CN116611011A - Transient overvoltage amplitude prediction method, system, terminal and medium - Google Patents
Transient overvoltage amplitude prediction method, system, terminal and medium Download PDFInfo
- Publication number
- CN116611011A CN116611011A CN202310560067.XA CN202310560067A CN116611011A CN 116611011 A CN116611011 A CN 116611011A CN 202310560067 A CN202310560067 A CN 202310560067A CN 116611011 A CN116611011 A CN 116611011A
- Authority
- CN
- China
- Prior art keywords
- data set
- training
- decision tree
- prediction
- model
- 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
- 230000001052 transient effect Effects 0.000 title claims abstract description 73
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000012549 training Methods 0.000 claims abstract description 90
- 238000003066 decision tree Methods 0.000 claims abstract description 82
- 238000004364 calculation method Methods 0.000 claims abstract description 25
- 230000008569 process Effects 0.000 claims abstract description 24
- 230000004044 response Effects 0.000 claims abstract description 11
- 238000012360 testing method Methods 0.000 claims description 26
- 238000012545 processing Methods 0.000 claims description 23
- 238000004590 computer program Methods 0.000 claims description 20
- 230000000694 effects Effects 0.000 claims description 20
- 238000010606 normalization Methods 0.000 claims description 18
- 238000004088 simulation Methods 0.000 claims description 8
- 238000011156 evaluation Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000012886 linear function Methods 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 abstract description 2
- 230000005540 biological transmission Effects 0.000 description 15
- 230000006870 function Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 229910052799 carbon Inorganic materials 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 230000006698 induction Effects 0.000 description 3
- 230000035699 permeability Effects 0.000 description 3
- 101000912561 Bos taurus Fibrinogen gamma-B chain Proteins 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000006386 neutralization reaction Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000013442 quality metrics Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the technical field of power systems, and discloses a method, a system, a terminal and a medium for predicting transient overvoltage amplitude, wherein in a sample data set generation stage, modeling and manual extraction of input characteristics are not needed, and only partial real-time response to electric quantity at the occurrence time of a fault is needed, so that the transient overvoltage amplitude of a transmitting end can be directly predicted; in the model training stage, the characteristic that the higher the overvoltage level is, the lower the tolerance of calculation errors is, and the training target of the traditional decision tree is improved, so that the weight of the prediction error of the high-risk scene is increased, and the attention to the accuracy of the prediction of the overvoltage amplitude of the high-risk scene is increased; in the prediction stage, compared with the traditional decision tree model, the improved model provided by the invention has higher prediction accuracy, the result is closer to a true value, the fan off-grid risk in the disturbance process can be more effectively estimated, and the guidance is provided for formulating reasonable emergency control measures.
Description
Technical Field
The invention relates to the technical field of power systems, in particular to a method, a system, a terminal and a medium for predicting transient overvoltage amplitude.
Background
In China, the energy resources and the load center are reversely distributed. Under the promotion of the 'double carbon' targets of carbon neutralization and carbon peak, new energy sources such as wind power, photovoltaic and the like are increased in a spanning manner, the power patterns of 'west electric east transmission' and 'north electric south transmission' are formed in China as a whole, and the extra-high voltage direct current transmission project is widely applied due to the advantages of large capacity and long-distance transmission. When the direct current fails to commutate, the direct current is blocked, and the direct current power transmission is blocked, a large amount of reactive power surplus of the alternating current filter matched with the rectifying station is injected into the power grid, so that the transient overvoltage is generated on the alternating current bus at the transmitting end. Particularly, for a system with a large number of fans connected to a feed port, once transient voltage rises above a fan voltage-resistant value, large-scale interlocking off-grid of the fans is caused, and serious system safety and stability problems are caused. In the running process of the power grid, the problem of transient overvoltage at the transmitting end has become a direct influence factor for restricting the direct current power transmission capacity and the internet power of a wind power plant. Therefore, the method accurately analyzes the transient overvoltage amplitude of the sending end, and has important guiding significance for effectively evaluating and reducing the off-grid risk of the fan in the disturbance process and making reasonable emergency control measures to maintain safe and stable operation of the system.
In the prior art, the amplitude of the transient overvoltage at the transmitting end is mainly analyzed based on a physical mechanism model, and can be roughly divided into two types. The method has certain physical significance, is simple and convenient to calculate, has low accuracy and is only suitable for preliminary qualitative analysis. The other type is transient overvoltage amplitude quantitative research based on detailed modeling, and the sending-end transient overvoltage is solved by reasonably performing equivalent modeling on an alternating-current/direct-current series-parallel system and considering related operation characteristics. The modeling quantitative analysis method has firm mathematical model foundation, improves the calculation accuracy to a certain extent, but the calculation result is closely connected with the accuracy of the model, a large number of algebraic equations and nonlinear equations are required to be solved in the calculation process, and the calculation speed is low. When the system scale is increased, the dimension is increased continuously, the calculation time required by solving is increased sharply, the requirement of transient voltage calculation in the existing power system cannot be met, and the equivalent simplified modeling can also have a certain influence on the calculation accuracy.
With the popularization and application of Wide Area Measurement Systems (WAMS) based on millisecond synchronous measurement in power systems, transient overvoltage amplitude prediction methods based on response and artificial intelligence have been developed to a certain extent. According to the real-time response characteristic quantity acquired by WAMS, the transient overvoltage is predicted based on an artificial intelligence method, such as fan transient overvoltage amplitude prediction based on a decision tree, transient overvoltage prediction based on an online sequence extreme learning machine and the like. The method based on response and artificial intelligence does not need complex mathematical modeling, has higher estimation precision, can be used for predicting the transient overvoltage amplitude of the sending end more quickly and effectively so as to better evaluate and reduce the off-grid risk of the fan, and has a certain improvement on the calculation efficiency and the calculation precision compared with the mechanism method. However, in the actual operation process of the power grid, the higher the overvoltage level of the fan machine end is, the larger the influence on the safe operation of the system is, and the lower the tolerance on the error of the calculation result is. The existing artificial intelligent prediction method is classical, the specific actual scene characteristics of the transient overvoltage of the sending end are not considered enough, and the prediction accuracy of the model still need to be improved.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention aims to provide a method, a system, a terminal and a medium for predicting transient overvoltage amplitude, so as to solve the technical problems of insufficient consideration of specific actual scene characteristics of the transient overvoltage of a transmitting end and low prediction accuracy and prediction accuracy of a model in the prior art.
The invention is realized by the following technical scheme:
a method for predicting transient overvoltage amplitude comprises the following steps:
step 1, generating a sample data set, carrying out normalization processing on the sample data set, and dividing the sample data set into a training data set and a test data set;
step 2, adopting a decision tree model, improving a training target of the decision tree model, and training the improved decision tree model by using a training data set;
and 3, predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect, and completing the prediction work of the transient overvoltage amplitude.
Preferably, in step 1, the generation process of the sample data set is as follows:
the method comprises the steps of selecting active power, reactive power, machine side voltage, machine side current, converter bus voltage and electric quantity of system frequency of a fan at fault moment as input characteristics of a prediction model, selecting transient overvoltage amplitude after fault as output of the prediction model, generating an initial sample data set, and forming the sample data set by utilizing historical response electric quantity data acquired by a WAMS system.
Preferably, in step 1, the sample data is normalized by using a linear function conversion method, where the expression is as follows:
wherein x is i The actual value of a certain electric quantity in the input characteristics; x is x max And x min Maximum and minimum values for the electrical quantity in all samples; y is i The normalization processing result is obtained.
Preferably, in step 2, the decision tree model is based on the set Sample i Dividing a certain element in the model, and determining dividing attributes and dividing standards gamma by calculating purity loss after sample division so as to obtain a well-trained decision tree model;
wherein Sample is assembled i Is Sample i ＝{f WT1 ,V WT1 ,I WT1 ,P WT1 ,Q WT1 ,…,f WTn ,V WTn ,I WTn ,P WTn ,Q WTn }；
Wherein n is the number of fans; subscript WTn is the nth fan; f (f) WTn ,V WTn ,I WTn ,P WTn ,Q WTn The frequency, the machine end voltage, the machine end current, the active power and the reactive power of the nth fan at the fault moment are respectively.
Preferably, in step 2, the calculation formula for improving the training target of the decision tree model is as follows:
wherein t is a node in the regression tree, R' (t) is a variance of the improved regression tree at the t node, and N is the number of samples corresponding to the node t; y is i Fan overvoltage amplitude obtained by simulation for a certain sample;the average value of the overvoltage amplitude values corresponding to all the samples in the node t; ΔR' (t) is the decrease in variance of the improved decision regression tree at the t-th node; r' (t) R )、R′(t L ) The variances of the right subtree and the left subtree of the improved regression tree after division are respectively; n (N) R 、N L The number of samples of the divided right subtree and left subtree respectively; r (t) is the variance of the conventional regression tree at the t-th node.
Preferably, in step 2, the improved decision tree model is trained by using a training data set, and a gradient descent method is adopted in the training process until the dividing condition of the training sample meets the purity requirement, so as to finally obtain the improved decision tree model with perfect training, wherein the dividing condition is gamma '' max Satisfies the following formula:
γ′ max ＝argmaxΔR′(γ,t)
in the formula, argmax is a variable γ corresponding to a maximum value of the variance reduction amount, that is, the purity loss Δr' (γ, t).
Preferably, in step 3, the trained decision tree model predicts the overvoltage of the test data set, and reflects the regression prediction effect of the model through the prediction accuracy index, and the evaluation index has the following calculation formula:
A S ＝(1-RMSE)×100％
where RMSE represents root mean square error; a is that S The prediction accuracy is obtained; n is the number of fans; y' (i) and y (i) are respectively a transient overvoltage predicted value and a true value of the ith fan machine end bus.
A system for predicting transient overvoltage magnitude, comprising:
the data set processing module is used for generating a sample data set, carrying out normalization processing on the sample data set and dividing the sample data set into a training data set and a test data set;
the model training module is used for adopting the decision tree model, improving the training target of the decision tree model and training the improved decision tree model by utilizing the training data set;
the data prediction module is used for predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect and completing the prediction work of the transient overvoltage amplitude.
A mobile terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of a method of predicting a transient overvoltage magnitude as described above when the computer program is executed.
A computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of a method of predicting a transient overvoltage magnitude as described above.
Compared with the prior art, the invention has the following beneficial technical effects:
the invention provides a method for predicting transient overvoltage amplitude, which is characterized in that in a sample data set generation stage, modeling and manual extraction of input characteristics are not needed, and only partial real-time response electric quantity at the occurrence time of a fault is needed, the transient overvoltage amplitude of a transmitting end can be directly predicted; in the model training stage, the characteristic that the higher the overvoltage level is, the lower the tolerance of calculation errors is, and the training target of the traditional decision tree is improved, so that the weight of the prediction error of the high-risk scene is increased, and the attention to the accuracy of the prediction of the overvoltage amplitude of the high-risk scene is increased; in the prediction stage, compared with the traditional decision tree model, the improved model provided by the invention has higher prediction accuracy, the result is closer to a true value, the fan off-grid risk in the disturbance process can be more effectively estimated, and the guidance is provided for formulating reasonable emergency control measures.
Drawings
FIG. 1 is a flow chart of a method for predicting transient overvoltage amplitude in the present invention;
FIG. 2 is a schematic diagram of an overvoltage regression prediction decision tree structure of a single fan in an embodiment of the present invention;
FIG. 3 is a schematic diagram of an AC/DC hybrid power grid power transmission end system containing large-scale wind power in an embodiment of the invention;
FIG. 4 is a schematic diagram of the voltage of the machine side bus of the blower in the case of a DC commutation failure fault in an embodiment of the present invention;
FIG. 5 is a schematic diagram of a training process for improving a decision tree in an embodiment of the present invention;
FIG. 6 is a regression prediction result of the transient overvoltage amplitude of the transmitting end in the embodiment of the invention;
FIG. 7 is a graph showing the relative error of the low risk interval over-voltage amplitude prediction in accordance with an embodiment of the present invention;
FIG. 8 is a graph showing the relative error of the prediction of the overvoltage amplitude in the high risk interval according to the embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
The invention is described in further detail below with reference to the attached drawing figures:
the invention aims to provide a method, a system, a terminal and a medium for predicting transient overvoltage amplitude, which are used for solving the technical problems of insufficient consideration of specific actual scene characteristics of the transient overvoltage of a transmitting end and low prediction accuracy and prediction accuracy of a model in the prior art.
Specifically, according to fig. 1, the method for predicting the transient overvoltage amplitude includes the following steps:
firstly, generating a sample data set by using time domain transient stability simulation software, or acquiring historical data of a power grid by using a WAMS system as the sample data set, carrying out normalization processing on the sample data set by adopting a normalization method, and dividing the sample set into a training data set and a test data set.
And taking various operation scenes such as different new energy permeabilities, direct current transmission power, load levels, induction motor duty ratio and the like of the power grid into consideration, combining different fault types, fault positions, fault duration and other factors, carrying out off-line simulation by using simulation software based on typical alternating current-direct current series-parallel scenes, collecting key response electric quantity data at fault moment, and reconstructing the collected data according to the input and output requirements of a model to generate a sample data set. Aiming at a transient overvoltage amplitude prediction scene that a sending end contains a large-scale fan, a plurality of electric quantities such as active power, reactive power, machine end voltage, machine end current, converter bus voltage, system frequency and the like of the fan at the moment of failure are generally selected as input characteristics of a prediction model, the operation characteristics of a system are better represented, the transient overvoltage amplitude after failure is selected as output of the prediction model, and therefore an initial sample data set is generated. For an actual power grid, the historical response electric quantity data acquired by the WAMS system can be utilized to form a sample data set.
After the initial sample data set is generated, sample data needs to be normalized so as to adapt to the requirement of the decision tree model on the input range, and meanwhile, the training time of the model is shortened by reducing the data range, so that the convergence of the model is accelerated, and the model prediction precision is improved. The invention adopts a linear function conversion method to carry out normalization processing, and the expression is as follows:
wherein x is i The actual value of a certain electric quantity in the input characteristics; x is x max And x min Maximum and minimum values for the electrical quantity in all samples; y is i The normalization processing result is obtained.
The sample set after normalization treatment can be input into the built model for training after reasonable division, and is usually divided into a training data set and a test data set according to a certain proportion. The training set is a data sample for model fitting, and the training set is used for determining built-in parameters of neurons; the test set is used for evaluating the final generalization capability of the model, and is only used for evaluating the quality of the model, and is not used as a selection basis related to algorithms such as parameter adjustment, feature selection and the like.
Secondly, adopting a decision tree, improving a training target, namely a loss function of a back propagation algorithm, and training an improved decision tree model by using a training data set; in the training process, aiming at the characteristic that the tolerance of the high-risk scene to the error of the overvoltage amplitude calculation result is low, the purity calculation formula in the training target is corrected so as to improve the weight of the prediction error of the high-risk scene.
The structure of the over-voltage amplitude regression prediction decision tree of the single fan is shown in figure 2. The regression tree training samples consisted of: the input characteristics are the frequency, voltage, current, active power and reactive power of n fans at the occurrence time of faults, the output label is the overvoltage amplitude obtained by adopting simulation software as shown in the following formula (2).
Sample i ＝{f WT1 ,V WT1 ,I WT1 ,P WT1 ,Q WT1 ,…,f WTn ,V WTn ,I WTn ,P WTn ,Q WTn }(2)
Wherein n is the number of fans; subscript WTn is the nth fan; f (f) WTn ,V WTn ,I WTn ,P WTn ,Q WTn The frequency, the machine end voltage, the machine end current, the active power and the reactive power of the nth fan at the fault moment are respectively.
In the training process, the decision tree is according to the set Sample i The method comprises the steps of dividing a certain element, determining dividing attributes and dividing standards gamma by calculating purity loss after sample division, obtaining a regression tree with perfect training, and further carrying out overvoltage amplitude prediction.
The training process of the traditional regression decision tree is as follows: when the regression tree is trained and constructed, the division attribute and the division standard gamma thereof are determined by calculating the purity loss after the sample division, and the purity loss is calculated based on the node variance. Let t be a node in the regression tree, the variance of that node be
Wherein R (t) is the variance of the traditional regression tree at the t-th node, and N is the number of samples corresponding to the node t; y is i For a certain purposeFan overvoltage amplitude obtained by simulation of a sample;the average value of the overvoltage amplitude values corresponding to all the samples in the node t. The loss of sample purity after division by node t is characterized by a decrease in variance Δr, also known as a branch quality metric.
Wherein R (t) R )、R(t L ) Variance of right subtree and left subtree of traditional regression tree after division respectively, N R 、N L The sample numbers of the right subtree and the left subtree after division are respectively, R (t) is the variance of the traditional regression tree at the t-th node, and N is the sample number corresponding to the node t.
In order to make the division purity of the sample high, a division condition gamma should be selected max The purity loss Δr (γ, t) is trained to be as large as possible, that is, to satisfy the following expression.
γ max ＝argmaxΔR(γ,t) (5)
In the formula, argmax is a variable γ corresponding to a maximum value of Δr (γ, t) which is a variance reduction amount.
In determining the dividing condition gamma max And then, continuously solving the dividing condition of the child node of the node t until the division of the training sample meets the purity requirement.
From the above process, it can be seen that the branch quality measurement index is an important basis for training and generating the decision tree. In the actual power system operation, the higher the overvoltage level of the fan machine end is, the larger the influence on the safe operation of the system is, and the lower the tolerance on the error of the calculation result is. The traditional decision tree method treats errors of different samples equally in the training process, and cannot show attention to accuracy of high-risk scenes. Therefore, the decision tree method needs to be improved, so that the decision tree method has higher accuracy for scenes with higher safety operation risks, and the actual requirements of engineering are met. Based on the method, a purity calculation formula in the branch quality measurement index is modified to improve the weight of the prediction error of the high-risk scene, and the modified purity and branch quality calculation formula is as follows:
wherein R' (t) is the variance of the improved regression tree at the t-th node, and N is the number of samples corresponding to the node t; y is i Fan overvoltage amplitude obtained by simulation for a certain sample;the average value of the overvoltage amplitude values corresponding to all the samples in the node t; ΔR' (t) is the decrease in variance of the improved decision regression tree at the t-th node; r' (t) R )、R′(t L ) The variances of the right subtree and the left subtree of the improved regression tree after division are respectively; n (N) R 、N L The number of samples of the divided right subtree and left subtree respectively; r (t) is the variance of the conventional regression tree at the t-th node.
From the improved model expression, it can be seen that the samples with smaller output results have larger error weight, and in the case of high risk scenes, namely, larger overvoltage amplitude, the purity loss is larger, so that the high risk scenes are better classified and predicted, the errors of the high risk samples are limited, and the introduction of risk difference knowledge in the overvoltage problem is realized.
Training an objective function based on an improved decision tree, so that the improved purity loss is as large as possible, training by adopting a gradient descent method in the training process until the division of training samples meets the purity requirement, and finally obtaining an improved decision tree model with perfect training and a division condition gamma '' max The following formula is satisfied.
γ′ max ＝argmaxΔR′(γ,t) (8)
In the formula, argmax is a variable γ corresponding to a maximum value of the variance reduction amount, that is, the purity loss Δr' (γ, t).
And finally, predicting the overvoltage amplitude of the test data set by adopting a well-trained improved decision tree model to obtain a transient overvoltage prediction result, and evaluating the model prediction effect by taking the prediction accuracy as an evaluation index.
When the improved decision tree model is used for predicting the transient overvoltage amplitude of the sending end, the regression prediction effect of the model is reflected through the prediction accuracy index, and the evaluation index has the following calculation formula:
A S ＝(1-RMSE)×100％ (10)
where RMSE represents root mean square error; a is that S The prediction accuracy is obtained; n is the number of fans; y' (i) and y (i) are respectively a transient overvoltage predicted value and a true value of the ith fan machine end bus. The index can reflect the overall effect of transient overvoltage amplitude prediction, and the higher the prediction accuracy is, the better the model prediction effect is.
In the actual power grid operation process, based on response electric quantity acquired by the WAMS system in real time, the transmission end transient overvoltage amplitude can be predicted by utilizing a well-trained improved decision tree model, so that a foundation is laid for the establishment and implementation of follow-up related control measures.
Examples
A schematic diagram of an AC/DC series-parallel power grid power transmission end system containing large-scale wind power is shown in FIG. 3, wherein the system reference capacity is 100MW, +/-800 kV Qinghai new energy base and northwest power grid are connected by the AC/DC series-parallel power grid power transmission end system, a large-scale wind power generation system is contained, a Henan load center and a Huazhong power grid are connected by a receiving end system, and when AC/DC faults with different severity degrees such as commutation failure, DC blocking, short circuit fault and the like occur near Qinghai-Yuan DC, transient overvoltage problems can occur in the power transmission end system. After the primary commutation failure fault occurs in the direct current receiving end system, the voltage of the bus of the machine end of the blower is shown as figure 4, the transient voltage of the machine end of the direct current near-area new energy exceeds 1.3p.u., and the problem of serious transient overvoltage exists.
Firstly, generating a sample data set by using time domain transient stability simulation software, carrying out normalization processing on the sample data set by adopting a normalization method, and dividing the sample set into a training data set and a test data set.
And comprehensively considering factors such as different load levels, new energy permeability, direct current transmission power, induction motor duty ratio, fault position, fault duration and the like, selecting machine end voltage, machine end current, active power, reactive power and system frequency of each fan at the sending end at the moment of fault occurrence as input, selecting sending end overvoltage amplitude in the transient process as output, and carrying out simulation by using simulation software PSASP to generate a sample set. The load levels were set to 90%, 100% and 110% of the rated level, the direct current transmission power was set to 60% -100% of the rated transmission power in 10% steps, the new energy permeability was set to 30%, 50% and 80%, and the induction motor load ratios were set to 50% and 65%. The faults mainly set direct current commutation failure faults, the direct current commutation failure times are set to be 1, 2 and 3 times, and the single commutation failure duration is set to be 0.15s, 0.2s and 0.25s. The simulation time was 10s, the rated frequency of the system was 50Hz, and the total number of samples was 4480.
And carrying out normalization processing on the sample data set by adopting a linear function conversion method, selecting 80% of data of the sample set as a training set, and the rest 20% of data as a test set, wherein the training set is used for model fitting data samples, and the test set evaluates the final generalization capability of the model.
Secondly, a decision tree is adopted, a training target, namely a loss function of a back propagation algorithm, is improved, and the improved decision tree model is trained by utilizing a training data set. In the training process, aiming at the characteristic that the tolerance of the high-risk scene to the error of the overvoltage amplitude calculation result is low, the purity calculation formula in the training target is corrected so as to improve the weight of the prediction error of the high-risk scene.
As described above, in order to raise the attention to the accuracy of the prediction of the overvoltage amplitude of the high-risk scene, the branch quality measurement index of the conventional decision tree is modified to raise the weight of the prediction error of the high-risk scene, and the calculation formulas of the purity and the branch quality after modification are as follows:
wherein R' (t) is the variance of the improved regression tree at the t-th node, and N is the number of samples corresponding to the node t; y is i Fan overvoltage amplitude obtained by simulation for a certain sample;the average value of the overvoltage amplitude values corresponding to all the samples in the node t; ΔR' (t) is the decrease in variance of the improved decision regression tree at the t-th node; r' (t) R )、R′(t L ) The variances of the right subtree and the left subtree of the improved regression tree after division are respectively; n (N) R 、N L The number of samples of the divided right subtree and left subtree respectively; r (t) is the variance of the conventional regression tree at the t-th node.
Based on improved decision tree training objective function, dividing condition gamma' max The training should be performed with as great a loss of purity as possible, and the training process is performed by gradient descent until the purity requirement is met by the division of training samples, and the training process of the improved decision tree is shown in fig. 5.
γ′ max ＝argmaxΔR′(γ,t) (13)
In the formula, argmax is a variable γ corresponding to a maximum value of the variance reduction amount, that is, the purity loss Δr' (γ, t).
And finally, predicting the overvoltage amplitude of the test data set by adopting a well-trained improved decision tree model to obtain a transient overvoltage prediction result, and evaluating the model prediction effect by taking the prediction accuracy as an evaluation index.
Comparing the regression prediction result of the improved decision tree model with the traditional decision tree model, and predicting transient overvoltage amplitude results by adopting different models is shown in fig. 6. The result shows that the regression prediction effect of the proposed evaluation model is relatively good, and the predicted value is closer to the true value. In addition, fig. 7 and 8 show the relative error results of different model predicted overvoltage magnitudes for the low risk interval and the high risk interval, respectively. It can be seen that the improved decision tree model shows a better regression prediction effect in both the low risk interval and the high risk interval, and the error is relatively small; meanwhile, by comparing risk intervals of different degrees, the improved decision tree model is found to have more obvious improvement on the prediction effect of the traditional decision tree model in a high risk interval, namely, in a scene with larger overvoltage amplitude, and has lower relative error, so that the effectiveness of the improved decision tree model is further verified, and the attention to the accuracy of the prediction of the overvoltage amplitude of the high risk scene is improved.
The average prediction accuracy of the different models was calculated using the evaluation index, and the results are shown in table 1.
Table 1 average prediction accuracy before and after improvement of decision tree model
As can be seen from Table 1, the regression prediction accuracy of the model is 98.1627%, which is higher than that of the traditional decision tree model, and further proves that the regression prediction effect of the improved decision tree model is relatively good.
In the actual power grid operation process, based on response electric quantity acquired by the WAMS system in real time, the transient overvoltage amplitude of the transmitting end can be predicted by utilizing a well-trained improved decision tree model, so that the fan off-grid risk in the disturbance process is effectively estimated, and important guidance is provided for formulating reasonable emergency control measures.
The invention also provides a prediction system of the transient overvoltage amplitude, which comprises a data set processing module, a model training module and a data prediction module;
the data set processing module is used for generating a sample data set, carrying out normalization processing on the sample data set and dividing the sample data set into a training data set and a test data set;
the model training module is used for adopting the decision tree model, improving the training target of the decision tree model and training the improved decision tree model by utilizing the training data set;
the data prediction module is used for predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect and completing the prediction work of the transient overvoltage amplitude.
The invention also provides a mobile terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, such as a prediction program of the transient overvoltage magnitude.
The steps of the method for predicting the transient overvoltage amplitude are implemented when the processor executes the computer program, for example:
generating a sample data set, carrying out normalization processing on the sample data set, and dividing the sample data set into a training data set and a test data set;
adopting a decision tree model, improving a training target of the decision tree model, and training the improved decision tree model by using a training data set;
and predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect, and completing the prediction work of the transient overvoltage amplitude.
Alternatively, the processor may implement functions of each module in the above system when executing the computer program, for example: the data set processing module is used for generating a sample data set, carrying out normalization processing on the sample data set and dividing the sample data set into a training data set and a test data set;
the model training module is used for adopting the decision tree model, improving the training target of the decision tree model and training the improved decision tree model by utilizing the training data set;
the data prediction module is used for predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect and completing the prediction work of the transient overvoltage amplitude.
The computer program may be divided into one or more modules/units, which are stored in the memory and executed by the processor to accomplish the present invention, for example. The one or more modules/units may be a series of computer program instruction segments capable of performing the specified functions, which instruction segments are used for describing the execution of the computer program in the mobile terminal. For example, the computer program may be partitioned into a data set processing module, a model training module, and a data prediction module, each module functioning specifically as follows:
the data set processing module is used for generating a sample data set, carrying out normalization processing on the sample data set and dividing the sample data set into a training data set and a test data set;
the model training module is used for adopting the decision tree model, improving the training target of the decision tree model and training the improved decision tree model by utilizing the training data set;
the data prediction module is used for predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect and completing the prediction work of the transient overvoltage amplitude.
The mobile terminal can be a computing device such as a desktop computer, a notebook computer, a palm computer, a cloud server and the like. The mobile terminal may include, but is not limited to, a processor, memory.
The processor may be a central processing unit (CentralProcessingUnit, CPU), other general purpose processors, digital signal processors (DigitalSignalProcessor, DSP), application specific integrated circuits (ApplicationSpecificIntegratedCircuit, ASIC), off-the-shelf programmable gate arrays (Field-ProgrammableGateArray, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like, which is a control center of the mobile terminal, connecting various parts of the entire mobile terminal using various interfaces and lines.
The memory may be used to store the computer program and/or module, and the processor may implement various functions of the mobile terminal by running or executing the computer program and/or module stored in the memory and invoking data stored in the memory.
The memory may include primarily a program area and a data area, and may include high-speed random access memory, but may also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart memory card (SmartMediaCard, SMC), secure digital (SecureDigital, SD) card, flash card (FlashCard), at least one disk storage device, flash memory device, or other volatile solid state memory device.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor implements the steps of the method of predicting a transient overvoltage magnitude.
The mobile terminal integrated modules/units may be stored in a computer readable storage medium if implemented in the form of software functional units and sold or used as a stand alone product.
Based on such understanding, the present invention may implement all or part of the above-mentioned method flow, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and the computer program may implement the steps of the above-mentioned method for predicting transient overvoltage amplitude when executed by a processor.
It should be noted that the computer readable medium contains content that can be appropriately scaled according to the requirements of jurisdictions in which such content is subject to legislation and patent practice, such as in certain jurisdictions in which such content is subject to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.
Finally, it should be noted that: the above embodiments are only for illustrating the technical aspects of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.
Claims (10)
1. The method for predicting the transient overvoltage amplitude is characterized by comprising the following steps:
step 1, generating a sample data set, carrying out normalization processing on the sample data set, and dividing the sample data set into a training data set and a test data set;
step 2, adopting a decision tree model, improving a training target of the decision tree model, and training the improved decision tree model by using a training data set;
and 3, predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect, and completing the prediction work of the transient overvoltage amplitude.
2. The method for predicting transient overvoltage amplitude according to claim 1, wherein in step 1, the generation process of the sample data set is as follows:
the method comprises the steps of selecting active power, reactive power, machine side voltage, machine side current, converter bus voltage and electric quantity of system frequency of a fan at fault moment as input characteristics of a prediction model, selecting transient overvoltage amplitude after fault as output of the prediction model, generating an initial sample data set, and forming the sample data set by utilizing historical response electric quantity data acquired by a WAMS system.
3. The method for predicting transient overvoltage amplitude according to claim 1, wherein in step 1, the sample data is normalized by using a linear function conversion method, and the expression is as follows:
wherein x is i The actual value of a certain electric quantity in the input characteristics; x is x max And x min Maximum and minimum values for the electrical quantity in all samples; y is i The normalization processing result is obtained.
4. The method of claim 1, wherein step 2, the decision tree model is based on a set Sample i Dividing a certain element in the model, and determining dividing attributes and dividing standards gamma by calculating purity loss after sample division so as to obtain a well-trained decision tree model;
wherein Sample is assembled i Is Sample i ＝{f WT1 ,V WT1 ,I WT1 ,P WT1 ,Q WT1 ,…,f WTn ,V WTn ,I WTn ,P WTn ,Q WTn }；
Wherein n is the number of fans; subscript WTn is the nth fan; f (f) WTn ,V WTn ,I WTn ,P WTn ,Q WTn The frequency, the machine end voltage, the machine end current, the active power and the reactive power of the nth fan at the fault moment are respectively.
5. The method according to claim 1, wherein in step 2, the calculation formula for improving the training target of the decision tree model is as follows:
wherein t is a node in the regression tree, R' (t) is a variance of the improved regression tree at the t node, and N is the number of samples corresponding to the node t; y is i Fan overvoltage amplitude obtained by simulation for a certain sample;the average value of the overvoltage amplitude values corresponding to all the samples in the node t; ΔR' (t) is the decrease in variance of the improved decision regression tree at the t-th node; r' (t) R )、R′(t L ) The variances of the right subtree and the left subtree of the improved regression tree after division are respectively; n (N) R 、N L The number of samples of the divided right subtree and left subtree respectively; r (t) is the variance of the conventional regression tree at the t-th node.
6. The method for predicting transient overvoltage amplitude according to claim 1, wherein in step 2, training the improved decision tree model by using a training data set, wherein the training process adopts a gradient descent method until the dividing condition of the training sample meets the purity requirement, and finally obtaining the improved decision tree model with perfect training, wherein the dividing condition is gamma' max Satisfies the following formula:
γ′ max ＝argmaxΔR′(γ,t)
in the formula, argmax is a variable γ corresponding to a maximum value of the variance reduction amount, that is, the purity loss Δr' (γ, t).
7. The method for predicting transient overvoltage amplitude according to claim 1, wherein in step 3, the trained decision tree model predicts the overvoltage of the test data set, and the regression prediction effect of the model is reflected by the prediction accuracy index, and the evaluation index calculation formula is as follows:
A S ＝(1-RMSE)×100％
where RMSE represents root mean square error; a is that S The prediction accuracy is obtained; n is the number of fans; y' (i) and y (i) are respectively a transient overvoltage predicted value and a true value of the ith fan machine end bus.
8. A system for predicting the magnitude of a transient overvoltage, comprising:
the data set processing module is used for generating a sample data set, carrying out normalization processing on the sample data set and dividing the sample data set into a training data set and a test data set;
the model training module is used for adopting the decision tree model, improving the training target of the decision tree model and training the improved decision tree model by utilizing the training data set;
the data prediction module is used for predicting the overvoltage of the test data set by the trained decision tree model, evaluating the prediction effect and completing the prediction work of the transient overvoltage amplitude.
9. A mobile terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of a method for predicting a transient overvoltage magnitude according to any one of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of a method of predicting a transient overvoltage magnitude according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310560067.XA CN116611011A (en) | 2023-05-17 | 2023-05-17 | Transient overvoltage amplitude prediction method, system, terminal and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310560067.XA CN116611011A (en) | 2023-05-17 | 2023-05-17 | Transient overvoltage amplitude prediction method, system, terminal and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116611011A true CN116611011A (en) | 2023-08-18 |
Family
ID=87679375
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310560067.XA Pending CN116611011A (en) | 2023-05-17 | 2023-05-17 | Transient overvoltage amplitude prediction method, system, terminal and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116611011A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117215205A (en) * | 2023-11-09 | 2023-12-12 | 国网经济技术研究院有限公司 | DC system control parameter analysis method based on decision tree and ISS theory |
-
2023
- 2023-05-17 CN CN202310560067.XA patent/CN116611011A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117215205A (en) * | 2023-11-09 | 2023-12-12 | 国网经济技术研究院有限公司 | DC system control parameter analysis method based on decision tree and ISS theory |
| CN117215205B (en) * | 2023-11-09 | 2024-02-06 | 国网经济技术研究院有限公司 | DC system control parameter analysis method based on decision tree and ISS theory |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103746583B (en) | The bridge arm equivalent emulation mode of a kind of MMC | |
| CN109474014B (en) | Quantitative evaluation method for dual-fed wind field access power grid friendliness | |
| CN109301814B (en) | Method and system for analyzing wind power capacity of access power grid | |
| CN116611011A (en) | Transient overvoltage amplitude prediction method, system, terminal and medium | |
| CN109301870B (en) | Capacity optimization method for power electronic multi-feed-in power system | |
| CN110941909A (en) | Method and system for identifying and calculating dominance of parameters of photovoltaic power station and wind power plant | |
| CN106410862B (en) | Based on the active wind power plant single machine equivalence method for restoring slope correction | |
| CN116388177A (en) | Method, device, equipment and medium for analyzing static voltage stability of power system | |
| CN105356508B (en) | Power grid wind access evaluation system and its method based on PSD-BPA | |
| CN109600064B (en) | Method and system for calculating alternating current asymmetric fault main loop of modular multilevel converter | |
| CN114491886A (en) | General modeling method and device for active power distribution network containing multi-type distributed new energy | |
| Tremblay et al. | Real-time simulation of a fully detailed type-IV wind turbine | |
| CN103825270B (en) | A kind of power distribution network three-phase state estimates the processing method of Jacobian matrix constant | |
| CN111697607A (en) | Multi-terminal flexible direct-current transmission receiving-end power grid access method and system | |
| CN113224788B (en) | Static voltage stability evaluation method for AC/DC hybrid power grid with wind power access | |
| CN112217220B (en) | Safety and stability defense method and device for new energy source sending multi-terminal flexible direct current power grid | |
| CN118094189A (en) | Transient overvoltage feature extraction method, system, terminal and medium | |
| CN111725834B (en) | Method for establishing simulation model of doubly-fed fan converter of wind power plant | |
| CN118868027A (en) | Method, device and equipment for screening intranet topological structure of direct-current offshore wind farm | |
| CN117559419A (en) | Prevention control method and device for restraining transient overvoltage of direct current end system | |
| CN116316520A (en) | Multi-module harmonic elimination method, device and equipment of direct-current chain type energy storage system and storage medium | |
| CN115954929A (en) | Topological structure optimization method for onshore wind power direct current collection system and related device | |
| CN117875038A (en) | RSCAPD rapid modeling method and system | |
| Guo et al. | Fault diagnosis of wind turbine power electronic equipment based on som neural network | |
| CN117057227A (en) | Equivalent modeling method and related device for photovoltaic power station |
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 |