safety correction optimization result which are manufactured in the step (B), and the output is the action state of each node in the safety correction optimization result;
step (D), based on the adjustment range of the safety correction node, establishing and training a DNN regression model for calculating a safety correction value, wherein the input of the NN regression model is effective out-of-limit data in the sample set manufactured in the step (B) and the action state of each node in the safety correction optimization result obtained in the step (C), and the action state is output as a system node safety correction adjustment value;
and (E) finishing the safety correction operation of the power system containing the UPFC by using the output system node safety correction adjustment quantity.
2. The safety correction method for the power system with the UPFC based on the SAE two-classification model as claimed in claim 1, characterized in that: step (A), constructing a safety correction optimization model of the power system containing the UPFC, as shown in a formula (1),
wherein the content of the first and second substances,
minimization of equipment for power systems with UPFCAdjusting the amount;
representing the number of system nodes;
is an integer variable 0-1 and represents whether the node i participates in the adjustment;
and
respectively representing active and reactive correction values of a node i; introducing weight coefficients
The method is used for improving the priority of node correction state calculation and meeting the safety correction demand; introducing node correction state weights
And correction amount weight
The method is used for ensuring that the generator nodes are preferentially adjusted in the correction process and avoiding the action of the load nodes, and the weighted values of the generator nodes are all smaller than the load nodes.
3. The safety correction method for the power system with the UPFC based on the SAE two-classification model as claimed in claim 2, characterized in that: step (B), the constructed security correction optimization model of the power system containing the UPFC is solved, the adjustment range of the security correction node when the power system containing the UPFC works is obtained, effective out-of-limit data under different fault conditions, power flow data at fault moments and corresponding security correction optimization results are recorded to prepare a sample, wherein the specific steps are as follows,
step (B1), solving the constructed safety correction optimization model of the power system containing the UPFC, as shown in formulas (2) and (3),
wherein, superscript denotes conjugation;
and
respectively the voltage of the head end and the tail end of the serial side of the branch where the UPFC is located;
representing the virtual node voltage introduced by the serial side of the UPFC;
and
and
and
respectively representing the conductance and sodium conductivity of the circuit on the serial side and the parallel side of the UPFC;
and
and
and
respectively representing initial values of the node generator and the load;
and
respectively representing the real part and the imaginary part of the element at the position i-j in the admittance matrix;
representing the phase angle difference between the node i and the node j;
step (B2), obtaining the adjustment range of the safety correction node when the power system containing the UPFC works and operates, as shown in formulas (4), (5) and (6),
wherein, the first and the second end of the pipe are connected with each other,
and
respectively representing the upper limit values of the capacities of the serial and parallel sides of the UPFC;
；
；
and
and
and
respectively representing the upper limit and the lower limit of the adjustable active power and the adjustable reactive power of the node i;
characterize the active power of the line i-j, and
and
respectively, its upper and lower limits;
and (B3) recording effective out-of-limit data under different fault conditions, tidal current data at fault time and corresponding safety correction optimization results to prepare samples, wherein the sample generation method is to randomly adjust loads to enable the system to have heavy load and N-1 and N-2 fault conditions, obtain the effective out-of-limit data of each branch and the tidal current data at the fault time, the effective out-of-limit data is that the out-of-limit volume exceeds 1% -30% of the self-stability limit value of the line, then, the out-of-limit conditions of the line are optimized and solved by a method for solving an optimization model, the safety correction optimization results of each node under the corresponding fault are obtained, and the safety correction optimization results comprise correction states and correction values.
4. The safety correction method for the UPFC-containing power system based on the SAE two-class model as claimed in claim 3, characterized in that: step (C), an SAE two-classification model for judging the correction state of the node is established and trained, the input of the SAE two-classification model is the load flow data of the fault moment in the sample set and the corresponding safety correction optimization result made in step (B), and the output is the action state of each node in the safety correction optimization result, wherein the specific steps are as follows,
step (C1), an SAE two-classification model for judging the node correction state is established, wherein the SAE two-classification model comprises an activation function and a classifier, the activation function selects a Sigmoid function, the top classifier adopts a Softmax regression layer, and the activation function is used for performing layer-by-layer feature extraction on high-dimensional input data and inputting the high-dimensional input data into the top classifier to finish classification;
step (C2), training the established SAE two-classification model, wherein the training of the SAE two-classification model comprises two stages of pre-training and fine-tuning, the concrete steps are as follows,
step (C21), in the pre-training stage, using a label-free sample to perform unsupervised training, initializing each layer parameter of the SAE two-classification model, and converting high-dimensional input into low-dimensional hidden layer feature expression;
and (C22) in the fine adjustment stage, supervised training is carried out by using the samples with the labels, and the parameters of the whole network are fine adjusted.
5. The safety correction method for the power system with the UPFC based on the SAE two-classification model as claimed in claim 4, characterized in that: step (D), based on the adjustment range of the safety correction node, establishing and training a DNN regression model for calculating a safety correction value, wherein the input of the NN regression model is effective out-of-limit data in a sample set manufactured in the step (B), the action state of each node in the safety correction optimization result obtained in the step (C) is output as a safety correction adjustment quantity of the system node, and the specific steps are as follows,
step (D1), a DNN regression model for calculating a safety correction value is established, wherein the DNN regression model shares six layers and comprises an input layer, an output layer and four hidden layers, the number of neurons of the hidden layers is 1024, 512, 512 and 512, and the activating functions are non-saturated and non-linear ReLU functions;
and (D2) training the constructed DNN regression model, wherein the DNN regression model is obtained by sending input data into a network according to batches, calculating layer by layer forwards until an output layer, calculating the value of a loss function, reversely propagating and updating the parameters of the network to achieve the optimal characterization capability of the model, the training process is off-line training, and the correction value of the DNN regression model is determined by using the learning capability of the DNN model aiming at action nodes.
6. The safety correction method for the power system with the UPFC based on the SAE two-classification model as claimed in claim 4, characterized in that: the labeled sample in the step (C2) is characterized by system flow data under fault conditions, and can be generated by solving a safety correction model under each fault condition by using sample data labeled whether a node acts or not; the label-free sample only needs system flow data under fault conditions, labels whether nodes act or not are not needed, and the label-free sample can be generated through flow calculation under each fault condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211265207.2A CN115329911A (en) | 2022-10-14 | 2022-10-14 | Safety correction method for UPFC-containing power system based on SAE two-classification model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211265207.2A CN115329911A (en) | 2022-10-14 | 2022-10-14 | Safety correction method for UPFC-containing power system based on SAE two-classification model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115329911A true CN115329911A (en) | 2022-11-11 |
Family
ID=83915343
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211265207.2A Pending CN115329911A (en) | 2022-10-14 | 2022-10-14 | Safety correction method for UPFC-containing power system based on SAE two-classification model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115329911A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115632406A (en) * | 2022-12-21 | 2023-01-20 | 国网天津市电力公司城东供电分公司 | Reactive voltage control method and system based on digital-mechanism fusion drive modeling |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111985091A (en) * | 2020-07-29 | 2020-11-24 | 河海大学 | Rapid safety correction method for power system containing UPFC |
-
2022
- 2022-10-14 CN CN202211265207.2A patent/CN115329911A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111985091A (en) * | 2020-07-29 | 2020-11-24 | 河海大学 | Rapid safety correction method for power system containing UPFC |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115632406A (en) * | 2022-12-21 | 2023-01-20 | 国网天津市电力公司城东供电分公司 | Reactive voltage control method and system based on digital-mechanism fusion drive modeling |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110535146B (en) | Electric power system reactive power optimization method based on depth determination strategy gradient reinforcement learning | |
| CN102074955B (en) | Method based on knowledge discovery technology for stability assessment and control of electric system | |
| CN110542819B (en) | Transformer fault type diagnosis method based on semi-supervised DBNC | |
| CN110994604A (en) | Electric power system transient stability evaluation method based on LSTM-DNN model | |
| CN105938578A (en) | Large-scale photovoltaic power station equivalent modeling method based on clustering analysis | |
| CN112330020B (en) | Collaborative optimization method for electricity-gas comprehensive energy system | |
| CN114006370B (en) | Power system transient stability analysis and evaluation method and system | |
| CN113300380B (en) | Load curve segmentation-based power distribution network reactive power optimization compensation method | |
| CN112330165A (en) | Power grid transient stability evaluation method and system based on feature separation type neural network | |
| CN112287605B (en) | Power flow checking method based on graph convolution network acceleration | |
| CN115329911A (en) | Safety correction method for UPFC-containing power system based on SAE two-classification model | |
| CN104899796A (en) | Method for optimizing LVQ neutral network based on particle swarm, and disturbance and harmonic wave detection methods | |
| CN117874615A (en) | Photovoltaic fault diagnosis method and system based on depth digital twin | |
| CN111985091A (en) | Rapid safety correction method for power system containing UPFC | |
| CN116227320A (en) | Double-fed fan control parameter identification method based on LSTM-IPSO | |
| CN114552653B (en) | Grid-connected inverter system control parameter multi-objective optimization method based on intelligent algorithm | |
| CN116306372A (en) | Electric-gas area comprehensive energy system safety correction control method based on DDPG algorithm | |
| Fu et al. | Fault diagnosis of photovoltaic array based on SE-ResNet | |
| CN116204771A (en) | Power system transient stability key feature selection method, device and product | |
| CN113991652B (en) | Data-driven multi-output calculation method for short-circuit current of distribution network containing IIDG | |
| CN116305683A (en) | Power system transient stability evaluation method and system based on sample equalization | |
| CN114139809A (en) | Dynamic frequency response curve prediction method and system based on XGboost power system | |
| CN108988319B (en) | Rapid emergency control method based on deep feedforward neural network and numerical integration sensitivity | |
| CN103838140B (en) | Weak nonlinear network control method based on direct inverse control algorithm | |
| CN112651628A (en) | Power system transient stability evaluation method based on capsule 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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221111 |
|
| RJ01 | Rejection of invention patent application after publication |