CN112883655A - DC-DC converter parameter identification method based on dendritic network - Google Patents
DC-DC converter parameter identification method based on dendritic network Download PDFInfo
- Publication number
- CN112883655A CN112883655A CN202110383596.8A CN202110383596A CN112883655A CN 112883655 A CN112883655 A CN 112883655A CN 202110383596 A CN202110383596 A CN 202110383596A CN 112883655 A CN112883655 A CN 112883655A
- Authority
- CN
- China
- Prior art keywords
- input voltage
- converter
- network
- parameter identification
- output signal
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
-
- 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
Abstract
A DC-DC converter parameter identification method based on a dendritic network belongs to the technical field of component parameter identification. The method solves the problem that the accuracy of identifying the DC-DC converter parameters by adopting the existing method is low under the condition of input voltage fluctuation. The method selects key characteristics capable of representing the output signal of the DC-DC converter by taking the change coefficient as an index, thereby improving the precision of parameter identification and shortening the training time of the network; by constructing the input voltage identification network and taking the actual DC-DC converter input voltage as the characteristic input parameter identification network, the parameter identification precision is effectively improved. The method can effectively identify the current parameter value of the degradation element in the DC-DC converter when the input voltage fluctuates, and has the characteristics of high identification precision and small calculation amount. The invention obtains higher parameter identification precision under different input voltages and noise amplitudes, and the relative error of parameter identification is less than or equal to 7 percent. The method can be applied to component parameter identification in the DC-DC converter.
Description
Technical Field
The invention belongs to the technical field of component parameter identification, and particularly relates to a parameter identification method for a DC-DC converter.
Background
DC-DC converters (DC-DC converters) are important components of power supplies for electronic systems, and their operating states directly affect the performance of the back-end electronic components and even the entire electronic system. In order to monitor the health of the DC-DC converter, it is essential to efficiently identify the parameters of the critical components. There are many factors that affect the effect of parameter identification. Input voltage fluctuations are a major factor. Input voltage fluctuations often occur in renewable energy systems because the input signal of the DC-DC converter is the output power, which is not constant due to changes in wind speed, solar intensity. There are certain limitations in identifying parameters under fluctuating input voltages using common identification methods based on models and data driving.
The model-based approach is to build a mathematical model between the unidentified parameters and the output from the structure of the DC-DC converter. When identifying parameters under input voltage fluctuations using a model-based approach, the accuracy of the DC-DC converter parameter identification is reduced. The main reasons are as follows. In the established calculation model, the theoretical input voltage value is incorporated into the model calculation parameters. As soon as the actual input voltage fluctuates compared to the theoretical value, the calculation results deviate from the actual parameter values. Meanwhile, the sampling precision of the output signal is affected by inevitable noise in actual output. In addition, the additional circuit blocks increase the hardware cost and the circuit structure must be understood. These are limitations of model-based approaches.
The data-driven based identification method does not require an accurate DC-DC converter structure, but establishes a relationship between the output and unknown parameters by collecting data. The trained network may identify parameters by learning the regularity of the training data. Under the condition that the test data and the training data operate under the same operation condition, the data driving method can obtain higher identification precision. However, once the actual input voltage deviates from the operating conditions of the training data, the test data is in an untrained state for the trained network, and the accuracy of the network for identifying the parameters of the DC-DC converter is significantly reduced.
Disclosure of Invention
The invention aims to solve the problem that the accuracy of identifying DC-DC converter parameters by adopting the existing method is low under the condition of input voltage fluctuation, and provides a DC-DC converter parameter identification method based on a dendritic network.
The technical scheme adopted by the invention for solving the technical problems is as follows: a DC-DC converter parameter identification method based on a dendritic network specifically comprises the following steps:
inputting different voltages to a DC-DC converter under the state of fixed components, and respectively obtaining output signals of the DC-DC converter under each input voltage;
step two, respectively carrying out denoising processing on output signals of the DC-DC converter under each input voltage to obtain denoised output signals corresponding to each input voltage;
respectively acquiring time domain characteristics and frequency domain characteristics of the output signal after denoising processing corresponding to each input voltage;
selecting a characteristic sensitive to the input voltage change from the characteristics acquired in the step three based on the evaluation method of the coefficient of change value;
step five, under the condition that the input voltage of the DC-DC converter is fixed, respectively obtaining output signals of the DC-DC converter under different component states;
step six, respectively carrying out denoising processing on output signals of the DC-DC converter in each component state to obtain denoised output signals corresponding to each component state;
respectively acquiring time domain characteristics and frequency domain characteristics of the output signal after denoising processing corresponding to each component state;
step eight, screening out key features from the features obtained in the step seven by using an evaluation method based on the coefficient of variation value;
step nine, respectively constructing an input voltage identification network of the DC-DC converter and a parameter identification network of the DC-DC converter according to the principle of the dendritic network;
training the constructed input voltage identification network by using the characteristic value which is selected from the step four and is sensitive to the input voltage change and the corresponding input voltage; training the parameter identification network by using the key characteristic values screened in the step eight and the corresponding input voltage;
tenthly, denoising an actual output signal of the DC-DC converter to be subjected to parameter identification, and acquiring time domain characteristics and frequency domain characteristics of the denoised actual output signal;
inputting a characteristic value corresponding to a characteristic sensitive to input voltage change in an actual output signal into a trained input voltage identification network to obtain an input voltage value output by the input voltage identification network;
and the characteristic value corresponding to the key characteristic in the actual output signal and the input voltage value output by the input voltage identification network are used as the input of the trained parameter identification network, and the identification result of the element parameter is output through the parameter identification network.
The invention has the beneficial effects that: the invention provides a DC-DC converter parameter identification method based on a dendritic network, which adopts a variation coefficient as an index to select key characteristics capable of representing output signals of a DC-DC converter, thereby improving the precision of parameter identification and shortening the training time of the network; by constructing the input voltage identification network and taking the actual DC-DC converter input voltage as the characteristic input parameter identification network, the parameter identification precision is effectively improved. In addition, the method can still effectively identify the current parameter value of the internal degradation element of the DC-DC converter under the condition that the input voltage has fluctuation, and has the characteristics of high identification precision and small calculation amount. By adopting the method, higher parameter identification precision can be obtained under the conditions of different input voltages and noise amplitudes, and the obtained parameter identification relative error is less than or equal to 7 percent.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a Buck converter circuit diagram;
in the figure, TD is a delay time of a pulse, TR is a pulse rise time, TF is a fall time, PW is a pulse width, PER is a period, V1 is an initial value of a pulse, and V2 is a pulse amplitude.
FIG. 3 is a diagram of a hardware experimental platform adopted by the present invention.
Detailed Description
First embodiment this embodiment will be described with reference to fig. 1 and 3. The DC-DC converter parameter identification method based on the dendritic network according to the present embodiment specifically includes the following steps:
inputting different voltages to a DC-DC converter under the state of fixed components, and respectively obtaining output signals of the DC-DC converter under each input voltage;
step two, respectively carrying out denoising processing on output signals of the DC-DC converter under each input voltage to obtain denoised output signals corresponding to each input voltage;
respectively acquiring time domain characteristics and frequency domain characteristics of the output signal after denoising processing corresponding to each input voltage;
selecting a characteristic sensitive to the input voltage change from the characteristics acquired in the step three based on the evaluation method of the coefficient of change value;
step five, under the condition that the input voltage of the DC-DC converter is fixed, respectively obtaining output signals of the DC-DC converter under different component states;
step six, respectively carrying out denoising processing on output signals of the DC-DC converter in each component state to obtain denoised output signals corresponding to each component state;
respectively acquiring time domain characteristics and frequency domain characteristics of the output signal after denoising processing corresponding to each component state;
step eight, screening out key features from the features obtained in the step seven by using an evaluation method based on the coefficient of variation value;
the method for screening the key characteristics is the same as the fourth step;
step nine, respectively constructing an input voltage identification network of the DC-DC converter and a parameter identification network of the DC-DC converter according to the principle of the dendritic network;
training the constructed input voltage identification network by using the characteristic value which is selected from the step four and is sensitive to the input voltage change and the corresponding input voltage; training the parameter identification network by using the key characteristic values screened in the step eight and the corresponding input voltage;
when the input voltage identification network is trained, the selected characteristic value sensitive to the input voltage change is used as input, and the corresponding input voltage is used as output. When the parameter identification network is trained, the actual input voltage and the screened key characteristic value are used as input, and the output result is a parameter value to be identified;
tenthly, denoising an actual output signal of the DC-DC converter to be subjected to parameter identification, and acquiring time domain characteristics and frequency domain characteristics of the denoised actual output signal;
inputting the characteristic value corresponding to the characteristic sensitive to the input voltage change selected in the fourth step in the actual output signal into the trained input voltage identification network to obtain the input voltage value output by the input voltage identification network;
and e, taking the characteristic value corresponding to the key characteristic screened in the step eight in the actual output signal and the input voltage value output by the input voltage identification network as the input of the trained parameter identification network, and outputting the identification result of the element parameter through the parameter identification network.
The second embodiment is as follows: the present embodiment is different from the first embodiment in that a method for performing denoising processing on the output signal of the DC-DC converter is a wavelet packet analysis method.
The third concrete implementation mode: the difference between the present embodiment and the specific embodiment is that the time domain characteristics include a ripple amplitude r of the output signal after the denoising process, a mean value m of the output signal after the denoising process, a standard deviation v of the output signal after the denoising process, a skewness s of the output signal after the denoising process, a kurtosis k of the output signal after the denoising process, an information entropy e of the output signal after the denoising process, and a centroid c of the output signal after the denoising process.
The fourth concrete implementation mode: in this embodiment, the frequency domain feature is a local energy value selected in the denoising process.
The fifth concrete implementation mode: the fourth difference between this embodiment and the specific embodiment is that the expression of the frequency domain characteristic is:
in the formula, Ei' denotes a local energy value of an ith frequency band of a jth layer of a wavelet packet,a decomposition coefficient representing the ith frequency band of the jth layer of the wavelet packet;a kth decomposition coefficient of the decomposition coefficients representing the jth layer ith frequency band of the wavelet packet; m1And M2Indicating the serial number of the corresponding discrete point of the decomposed signal.
The local energy value reflects the energy of the original signal in a certain time period of the frequency band, and can fully reflect the time-varying property of the non-stationary time-varying signal.
The sixth specific implementation mode: the difference between this embodiment and the first embodiment is that the specific process of step four is as follows:
step four, setting the number of input voltage values as a, and defining all the input voltage values as a matrix P ═ P (P)1,p2,…,pa) For the i' th input voltage value pi’And i' is 1,2, …, a, and p is obtained according to the time domain characteristics and the frequency domain characteristics obtained in the step threei’Corresponding features are denoted by si’1,si’2,…,si’bWherein b is the number of all features (i.e.The total number of time domain features and frequency domain features);
then p isi’I' is 1,2, …, where S is a feature matrix formed by features corresponding to a, and the size of the feature matrix S is a × b;
step four and two, calculating a change coefficient matrix cv according to the feature matrix S, wherein cv is equal to (cv)1 cv2 … cvb),cvj’Is the coefficient of variation of the j 'th feature, j' is 1,2, …, b;
step four and three, for cvj’Taking the absolute value, if cvj’If the absolute value of the sum is not less than 15%, then cv is calculatedj’The corresponding characteristic serves as a characteristic that is sensitive to input voltage variations.
The seventh embodiment: the difference between this embodiment and the sixth embodiment is that the specific process in the fourth step is as follows:
the expression of the coefficient of variation matrix cv is as shown in equation (2):
the calculation expression of the coefficient of variation is shown in equation (3):
wherein v isj’Denotes s1j’,s2j’,…,saj’Standard deviation of (1), mj’Denotes s1j’,s2j’,…,saj’Is measured.
The specific implementation mode is eight: the difference between this embodiment and the first embodiment is that the specific process of step eight is:
step eight, setting the number of the component states as a ', and defining all the component states as a matrix P ═ P'1,p’2,…,p’a) For the i ' th ' component state p 'iAnd 1,2, …, a', according to the time domain feature and the frequency domain feature obtained in step seven,p'i”Corresponding characteristic is denoted as s'i”1,s’i”2,…,s’i”bWherein b is the number of all the characteristics;
then p'i”The feature matrix formed by the features corresponding to the features i ″, 1,2, …, a 'is S', and the size of the feature matrix S 'is a' × b;
step eight two, calculating a variation coefficient matrix cv ', cv ' ═ (cv '1 cv’2 … cv’b),cv’j’Is the coefficient of variation of the j 'th feature, j' is 1,2, …, b;
step eight three, to cv'j’Taking absolute value, if cv'j’Is not less than 15%, then cv'j’As a key feature.
The specific implementation method nine: the difference between this embodiment and the eighth embodiment is that the specific process of the eighth two step is:
the expression of the coefficient of variation matrix cv' is shown in formula (4):
the calculation expression of the coefficient of variation is shown in equation (5):
wherein, v'j’Denotes s1j’,s2j’,…,sa’j’Standard deviation of (1), m'j’Denotes s1j’,s2j’,…,sa’j’Is measured.
Assuming that R, C, L exist in the parameters to be identified, and there may be several values for each parameter, all possible combinations of the parameters R, C, L are obtained, each combination corresponds to one component state, and all component states form a state matrix P;
the detailed implementation mode is ten: the difference between this embodiment and the ninth embodiment is that, in the step ten, the characteristic value corresponding to the characteristic sensitive to the input voltage change in the actual output signal is input to the trained input voltage identification network, so as to obtain the input voltage value output by the input voltage identification network; the specific process comprises the following steps:
wherein X is the input vector of the input voltage identification network, the input vector is composed of the characteristic value corresponding to the characteristic sensitive to the input voltage variation in the actual output signal, and W is the input vectorl,l-1Representing a weight matrix from the L-1 th module to the L-th module of the network, L being 1,2, …, L being the number of modules in the network, Y being the output value of the input voltage identification network, o being the hadamard product, N+Is a positive integer.
The concrete implementation mode eleven: the present embodiment is very different from the specific embodiment in that the parameter identification network includes N parallel sub-networks, where the input of each sub-network is the same, and the input is the characteristic value corresponding to the key characteristic in the actual output signal and the input voltage value output by the input voltage identification network.
The expression of the parameter identification network is the same as the formula (6), except that the input vector of the parameter identification network is: the vector formed by the characteristic value corresponding to the key characteristic in the actual output signal and the input voltage value output by the input voltage identification network is output as the identification result of the element parameter.
The specific implementation mode twelve: the difference between this embodiment and the eleventh embodiment is that the value of N is the number of the parameters to be identified.
The following examples were used to demonstrate the beneficial effects of the present invention:
the first embodiment is as follows:
the invention takes a Buck converter as an example to explain the DC-DC converter parameter identification method based on the input voltage fluctuation condition of the dendritic network in detail. The Buck converter circuit is shown in fig. 2. Wherein Ron and RLRespectively MOS transistor andthe internal resistance of the diode, Rc, is the equivalent series resistance of the electrolytic capacitor. And performing functional simulation on the DC-DC converter by adopting Pspice simulation software, wherein simulation parameter settings are shown in Table 1. In the present simulation, the capacitance and inductance are degraded simultaneously. The parameters to be identified in the experiment are the capacitance C, the equivalent resistance Rc, the inductance L, the equivalent resistance R of the inductorL。
Table 1: buck converter parameter setting
To verify the effectiveness of the proposed method, first a simulation experiment was performed under ideal conditions. According to the parameter identification process, key features are extracted firstly. When a network is established, a large amount of training data needs to be acquired. First, component parameter states are set. In order to ensure that the selected key features have the universality of identifying various parameter values, the set parameter state is randomly selected within the failure range of the element parameters. The invention mainly identifies weak deviation of component parameters, so the parameter deviation range is set according to the failure threshold value of the component as follows: c: -5% to-20%, Rc: + 20% -60%, L: -5% to 20%, RL: + 20% -60%. In the simulation experiment, four parameter values were randomly selected within the range at an input voltage of +10V, as shown in table 2. The parameter status is expressed as a parameter value and a deviation ratio from a normal value.
Table 2: training data parameter settings
The output of each parameter in the ideal state is sampled. The sampling period is 1 mus, and the number of sampling points is 200. And decomposing the output by adopting three layers of wavelet packets, and calculating an energy value through a third layer of wavelet coefficient. Frequency domain featuresThe number is 8, and is denoted as e _ ideal1-e_ideal8(ii) a And calculates the time domain features. Finally, key features are extracted according to the variation coefficients, and the cv value of each feature is shown in table 3.
Table 3: coefficient of variation of each feature under ideal conditions
The invention converts cv intothSet to 15%, when the absolute value of the corresponding cv is larger than cvthThe feature is selected. Therefore, the key features in the time domain include ripple amplitude, standard deviation and centroid, and the key feature in the frequency domain is e _ ideal2~e_ideal8. These key features are used to describe the deviation of the component parameters. To improve the applicability of the training data to fluctuating conditions, an input voltage is also added to the training data. The number of parallel dendritic networks is 4, with 11 inputs and 1 output per network. For networks that recognize input voltage, resulting in output at variable input voltage, the key features selected include mean, e _ ideal1And e _ ideal7. Depending on the key features selected, two networks may be established. Experiments were performed with input voltage fluctuations, and the input voltage value of the training data was + 10V. Setting the parameter to be identified as C-42.3 mu f, RC=0.265Ω,L=522μH,RL60.131 Ω as test data. To verify the recognition ability of the method under different input voltage fluctuations, test data under +8.0V, +8.5V, +9.0V, +9.5V, +9.9V, +10.5V, +11.0V, +11.5V and +12.0V 9 input voltages were sampled. On the basis of the established network, key characteristics of the test data input voltage are obtained, so that the actual input voltage is identified. The component parameters are then identified using a plurality of parallel dendritic networks. The results of the experiment were described using the average relative error, and the number of executions per network was 10, and the final results are shown in table 4.
Table 4: parameter identification result under condition of fluctuating input voltage
The results show that the average relative error at each input voltage is relatively low when the training voltage is + 10.0V. The simulation result fully proves that the method can obtain better identification effect under the condition of input voltage fluctuation.
Example two:
in contrast to the first embodiment, the second embodiment considers the case where noise exists in the output voltage of the DC-DC converter. Also taking the Buck converter as an example, the parameter settings are shown in table 1 and table 2. The training data was sampled at +10V input voltage and ± 5mV noise. The output signal is decomposed by adopting three-layer wavelet packet analysis, and the obtained energy is e _ actual1~e_actual8(ii) a After denoising, the reserved energy range is e _ actual1~e_actual6。
After the time domain features of the output signal are obtained, the variation coefficient of each feature is recalculated, and the result is shown in table 5. According to the threshold value cvthFeatures selected for noise conditions include ripple range, skewness, standard deviation, kurtosis, centroid, e _ actual2-e_acutal5. The input voltage is also added to the training data. Each network has ten inputs and one output. For networks that recognize input voltage, an output at a variable input voltage is obtained, the selected key features including mean, E _ actual1And E _ acutal6。
Test data is sampled under different input voltages and noise conditions, the input voltages including +10.5V, +11.0V, +8.5V, +9.0V, and + 11.5V. The noise amplitude is set to + -10 mV, + -20 mV, + -30 mV, + -40 mV, + -50 mV. The identification parameter is set to C-40.23 muf, RC=0.305Ω,L=614μH,RL0.188 Ω. After 10 network runs, the identified input voltages and parameters are shown in table 6.
Table 5: coefficient of variation of each feature under noisy conditions
Table 6: parameter identification result under input voltage fluctuation and output noise conditions
The results show that the average relative error is low when the fluctuating conditions are close to the training data. Then, as the noise amplitude and the input voltage gradually deviate from the original conditions, the average relative error rises. But the average relative error remains small at values < 7.0%. The invention has been shown to have a strong ability to identify fluctuations including various input voltages and noise. These situations can occur in the actual operation of the DC-DC converter. Meanwhile, the invention has important practical value.
The above-described calculation examples of the present invention are merely to explain the calculation model and the calculation flow of the present invention in detail, and are not intended to limit the embodiments of the present invention. It will be apparent to those skilled in the art that other variations and modifications of the present invention can be made based on the above description, and it is not intended to be exhaustive or to limit the invention to the precise form disclosed, and all such modifications and variations are possible and contemplated as falling within the scope of the invention.
Claims (12)
1. A DC-DC converter parameter identification method based on a dendritic network is characterized by specifically comprising the following steps:
inputting different voltages to a DC-DC converter under the state of fixed components, and respectively obtaining output signals of the DC-DC converter under each input voltage;
step two, respectively carrying out denoising processing on output signals of the DC-DC converter under each input voltage to obtain denoised output signals corresponding to each input voltage;
respectively acquiring time domain characteristics and frequency domain characteristics of the output signal after denoising processing corresponding to each input voltage;
selecting a characteristic sensitive to the input voltage change from the characteristics acquired in the step three based on the evaluation method of the coefficient of change value;
step five, under the condition that the input voltage of the DC-DC converter is fixed, respectively obtaining output signals of the DC-DC converter under different component states;
step six, respectively carrying out denoising processing on output signals of the DC-DC converter in each component state to obtain denoised output signals corresponding to each component state;
respectively acquiring time domain characteristics and frequency domain characteristics of the output signal after denoising processing corresponding to each component state;
step eight, screening out key features from the features obtained in the step seven by using an evaluation method based on the coefficient of variation value;
step nine, respectively constructing an input voltage identification network of the DC-DC converter and a parameter identification network of the DC-DC converter according to the principle of the dendritic network;
training the constructed input voltage identification network by using the characteristic value which is selected from the step four and is sensitive to the input voltage change and the corresponding input voltage; training the parameter identification network by using the key characteristic values screened in the step eight and the corresponding input voltage;
tenthly, denoising an actual output signal of the DC-DC converter to be subjected to parameter identification, and acquiring time domain characteristics and frequency domain characteristics of the denoised actual output signal;
inputting a characteristic value corresponding to a characteristic sensitive to input voltage change in an actual output signal into a trained input voltage identification network to obtain an input voltage value output by the input voltage identification network;
and the characteristic value corresponding to the key characteristic in the actual output signal and the input voltage value output by the input voltage identification network are used as the input of the trained parameter identification network, and the identification result of the element parameter is output through the parameter identification network.
2. The DC-DC converter parameter identification method based on the dendritic network as claimed in claim 1, wherein the method for denoising the output signal of the DC-DC converter is a wavelet packet analysis method.
3. The DC-DC converter parameter identification method based on the dendritic network as claimed in claim 1, wherein the time domain characteristics comprise a ripple amplitude r of the denoised output signal, a mean value m of the denoised output signal, a standard deviation v of the denoised output signal, a skewness s of the denoised output signal, a kurtosis k of the denoised output signal, an information entropy e of the denoised output signal and a centroid c of the denoised output signal.
4. The DC-DC converter parameter identification method based on the dendritic network as claimed in claim 1, wherein the frequency domain feature is a local energy value selected in a denoising process.
5. The DC-DC converter parameter identification method based on the dendritic network according to claim 4, wherein the expression of the frequency domain characteristics is as follows:
in the formula, Ei' denotes a local energy value of an ith frequency band of a jth layer of a wavelet packet,a kth decomposition coefficient of the decomposition coefficients representing the jth layer ith frequency band of the wavelet packet; m1And M2Indicating the serial number of the corresponding discrete point of the decomposed signal.
6. The method for identifying DC-DC converter parameters based on the dendritic network according to claim 1, wherein the specific process of the fourth step is as follows:
step four, setting the number of input voltage values as a, and defining all the input voltage values as a matrix P ═ P (P)1,p2,…,pa) For the i' th input voltage value pi’And i' is 1,2, …, a, and p is obtained according to the time domain characteristics and the frequency domain characteristics obtained in the step threei’Corresponding features are denoted by si’1,si’2,…,si’bWherein b is the number of all the characteristics;
then p isi’I' is 1,2, …, where S is a feature matrix formed by features corresponding to a, and the size of the feature matrix S is a × b;
step four and two, calculating a change coefficient matrix cv according to the feature matrix S, wherein cv is equal to (cv)1 cv2 … cvb),cvj’Is the coefficient of variation of the j 'th feature, j' is 1,2, …, b;
step four and three, for cvj’Taking the absolute value, if cvj’If the absolute value of the sum is not less than 15%, then cv is calculatedj’The corresponding characteristic serves as a characteristic that is sensitive to input voltage variations.
7. The DC-DC converter parameter identification method based on the dendritic network according to claim 6, wherein the specific process of the second step is as follows:
the expression of the coefficient of variation matrix cv is as shown in equation (2):
the calculation expression of the coefficient of variation is shown in equation (3):
wherein v isj’Denotes s1j’,s2j’,…,saj’Standard deviation of (1), mj’Denotes s1j’,s2j’,…,saj’Is measured.
8. The DC-DC converter parameter identification method based on the dendritic network according to claim 1, wherein the specific process of the step eight is as follows:
step eight, setting the number of the component states as a ', and defining all the component states as a matrix P ═ P'1,p’2,…,p’a) For the i ' th ' component state p 'i”And i ″ -1, 2, …, a ', and p ' is determined according to the time domain feature and the frequency domain feature obtained in the step seven 'i”Corresponding characteristic is denoted as s'i”1,s’i”2,…,s’i”bWherein b is the number of all the characteristics;
then p'i”The feature matrix formed by the features corresponding to the features i ″, 1,2, …, a 'is S', and the size of the feature matrix S 'is a' × b;
step eight two, calculating a variation coefficient matrix cv ', cv ' ═ (cv '1 cv’2 … cv’b),cv’j’Is the coefficient of variation of the j 'th feature, j' is 1,2, …, b;
step eight three, to cv'j’Taking absolute value, if cv'j’Is not less than 15%, then cv'j’As a key feature.
9. The method according to claim 8, wherein the specific process of step eight two is as follows:
the expression of the coefficient of variation matrix cv' is shown in formula (4):
the calculation expression of the coefficient of variation is shown in equation (5):
wherein, v'j’Denotes s1j’,s2j’,…,sa’j’Standard deviation of (1), m'j’Denotes s1j’,s2j’,…,sa’j’Is measured.
10. The DC-DC converter parameter identification method based on the dendritic network according to claim 9, wherein in the step ten, the characteristic value corresponding to the characteristic sensitive to the input voltage variation in the actual output signal is input into the trained input voltage identification network, so as to obtain the input voltage value output by the input voltage identification network; the specific process comprises the following steps:
wherein X is the input vector of the input voltage identification network, Wl,l-1The weight matrix from the L-1 th module to the L-th module of the network is shown, L is 1,2, …, L is the number of modules in the network, Y is the output value of the input voltage identification network,is the product of Hadamard, N+Is a positive integer.
11. The DC-DC converter parameter identification method based on dendritic network of claim 10, wherein the parameter identification network comprises N parallel sub-networks, wherein the input of each sub-network is the same, and the input is the corresponding characteristic value of the key characteristic in the actual output signal and the input voltage value of the input voltage identification network output.
12. The DC-DC converter parameter identification method based on the dendritic network according to claim 11, wherein the value of N is the number of the parameters to be identified.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110383596.8A CN112883655B (en) | 2021-04-09 | 2021-04-09 | DC-DC converter parameter identification method based on dendritic network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110383596.8A CN112883655B (en) | 2021-04-09 | 2021-04-09 | DC-DC converter parameter identification method based on dendritic network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112883655A true CN112883655A (en) | 2021-06-01 |
CN112883655B CN112883655B (en) | 2022-07-01 |
Family
ID=76039950
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110383596.8A Active CN112883655B (en) | 2021-04-09 | 2021-04-09 | DC-DC converter parameter identification method based on dendritic network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112883655B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108702145A (en) * | 2016-02-29 | 2018-10-23 | 脉冲生物科学有限公司 | High voltage analog circuits impulse generator with feedback control |
CN110808580A (en) * | 2019-10-25 | 2020-02-18 | 国网天津市电力公司电力科学研究院 | Quick identification method for voltage sag source based on wavelet transformation and extreme learning machine |
CN111464022A (en) * | 2020-03-25 | 2020-07-28 | 哈尔滨工业大学 | Multi-feature-based parameter identification method for key devices of DC-DC converter |
US20200265319A1 (en) * | 2019-02-19 | 2020-08-20 | Volodymyr Bykov | Domain-Based Dendral Network |
-
2021
- 2021-04-09 CN CN202110383596.8A patent/CN112883655B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108702145A (en) * | 2016-02-29 | 2018-10-23 | 脉冲生物科学有限公司 | High voltage analog circuits impulse generator with feedback control |
US20200265319A1 (en) * | 2019-02-19 | 2020-08-20 | Volodymyr Bykov | Domain-Based Dendral Network |
CN110808580A (en) * | 2019-10-25 | 2020-02-18 | 国网天津市电力公司电力科学研究院 | Quick identification method for voltage sag source based on wavelet transformation and extreme learning machine |
CN111464022A (en) * | 2020-03-25 | 2020-07-28 | 哈尔滨工业大学 | Multi-feature-based parameter identification method for key devices of DC-DC converter |
Non-Patent Citations (2)
Title |
---|
MOHAMED AHMEID 等: "Real-Time Parameter Estimation of DC–DC Converters Using a Self-Tuned Kalman Filter", 《IEEE TRANSACTIONS ON POWER ELECTRONICS》 * |
罗震雄: "分布式DC-DC开关变换器非线性动力学特征的量化与识别", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Also Published As
Publication number | Publication date |
---|---|
CN112883655B (en) | 2022-07-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112685961B (en) | Method and system for predicting remaining service life of analog circuit | |
Jiang et al. | An efficient fault diagnostic method for three-phase induction motors based on incremental broad learning and non-negative matrix factorization | |
CN110346736B (en) | NPC three-level inverter fault diagnosis method based on improved treelet transformation | |
CN112327200B (en) | Health characterization and health baseline construction method of DC-AC inverter circuit | |
CN112883655B (en) | DC-DC converter parameter identification method based on dendritic network | |
Zhang et al. | Bearing fault diagnosis using a novel classifier ensemble based on lifting wavelet packet transforms and sample entropy | |
CN116167007A (en) | Analog circuit detection method based on gating recursion unit self-coding neural network | |
CN114966224A (en) | Improved neural network-based modular multilevel converter capacitance monitoring method | |
Guangquan et al. | Analog circuit incipient fault diagnosis using deep neural Network | |
CN112115663A (en) | Feature extraction method for early multi-source fault diagnosis of DC-DC converter | |
CN114931390A (en) | Muscle force estimation method based on fatigue analysis | |
DE102012016279A1 (en) | Method for parameterization of impedance-based model of e.g. lithium ion battery, involves determining parameter by adjusting response with time response of system with impedance elements e.g. resistors and constant phase elements | |
CN104200002B (en) | Method for extracting modal parameter from viscous damping vibration signals | |
CN100464500C (en) | Geometrical characteristic filtering method for time-frequency aliasing signal | |
CN113689044A (en) | Method and system for predicting residual service life of switching power supply | |
Jiang et al. | Parameter identification of DC–DC converter based on dendrite net under fluctuating input voltages | |
CN112305343B (en) | ESR and C monitoring method for output capacitor switching frequency point of Buck converter | |
CN111464022A (en) | Multi-feature-based parameter identification method for key devices of DC-DC converter | |
CN111426880A (en) | Method for monitoring ESR and C of output capacitor of Buck-type DC-DC converter | |
Guo et al. | Deep Degradation Feature Extraction and RUL Estimation for Switching Power Unit | |
Mosin | An approach to construction the neuromorphic classifier for analog fault testing and diagnosis | |
Ghaderi et al. | A Novel Mathematical Analysis for Electrical Specifications of Step-up Converter | |
CN117031259A (en) | Analog circuit fault diagnosis method, system and electronic equipment | |
CN114487603A (en) | Method for monitoring ESR (equivalent series resistance) of output capacitor of Buck converter | |
CN114563628A (en) | Neural network-based modular multilevel converter capacitance fault diagnosis method |
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 |