CN115411775A - Doubly-fed wind turbine control parameter identification method based on LSTM neural network - Google Patents
Doubly-fed wind turbine control parameter identification method based on LSTM neural network Download PDFInfo
- Publication number
- CN115411775A CN115411775A CN202211181791.3A CN202211181791A CN115411775A CN 115411775 A CN115411775 A CN 115411775A CN 202211181791 A CN202211181791 A CN 202211181791A CN 115411775 A CN115411775 A CN 115411775A
- Authority
- CN
- China
- Prior art keywords
- model
- neural network
- data
- value
- lstm neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 34
- 238000000034 method Methods 0.000 title claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 239000003990 capacitor Substances 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 4
- 230000002411 adverse Effects 0.000 claims description 3
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 2
- 210000004027 cell Anatomy 0.000 description 10
- 238000010586 diagram Methods 0.000 description 10
- 238000004088 simulation Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 5
- 238000005259 measurement Methods 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000010248 power generation Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/28—The renewable source being wind energy
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Power Engineering (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a method for identifying parameters of a doubly-fed fan controller based on LSTM, which comprises the steps of obtaining hardware-in-loop test data of the doubly-fed fan controller by utilizing RT-LAB, extracting characteristic quantity with high correlation by using a Person correlation coefficient method, using the characteristic quantity for neural network training, identifying control parameters of a voltage outer loop and a current inner loop, and testing feasibility, effectiveness and practicability of an algorithm by using hardware-in-loop test data. Compared with the traditional parameter identification method, the operation characteristic of the fan control system can be simulated through training historical sample data, and the measured data is input into the LSTM neural network under the condition of not operating the fan model, so that the off-line identification of the control parameters is carried out.
Description
Technical Field
The invention relates to the technical field of new energy power generation parameter identification, in particular to a double-fed fan control parameter identification method based on an LSTM neural network.
Background
Because of instability of wind power and access of high-proportion power electronic equipment of a power grid, the influence of the operating characteristics on the safety and stability of a power system is increased, and a fan needs to be subjected to simulation analysis; since the control mode and control parameters of the controller have a great influence on model simulation, parameter identification is an important mode for obtaining accurate model parameters. The grid in China has been developed into an extra-high voltage alternating current-direct current hybrid grid, and meanwhile, with the large-scale access of new energy, the running characteristics of the grid become more complex, the control difficulty is increased, and the simulation of a large grid is seriously tested. In addition, the manufacturers of the grid-connected new energy power generation equipment in China are numerous, the types of the new energy power generation equipment exceed hundreds, the differences of the equipment structures and the grid-connected characteristics of different manufacturers in different models are obvious, the key grid-connected characteristics of different types cannot be accurately simulated by adopting typical model parameter simulation, and the requirements of power grid simulation calculation are difficult to meet.
Based on the method, aiming at the problems in the existing double-fed fan control system parameter identification research, an LSTM-based double-fed fan controller parameter identification method is provided, RT-LAB is used for obtaining double-fed fan controller hardware in-loop test data, a Person correlation coefficient method is used for extracting characteristic quantity with high correlation and using the characteristic quantity for neural network training, voltage outer loop and current inner loop control parameters are identified, and feasibility, effectiveness and practicability of an algorithm are tested through hardware in-loop test data.
Disclosure of Invention
The invention provides a method for identifying the control parameters of a doubly-fed fan based on an LSTM neural network, which can accurately and reliably identify the PI parameters of a doubly-fed fan control system; in order to realize the technical effects, the technical scheme adopted by the invention is as follows:
a doubly-fed wind turbine control parameter identification method based on an LSTM neural network comprises the following steps:
s1, building an identification model according to a double-fed fan hardware-in-the-loop physical model; wherein the electrical parameters of the transformer, the filter and other elements in the identification model are consistent with the electrical parameters of the semi-physical test model.
Further, in step S1, in order to maintain the voltage of the parallel capacitors in the "back-to-back" converter to be constant, and to control the vector control of the reactive power grid voltage orientation output by the converter; the voltage of the capacitor is controlled by the d-axis component of the grid-side converter current, while the port voltage of the wind power system is controlled by the q-axis component of the grid-side converter current, the reference value for the q-axis current being usually set to 0 in order to reduce losses.
The network side controller control equation can be written as:
i dg_ref =-K p1 Δu DC +K i1 x 1 (2)
thus, it is possible to obtain:
u dg =K p2 (-K p1 Δu DC +K i1 x 1 -i dg )+K i2 x 2 +X Tg i qg (5)
u qg =K p3 (i qg_ref -i qg )+K i3 x 3 -X Tg i dg (6)
in the formula, K p1 And K i1 Proportional coefficient and integral coefficient of the voltage controller respectively; k p2 And K i2 A first group of proportionality coefficients and integral coefficients of the current inner loop controller respectively; k i3 And K i3 A second group of proportional coefficients and integral coefficients of the current inner loop controller are respectively; u. u DC And u DC_ref Respectively an actual value and a reference value of the direct current bus voltage; i all right angle dg And i qg Actual values of the grid side d and q axis currents respectively; u. u dg And u qg Actual values of the grid side d and q axis voltages respectively; x Tg Is the reactance value of a transformer connected between the converter and the grid.
Under normal operating conditions, i qg_ref Normally 0, but in fault conditions, i qg_ref The method is directly given by a low-penetration control module in the following way:
wherein k is the reactive current support coefficient, U N Rated voltage for fan grid-connected point, I N The rated current of the grid-side converter. When low voltage ride through occurs, the grid-side converter can send out certain reactive power, the value of the reactive current can be obtained according to the formula (7), at the moment, the q-axis current inner loop is switched to a low-ride-through control mode, and the q-axis reference value is given by a low-ride-through control module.
S2, collecting u under the voltage drop degrees of 20%, 40%, 60% and 80% respectively DC 、i dg And i qg And collecting u DC 、i dg And i qg Error value R from hardware-in-the-loop experimental data e (ii) a Using the collected values as input characteristic values and network side control parameters K p1 、K i1 、K p2 、K i2 、K i3 And K i3 Processing the data set for outputting the data set; wherein u is DC The voltage of a direct current bus of a fan controller model network side controller capacitor; i.e. i dg And i qg Dq component of the output current for the grid-side converter; the method specifically comprises the following steps:
s201, outputting a mat file by using a Python calling model, and collecting a characteristic-output data set, wherein an input characteristic value is a direct-current bus voltage u of a controller on the network side of the fan controller model DC Dq component i of the output current of the grid-side converter dg 、i qg And u DC 、i dg 、i qg Error value R from hardware-in-the-loop experimental data e The output value is a network side control parameter; setting the voltage drop degrees of the fan grid-connected points at equal intervals to be 20%, 40%, 60% and 80%, and setting the duration to be 0.5s, and collecting 100 groups of data under each drop degree, wherein the total number of the data is 400;
s202, increasing dimensionality of an input feature set, and taking partitioned values of uDC, idg and iqg in the input feature value as the feature set;
s203, removing irrelevant features, and performing relevance analysis on the feature set obtained after dimensionality is increased by adopting a Person relevance coefficient method; estimating the covariance and standard deviation between the two variables, the Person correlation coefficient can be obtained:
in the formula (I), the compound is shown in the specification,the larger the absolute value of P is the sample mean value, the stronger the correlation between the two variables is;
finally obtaining an interval characteristic value with a larger P value as an input characteristic set by calculating a Person correlation coefficient of each interval characteristic value and an output control parameter;
s204, the acquired data is normalized, the preprocessed data is limited in the range of [0,1], and adverse effects caused by singular sample data are eliminated.
And S3, identifying the control parameters of the doubly-fed fan by adopting the LSTM neural network based on the processed data set, obtaining the structural parameters and the network training parameters of the LSTM neural network after a large amount of debugging work, and training and testing the LSTM neural network model.
Preferably, the LSTM neural network employed herein comprises 1 input and output layer and 2 hidden layers, the number of neurons of the hidden layers being 12. The neuron structure is shown in fig. 3, and the core cell state C is controlled by a forgetting gate, an input gate and an output gate. σ is sigmoid function, x k Is the input of the kth cell, C k Is the cell state of the kth cell, h k Is the hidden state of the kth cell;
the input gate is used for controlling the number of the current input data of the network flowing into the memory unit and is stored in c, and the input gate formula is as follows:
i t =σ(W i ·[h t-1 ,x t ]+b i ) (10)
the forgetting gate can judge the influence degree of the memory unit information ct-1 at the last moment on the current memory unit ct, and the formula is as follows:
f t =δ(W f ·[h t-1 ,x t ]+b f ) (11)
the influence of the output gate control memory unit ct on the output value ht can be determined by the following equation:
o t =σ(W o ·[h t-1 ,x t ]+b o ) (14)
h t =o t ⊙tanh(c t ) (15)
b1, dividing a sample training set and a verification set, randomly selecting 90% of data sets as the training set, and using the rest 10% of the data sets as the verification set to verify the accuracy of training results;
b2, initializing LSTM neural network parameters, setting a loss function MSE and an optimizer Adam, wherein the learning rate is 0.015, and the training times are 500;
b3, identifying the acquired sequence by adopting an LSTM neural network, and obtaining an identification result;
and B4, if the error is smaller than a certain value, outputting the optimal solution as an identification result, and otherwise, repeating the steps B2-B3.
And S4, comparing and verifying the identification result obtained by the LSTM neural network and hardware-in-loop experimental data under the voltage drop degrees of 20%, 40%, 60% and 80%.
Preferably, in step S4, in order to further verify the validity and practicability of the extracted model identification, the LSTM model of the present invention is compared with the BP model and the RNN model, and the model is trained using the same measured data.
Further, the above training is performed by u DC For example, the average deviations between the a, B, C, D and E intervals of the 3 model identifications at 20%, 40%, 60% and 80% voltage drop degrees were calculated according to the following formula.
In the formula, F Vi Is a section V i Average deviation of (u) M For actually measuring the DC capacitor voltage u i To identify the model DC capacitor voltage, K start And K end The first and last simulation data sequence numbers of the interval.
The invention has the following beneficial effects:
1. compared with the conventional parameter identification method, the doubly-fed fan controller parameter identification method based on the LSTM neural network can simulate the operation characteristics of a fan control system through training historical sample data, and inputs measured data to the LSTM neural network to perform off-line identification of control parameters under the condition that a fan model is not operated;
2. the method divides input data into 5 intervals according to simulation time and waveform states, is used for increasing dimensionality of an input feature set, and adopts a Person correlation coefficient method to perform correlation analysis on the feature set obtained after dimensionality is increased, so that irrelevant features of the input feature set are removed.
3. Aiming at the problem that the control parameters of the electromagnetic model of the doubly-fed wind turbine are difficult to obtain with high precision under the transient working condition, the method can comprehensively identify the control parameters of the wind turbine under different voltage drop conditions by setting the voltage drop degrees of the grid-connected points of the wind turbine at equal intervals to be 20%, 40%, 60% and 80%, and improve the identification precision of the control parameters under different voltage drop conditions.
Drawings
Fig. 1 is a schematic diagram of a method for identifying a control parameter of a doubly-fed wind turbine according to the present invention.
Fig. 2 shows the Person coefficient and value of each input feature value of the present invention.
FIG. 3 is an iterative diagram of the loss function in the LSTM model training of the present invention.
FIG. 4 is a diagram of the LSTM neural network pair parameter K of the present invention P1 The training recognition result.
FIG. 5 is a diagram of the LSTM neural network pair parameter K of the present invention i1 The training recognition result.
FIG. 6 is a diagram of the LSTM neural network pair parameter K of the present invention P2 The training recognition result.
FIG. 7 is a diagram of the LSTM neural network pair parameter K of the present invention i2 The training recognition result.
FIG. 8 is a diagram of the LSTM neural network of the present invention versus the parameter K P3 The training recognition result.
FIG. 9 is a diagram of the LSTM neural network pair parameter K of the present invention i3 The recognition result is trained.
FIG. 10 is a diagram showing the comparison result between the LSTM identification curve, the BP identification curve and the RNN identification curve of the present invention and the DC capacitance voltage of the actual measurement curve under the voltage drop degree of 20%.
FIG. 11 is a d-axis DC comparison result chart of the LSTM identification curve, the BP identification curve and the RNN identification curve of the present invention with the actual measurement curve under the voltage drop degree of 20%.
FIG. 12 is a q-axis DC comparison result plot of the LSTM, BP, and RNN identification curves of the present invention with the measured curves at 20% voltage droop.
FIG. 13 is a diagram showing the comparison result between the LSTM identification curve, the BP identification curve and the RNN identification curve of the present invention and the DC capacitance voltage of the actual measurement curve under the voltage drop degree of 80%.
FIG. 14 is a d-axis DC comparison result chart of the LSTM identification curve, the BP identification curve and the RNN identification curve of the present invention with the actual measurement curve at 80% voltage drop.
FIG. 15 is a q-axis DC comparison result plot of the LSTM, BP, and RNN identification curves of the present invention with the measured curves at 80% voltage droop.
Detailed Description
As shown in fig. 1 to 15, a method for identifying control parameters of a doubly-fed wind turbine based on an LSTM neural network includes the following steps:
s1, building an identification model according to a double-fed fan hardware-in-loop physical model; wherein the electrical parameters of the transformer, the filter and other elements in the identification model are consistent with the electrical parameters of the semi-physical test model.
Further, in step S1, in order to maintain the voltage of the parallel capacitors in the back-to-back converter to be constant and to control the vector control of the reactive power grid voltage orientation output by the converter; the voltage of the capacitor is controlled by the d-axis component of the grid-side converter current, while the port voltage of the wind power system is controlled by the q-axis component of the grid-side converter current, the reference value for the q-axis current being usually set to 0 in order to reduce losses.
The network side controller control equation can be written as:
i dg_ref =-K p1 Δu DC +K i1 x 1 (2)
thus, it is possible to obtain:
u dg =K p2 (-K p1 Δu DC +K i1 x 1 -i dg )+K i2 x 2 +X Tg i qg (5)
u qg =K p3 (i qg_ref -i qg )+K i3 x 3 -X Tg i dg (6)
in the formula, K p1 And K i1 Proportional coefficient and integral coefficient of the voltage controller respectively; k p2 And K i2 A first group of proportionality coefficients and integral coefficients of the current inner loop controller respectively; k i3 And K i3 A second group of proportional coefficients and integral coefficients of the current inner loop controller are respectively; u. of DC And u DC_ref Respectively an actual value and a reference value of the direct current bus voltage; i.e. i dg And i qg Actual values of the grid side d and q axis currents respectively; u. of dg And u qg Actual values of the grid side d and q axis voltages respectively; x Tg Is the reactance value of a transformer connected between the converter and the grid.
Under normal operating conditions, i qg_ref Typically 0, but under fault conditions, i qg_ref The low-penetration control module directly gives the following modes:
wherein k is the reactive current support coefficient, U N Rated voltage for fan grid-connected point, I N Rated current for the grid-side converter. When low voltage ride through occurs, the grid-side converter can send out certain reactive power, the value of the reactive current can be obtained according to the formula (7), at the moment, the q-axis current inner loop is switched to a low-ride-through control mode, and the q-axis reference value is given by a low-ride-through control module.
S2, collecting u under the voltage drop degrees of 20%, 40%, 60% and 80% respectively DC 、i dg And i qg And collecting u DC 、i dg And i qg Error value R of hardware-in-the-loop experimental data e (ii) a Using the collected values as input characteristic values and network side control parameters K p1 、K i1 、K p2 、K i2 、K i3 And K i3 Processing the data set for outputting the data set; wherein u DC The voltage of a direct current bus of a fan controller model network side controller capacitor; i.e. i dg And i qg Dq component of the output current for the grid-side converter; the method specifically comprises the following steps:
s201, outputting a mat file by using a Python calling model, and collecting a characteristic-output data set, wherein an input characteristic value is a direct-current bus voltage u of a controller on the network side of the fan controller model DC Dq component i of output current of grid-side converter dg 、i qg And u DC 、i dg 、i qg Error value R from hardware-in-the-loop experimental data e The output value is a network side control parameter; setting the voltage drop degrees of the fan grid-connected points at equal intervals to be 20%, 40%, 60% and 80%, and setting the duration to be 0.5s, and collecting 100 groups of data under each drop degree, wherein the total number of the data is 400;
s202, increasing dimensionality of an input feature set, and taking partitioned values of uDC, idg and iqg in the input feature value as the feature set;
s203, removing irrelevant features, and performing relevance analysis on the feature set obtained after dimensionality is increased by adopting a Person relevance coefficient method; estimating the covariance and standard deviation between the two variables, one can obtain the Person correlation coefficient:
in the formula (I), the compound is shown in the specification,the absolute value of P is the sample mean value, and the larger the absolute value of P is, the stronger the correlation between the two variables is;
finally, obtaining an interval characteristic value with a larger P value as an input characteristic set by calculating the Person correlation coefficient of each interval characteristic value and the output control parameter;
s204, carrying out normalization processing on the acquired data, limiting the preprocessed data in a range of [0,1], and eliminating adverse effects caused by singular sample data.
And S3, identifying the control parameters of the doubly-fed fan by adopting the LSTM neural network based on the processed data set, obtaining the structural parameters and the network training parameters of the LSTM neural network after a large amount of debugging work, and training and testing the LSTM neural network model.
Preferably, the LSTM neural network employed herein comprises 1 input, output layer and 2 hidden layers, the number of neurons of the hidden layers being 12. The neuron structure is shown in fig. 3, and the core cell state C is controlled by a forgetting gate, an input gate and an output gate. σ is sigmoid function, x k Is the input of the kth cell, C k Is the cell state of the kth cell, h k Is the hidden state of the kth cell;
the input gate is used for controlling the number of the current input data of the network flowing into the memory unit and is stored in c, and the formula of the input gate is as follows:
i t =σ(W i ·[h t-1 ,x t ]+b i ) (10)
the forgetting gate can judge the influence degree of the last moment memory unit information ct-1 on the current memory unit ct, and the formula is as follows:
f t =δ(W f ·[h t-1 ,x t ]+b f ) (11)
the influence of the output gate control memory unit ct on the output value ht can be determined by the following equation:
o t =σ(W o ·[h t-1 ,x t ]+b o ) (14)
h t =o t ⊙tanh(c t ) (15)
b1, dividing a sample training set and a verification set, randomly selecting 90% of data sets as the training set, and using the rest 10% of the data sets as the verification set to verify the accuracy of training results;
b2, initializing LSTM neural network parameters, setting a loss function MSE and an optimizer Adam, wherein the learning rate is 0.015, and the training times are 500;
b3, identifying the acquired sequence by adopting an LSTM neural network, and obtaining an identification result;
and B4, if the error is smaller than a certain value, outputting the optimal solution as an identification result, and otherwise, repeating the steps B2-B3.
And S4, comparing and verifying the identification result obtained by the LSTM neural network and hardware-in-the-loop experimental data under the voltage drop degrees of 20%, 40%, 60% and 80%.
Preferably, in step S4, in order to further verify the validity and practicability of the extracted model identification, the LSTM model of the present invention is compared with the BP model and the RNN model, and the model is trained using the same measured data.
Further, the above training is performed by u DC For example, the average deviations between the intervals a, B, C, D and E of the 3 model identification results at 20%, 40%, 60% and 80% voltage drop levels were calculated according to the following formula.
In the formula, F Vi Is a section V i Average deviation of (u) M For actually measuring the DC capacitor voltage u i To identify the model DC capacitor voltage, K start And K end The first and last simulation data sequence numbers of the interval.
Claims (2)
1. A doubly-fed wind turbine control parameter identification method based on an LSTM neural network is characterized by comprising the following steps: the method comprises the following steps:
s1, building an identification model according to a double-fed fan hardware-in-loop physical model; wherein the electrical parameters of the transformer, the filter and other elements in the identification model are consistent with the electrical parameters of the semi-physical test model;
s2, collecting u under the voltage drop degrees of 20%, 40%, 60% and 80% respectively DC 、i dg And i qg And collecting u DC 、i dg And i qg Error value R from hardware-in-the-loop experimental data e (ii) a Using the collected values as input characteristic values and network side control parameters K p1 、K i1 、K p2 、K i2 、K i3 And K i3 Processing the data set for outputting the data set; wherein u is DC The voltage of a direct current bus of a fan controller model network side controller capacitor; i.e. i dg And i qg Dq component of the output current for the grid-side converter; the method specifically comprises the following steps:
s201, outputting a mat file by using a Python calling model, and collecting a characteristic-output data set, wherein an input characteristic value is direct current bus voltage u of a fan controller model network side controller DC Dq component i of output current of grid-side converter dg 、i qg And u DC 、i dg 、i qg Error value R from hardware-in-the-loop experimental data e The output value is a network side control parameter; setting the voltage drop degrees of the fan grid-connected points at equal intervals to be 20%, 40%, 60% and 80%, and setting the duration to be 0.5s, and collecting 100 groups of data under each drop degree, wherein the total number of the data is 400;
s202, increasing the dimensionality of an input feature set, and taking the partitioned values of uDC, idg and iqg in the input feature value as the feature set;
s203, removing irrelevant features, and performing relevance analysis on the feature set obtained after dimensionality is increased by adopting a Person relevance coefficient method; estimating the covariance and standard deviation between the two variables, the Person correlation coefficient can be obtained:
in the formula (I), the compound is shown in the specification,the larger the absolute value of P is the sample mean value, the stronger the correlation between the two variables is;
finally, obtaining an interval characteristic value with a larger P value as an input characteristic set by calculating the Person correlation coefficient of each interval characteristic value and the output control parameter;
s204, carrying out normalization processing on the acquired data, limiting the preprocessed data in a range of [0,1], and eliminating adverse effects caused by singular sample data;
s3, identifying the control parameters of the doubly-fed wind turbine by adopting an LSTM neural network based on the processed data set, and training and testing an LSTM neural network model;
and S4, comparing and verifying the identification result obtained by the LSTM neural network and hardware-in-the-loop experimental data under the voltage drop degrees of 20%, 40%, 60% and 80%.
2. The LSTM neural network-based doubly-fed wind turbine control parameter identification method of claim 1, wherein the LSTM neural network-based doubly-fed wind turbine control parameter identification method comprises the following steps: in step S4, in order to further verify the validity and practicability of the model identification, the LSTM model of the present invention is compared with the BP model and the RNN model, and the models are trained using the same measured data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211181791.3A CN115411775B (en) | 2022-09-27 | 2022-09-27 | Double-fed fan control parameter identification method based on LSTM neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211181791.3A CN115411775B (en) | 2022-09-27 | 2022-09-27 | Double-fed fan control parameter identification method based on LSTM neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115411775A true CN115411775A (en) | 2022-11-29 |
CN115411775B CN115411775B (en) | 2024-04-26 |
Family
ID=84167324
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211181791.3A Active CN115411775B (en) | 2022-09-27 | 2022-09-27 | Double-fed fan control parameter identification method based on LSTM neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115411775B (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN204407916U (en) * | 2015-02-13 | 2015-06-17 | 南方电网科学研究院有限责任公司 | Comprehensive micro-grid experiment simulation platform containing wind, light and energy storage |
CN109738776A (en) * | 2019-01-02 | 2019-05-10 | 华南理工大学 | Fan converter open-circuit fault recognition methods based on LSTM |
CN111293693A (en) * | 2020-03-30 | 2020-06-16 | 华北电力大学 | Doubly-fed wind turbine converter control parameter identification method based on extended Kalman filtering |
WO2021068454A1 (en) * | 2019-10-12 | 2021-04-15 | 联合微电子中心有限责任公司 | Method for identifying energy of micro-energy device on basis of bp neural network |
US20210201155A1 (en) * | 2019-12-30 | 2021-07-01 | Dalian University Of Technology | Intelligent control method for dynamic neural network-based variable cycle engine |
CN113964885A (en) * | 2021-08-31 | 2022-01-21 | 国网山东省电力公司东营供电公司 | Reactive active prediction and control technology of power grid based on situation awareness |
CN114977939A (en) * | 2022-05-26 | 2022-08-30 | 三峡大学 | Doubly-fed wind turbine control parameter identification method based on improved multi-target particle swarm algorithm |
-
2022
- 2022-09-27 CN CN202211181791.3A patent/CN115411775B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN204407916U (en) * | 2015-02-13 | 2015-06-17 | 南方电网科学研究院有限责任公司 | Comprehensive micro-grid experiment simulation platform containing wind, light and energy storage |
CN109738776A (en) * | 2019-01-02 | 2019-05-10 | 华南理工大学 | Fan converter open-circuit fault recognition methods based on LSTM |
WO2021068454A1 (en) * | 2019-10-12 | 2021-04-15 | 联合微电子中心有限责任公司 | Method for identifying energy of micro-energy device on basis of bp neural network |
US20210201155A1 (en) * | 2019-12-30 | 2021-07-01 | Dalian University Of Technology | Intelligent control method for dynamic neural network-based variable cycle engine |
CN111293693A (en) * | 2020-03-30 | 2020-06-16 | 华北电力大学 | Doubly-fed wind turbine converter control parameter identification method based on extended Kalman filtering |
CN113964885A (en) * | 2021-08-31 | 2022-01-21 | 国网山东省电力公司东营供电公司 | Reactive active prediction and control technology of power grid based on situation awareness |
CN114977939A (en) * | 2022-05-26 | 2022-08-30 | 三峡大学 | Doubly-fed wind turbine control parameter identification method based on improved multi-target particle swarm algorithm |
Non-Patent Citations (1)
Title |
---|
潘学萍;鞠平;温荣超;吴峰;金宇清;: "解耦辨识双馈风电机组转子侧控制器参数的频域方法", 电力***自动化, no. 20, 25 October 2015 (2015-10-25) * |
Also Published As
Publication number | Publication date |
---|---|
CN115411775B (en) | 2024-04-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109165504B (en) | Power system false data attack identification method based on anti-generation network | |
CN109842373A (en) | Diagnosing failure of photovoltaic array method and device based on spatial and temporal distributions characteristic | |
CN108535572B (en) | Metering system secondary circuit monitoring method and device based on fundamental wave zero sequence characteristics | |
CN111160241B (en) | Power distribution network fault classification method, system and medium based on deep learning | |
CN112069727B (en) | Intelligent transient stability evaluation system and method with high reliability for power system | |
CN111553112A (en) | Power system fault identification method and device based on deep belief network | |
CN112649642A (en) | Electricity stealing position judging method, device, equipment and storage medium | |
CN110634082A (en) | Low-frequency load shedding system operation stage prediction method based on deep learning | |
CN111401792A (en) | Dynamic safety assessment method based on extreme gradient lifting decision tree | |
CN109921462B (en) | New energy consumption capability assessment method and system based on LSTM | |
CN114611676A (en) | New energy power generation system impedance model identification method and system based on neural network | |
CN114266396A (en) | Transient stability discrimination method based on intelligent screening of power grid characteristics | |
Rao et al. | Wideband impedance online identification of wind farms based on combined data-driven and knowledge-driven | |
CN115411775B (en) | Double-fed fan control parameter identification method based on LSTM neural network | |
CN116227320A (en) | Double-fed fan control parameter identification method based on LSTM-IPSO | |
CN105701265A (en) | Double-fed wind generator modeling method and apparatus | |
CN111293693A (en) | Doubly-fed wind turbine converter control parameter identification method based on extended Kalman filtering | |
CN112036010A (en) | Photovoltaic system dynamic process hybrid equivalent modeling method based on data driving | |
CN112910006B (en) | Universal electromagnetic transient modeling method for direct-drive wind turbine generator | |
CN116047222A (en) | Automatic identification method for voltage fault ride-through control parameters of new energy converter controller | |
CN114384319A (en) | Grid-connected inverter island detection method, system, terminal and medium | |
CN113536674A (en) | Line parameter identification method based on BP neural network and improved SCADA data | |
Fan et al. | Transmission Line Fault Type Identification based on Transfer Learning and Deep Learning | |
CN110988590A (en) | PCA-SVM model-based distribution network line selection method and system | |
Kamwa et al. | Sensorless ANN-based speed estimation of synchronous generators: Improved performance through physically motivated pre-filters |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |