CN109697302A - The behavior modeling method of radio-frequency power amplifier based on OP-ELM - Google Patents
The behavior modeling method of radio-frequency power amplifier based on OP-ELM Download PDFInfo
- Publication number
- CN109697302A CN109697302A CN201811366103.4A CN201811366103A CN109697302A CN 109697302 A CN109697302 A CN 109697302A CN 201811366103 A CN201811366103 A CN 201811366103A CN 109697302 A CN109697302 A CN 109697302A
- Authority
- CN
- China
- Prior art keywords
- elm
- behavior
- radio
- model
- power amplifier
- 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
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
-
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Amplifiers (AREA)
Abstract
The present invention discloses the behavior modeling method of the radio-frequency power amplifier based on OP-ELM, comprising steps of determining the input variable and corresponding output variable of the behavior model of radio-frequency power amplifier；Training data is imported in ELM algorithm model, to the training of ELM algorithm model；Assessment sequence is carried out by modeling result influence on the hidden layer neuron in ELM algorithm model network structure, the posterior predetermined hidden layer neuron of sequence is trimmed, forms the OP-ELM RF power amplification behavior model trimmed；Test data is imported in the OP-ELM RF power amplification behavior model trimmed, test error RMSE is calculated, compares the output of OP-ELM model as a result, verifying behavior modeling result.The present invention trims neuron wherein low to modeling result importance with and then sequence by the assessment to hidden layer neuron importance using OP-ELM algorithm, achieve the purpose that scientifically to reduce hidden layer nerve number, to realize easy rapidly and accurately to radio-frequency power amplifier progress external behavior modeling.
Description
Technical field
The present invention relates to fields of communication technology, more particularly to a kind of behavior of radio-frequency power amplifier based on OP-ELM
Modeling method.
Background technique
With the development of mobile communication technology, from second generation GSM to forth generation LTE communication technology or even current 5G all things on earth
Revolution is interconnected, in order to realize that the purpose for improving traffic rate, the bandwidth and power PAR of modulated signal all can significantly increase.This
All make the requirement to the radio-frequency power amplifier to play an important role in communication system further harsh a bit.Radio-frequency power amplifier
Non-linear and frequency dependence of itself etc. will all have an impact final communication efficiency.Therefore, accurately understand radio frequency function
The input-output characteristic of rate amplifier, i.e. behavior model just become ever more important.
Currently, the Behavioral Modeling Technique for radio-frequency power amplifier is varied, the wherein modeling side of neural network class
Method is the mainstream of educational circles.Recently, a kind of single hidden layer being extreme learning machine (extreme learning machine, ELM)
The learning algorithm of neuron neural network, can be in the case where obtaining same modeling accuracy, by feat of its extremely fast net
The advantage of network training speed, has been also introduced into the behavior modeling of radio-frequency power amplifier.But using ELM algorithm
It carries out, often in order to improve modeling accuracy, increasing hidden layer in single hidden layer neural network in the practical applications such as power amplifier modeling
Neuron leads to the not compact of network structure.
Therefore, extreme learning machine (Optimally Pruned Extreme Learning Machine, OP- are most preferably cut out
ELM it is that reply solves ELM algorithm in some realities earliest that) it is to efficiently solve that the innovatory algorithm as a kind of ELM, which is suggested,
In the application of border in the network structure that occurs the excessive problem of hidden layer neuron and be suggested.In OP-ELM algorithm, by
Multiple response sparse regression (multi-response sparse regression, MRSR) method is in ELM algorithm network structure
Effect of the hidden layer neuron in modeling is effectively assessed, judge each hidden layer neuron to modeling result
Importance.Then the low hidden layer neuron of those importance is removed, to reach reduction hidden layer neuron number and make net
The compact-sized purpose of network.
Therefore, it uses for reference this and most preferably cuts out extreme learning machine (Optimally Pruned Extreme Learning
Machine, OP-ELM), to propose the behavior modeling method of radio-frequency power amplifier, there is necessity.
Summary of the invention
In view of the technical drawbacks of the prior art, it is an object of the present invention to provide a kind of radio frequencies based on OP-ELM
The behavior modeling method of power amplifier, be by by the process most preferably trimmed in OP-ELM algorithm so that OP-ELM compared with
Existing ELM algorithm is preferably suitable for the behavior modeling method of radio-frequency power amplifier.
The technical solution adopted to achieve the purpose of the present invention is:
The behavior modeling method of radio-frequency power amplifier based on OP-ELM, comprising steps of
Determine the input variable and corresponding output variable of the behavior model of radio-frequency power amplifier；
Training data is imported in ELM algorithm model, to the training of ELM algorithm model；
Assessment sequence is carried out by modeling result influence on the hidden layer neuron in ELM algorithm model network structure, is repaired
The posterior predetermined hidden layer neuron of sequence is cut, the OP-ELM RF power amplification behavior model trimmed is formed；
Test data is imported in the OP-ELM RF power amplification behavior model trimmed, test error RMSE is calculated, compares
The output of OP-ELM model is as a result, verifying behavior modeling result.
The input variable is input power, and frequency input signal, the input variable is output power, PAE efficiency.
The test error RMSE calculation is as follows:.
Wherein, (Ymea)iIndicate the measured value of RF power amplification, (Ymdl)iIndicate building for the behavior model based on OP-ELM algorithm
Mould output is as a result, n indicates sample size.
Described comparison OP-ELM model output with measured result or simulation result the result is that be compared.
Using MRSR algorithm to the hidden layer neuron in ELM algorithm model network structure, according to hidden layer neuron pair
The influence of modeling result is assessed and is sorted.
The hidden layer neuron number needed in ELM algorithm model is determined using leaving-one method criterion LOO, it is extra to trim
The hidden layer neuron of sequence rearward.
A kind of RF power amplification behavior modeling method established based on OP-ELM of the invention, overcomes existing ELM to radio frequency function
Deficiency in rate amplifier behavior modeling method utilizes the assessment to the network concealed layer nerve net of the mono- hidden layer neural network of ELM
Sequence and optimum clipped solve in the existing RF power amplification modeling method based on ELM because that can not hide in network mechanism
The not compact technical problem of network structure, realizes for outside radio-frequency power amplifier caused by layer neuron Effective selection
Effective Accurate Model of behavior.
In the present invention, OP-ELM algorithm is trimmed wherein and then the assessment and sequence to hidden layer neuron importance
The neuron low to modeling result importance achievees the purpose that scientifically to reduce hidden layer nerve number, thus to be easy to be quick
External behavior modeling accurately is carried out to radio-frequency power amplifier, a kind of completely new solution is provided.
Detailed description of the invention
Fig. 1 show in the radio-frequency power amplifier behavior modeling method of the invention based on OP-ELM and most preferably trims network
The flow chart of neuronal process；
The network structure that Fig. 2 show after trimming in the behavior modeling method of the radio-frequency power amplifier based on OP-ELM is shown
It is intended to；
Fig. 3 show the modeling effect picture of the behavior modeling method of the radio-frequency power amplifier based on OP-ELM.
Specific embodiment
The present invention is described in further detail below in conjunction with the drawings and specific embodiments.It should be appreciated that described herein
Specific embodiment be only used to explain the present invention, be not intended to limit the present invention.
As shown in Figure 1, the radio-frequency power amplifier behavior modeling method based on OP-ELM, including with step:
1, the input variable (input power, frequency input signal) of the behavior model of radio-frequency power amplifier and corresponding is determined
Output variable (output power, PAE efficiency etc.)；
2, training data is imported in original EL M algorithm model, to the training of original EL M algorithm model.
3, it to hidden layer neuron in original EL M prototype network structure, is assessed according to its influence to modeling result
Sequence will affect the smaller posterior neuron of sequence and trim, forms the OP-ELM RF power amplification behavior model trimmed.
4, test data is imported in the OP-ELM RF power amplification behavior model most preferably trimmed, compares OP-ELM mould
Type reality output result and measured result (or simulation result) verify behavior modeling result；It wherein, is to calculate test according to the following formula
Error RMSE, verifying behavior modeling result:
Wherein, (Ymea)iIndicate the measured value of RF power amplification, (Ymdl)iIndicate building for the behavior model based on OP-ELM algorithm
Mould output is as a result, n indicates sample size.
The present invention is based on the measured data of radio-frequency power amplifier or by the emulation data training mould of ADS simulation software
Type obtains the behavior model of the power amplifier.The precision for the expected modeling that the model reaches, can satisfy auxiliary and design system
Grade communication transceiver necessary requirement.
In the following, the method for building up of the behavior model now in conjunction with a AB class wide-band radio frequency power amplifier, to the present invention into
Row is described in detail.The power amplifier can be incited somebody to action using the GaN HEMT RF power transistor of the model CGH40006P of CREE company
The power of radiofrequency signal on the broadband of 0.8GHz-2.14GHz amplifies.Implementation step is as follows:
1, the input/output variable of prediction model is selected.
In the present embodiment, the gain of the power amplifier is built with this behavior that the frequency of input signal increases and changes
Mould.Data are obtained by the emulation of ADS simulation software, input variable be power amplifier input signal frequency, variation range from
0.6GHz to 2.5GHz, stepping 0.01GHz.Output variable is the power gain of power amplifier.
2, training data is imported into ELM algorithm model, ELM algorithm model is trained, network structure such as Fig. 2 in algorithm
It is shown；
3, according to shown in Fig. 1, using MRSR algorithm to hidden layer neuron in the network structure of ELM algorithm model, according to
The importance of modeling result is assessed, and is ranked up from big to small by importance.
4, according to shown in Fig. 1, determine that OP-ELM is calculated using leaving-one method criterion (leave-one-out criterion, LOO)
Most suitable hidden layer neuron number in method, and the neuron of extra influence sequence rearward, trimming are trimmed according to this
Rear network structure is as shown in Fig. 2, OP-ELM algorithm model after being trimmed.
5, test data is imported into the OP-ELM algorithm model after trimming, calculates RMSE, assess modeling accuracy.
In the present embodiment, the hidden layer neuron number after trimming is 10, models RMSE=0.031, modeling result
As shown in figure 3, meeting modeling demand.
In conclusion establishing the behavior model of radio-frequency power amplifier using OP-ELM algorithm, that is, establishes radio-frequency power and put
Mapping relations one by one between big a pair of of input/output variable of device, can be right with the external behavior of the Accurate Prediction RF power amplification
The radio circuit of more scientific ground design communication transceiver realizes that more effectively wireless communication has great importance
The present invention is studied for the behavior modeling of radio-frequency power amplifier, for the first time by OP-ELM algorithm model application power amplifier
Behavior modeling solves deficiency present in existing tradition ELM algorithm model power amplifier behavior modeling, realizes to RF power amplification
External behavior is efficiently and accurate modeling, help understand the external behavior variation of RF power amplification, and then is wireless communication transceiver
The design of system provides effective guidance.
Compared to existing original EL M modeling method, hidden layer mind is had based on the power amplifier behavior model that OP-ELM is established
The advantage few through first number, network structure is compact enriches the RF power amplification behavior modeling of single hidden layer neural network class model
Method.Therefore the present invention has certain Value of Science & Technology and academic significance.
The above is only a preferred embodiment of the present invention, it is noted that for the common skill of the art
For art personnel, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications
Also it should be regarded as protection scope of the present invention.
Claims (6)
1. the behavior modeling method of the radio-frequency power amplifier based on OP-ELM, which is characterized in that comprising steps of
Determine the input variable and corresponding output variable of the behavior model of radio-frequency power amplifier；
Training data is imported in ELM algorithm model, to the training of ELM algorithm model；
Assessment sequence is carried out by modeling result influence on the hidden layer neuron in ELM algorithm model network structure, is trimmed
Sort posterior predetermined hidden layer neuron, forms the OP-ELM RF power amplification behavior model trimmed；
Test data is imported in the OP-ELM RF power amplification behavior model trimmed, test error RMSE is calculated, compares OP-
The output of ELM model is as a result, verifying behavior modeling result.
2. the behavior modeling method of the radio-frequency power amplifier based on OP-ELM as described in claim 1, which is characterized in that institute
Stating input variable is input power, and frequency input signal, the input variable is output power, PAE efficiency.
3. the behavior modeling method of the radio-frequency power amplifier based on OP-ELM as described in claim 1, which is characterized in that institute
The test error RMSE calculation stated is as follows:.
Wherein, (Ymea)iIndicate the measured value of RF power amplification, (Ymdl)iIndicate that the modeling of the behavior model based on OP-ELM algorithm is defeated
Out as a result, n indicates sample size.
4. the behavior modeling method of the radio-frequency power amplifier based on OP-ELM as described in claim 1, which is characterized in that institute
The comparison OP-ELM model output stated is the result is that be compared with measured result or simulation result.
5. the behavior modeling method of the radio-frequency power amplifier based on OP-ELM as described in claim 1, which is characterized in that make
With MRSR algorithm to the hidden layer neuron in ELM algorithm model network structure, according to hidden layer neuron to modeling result
Influence is assessed and is sorted.
6. the behavior modeling method of the radio-frequency power amplifier based on OP-ELM as described in claim 1, which is characterized in that make
The hidden layer neuron number needed in ELM algorithm model is determined with leaving-one method criterion LOO, trims extra sort rearward
Hidden layer neuron.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811366103.4A CN109697302A (en) | 2018-11-16 | 2018-11-16 | The behavior modeling method of radio-frequency power amplifier based on OP-ELM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811366103.4A CN109697302A (en) | 2018-11-16 | 2018-11-16 | The behavior modeling method of radio-frequency power amplifier based on OP-ELM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109697302A true CN109697302A (en) | 2019-04-30 |
Family
ID=66229877
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811366103.4A Pending CN109697302A (en) | 2018-11-16 | 2018-11-16 | The behavior modeling method of radio-frequency power amplifier based on OP-ELM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109697302A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110535486A (en) * | 2019-08-07 | 2019-12-03 | 东南大学 | The direct processing formula transceiver of radiofrequency signal based on super surface neural network |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160132635A1 (en) * | 2013-06-14 | 2016-05-12 | Keygene N.V. | Directed strategies for improving phenotypic traits |
| CN107977535A (en) * | 2017-12-22 | 2018-05-01 | 天津大学 | A kind of RF power amplification temperature characterisitic modeling method based on ELM |
-
2018
- 2018-11-16 CN CN201811366103.4A patent/CN109697302A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160132635A1 (en) * | 2013-06-14 | 2016-05-12 | Keygene N.V. | Directed strategies for improving phenotypic traits |
| CN107977535A (en) * | 2017-12-22 | 2018-05-01 | 天津大学 | A kind of RF power amplification temperature characterisitic modeling method based on ELM |
Non-Patent Citations (2)
| Title |
|---|
| CHENG-YU ZHANG等: "Extreme learning machine for the behavioral modeling of RF power amplifiers", 《2017 IEEE MTT-S INTERNATIONAL MICROWAVE SYMPOSIUM (IMS)》 * |
| YOAN MICHE等: "OP-ELM: Optimally Pruned Extreme Learning Machine", 《IEEE TRANSACTIONS ON NEURAL NETWORKS》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110535486A (en) * | 2019-08-07 | 2019-12-03 | 东南大学 | The direct processing formula transceiver of radiofrequency signal based on super surface neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kouhalvandi et al. | Automated deep neural learning-based optimization for high performance high power amplifier designs | |
| US11475607B2 (en) | Radio coverage map generation | |
| WO2020119372A1 (en) | Field intensity prediction method and apparatus, and device and storage medium | |
| CN110941896B (en) | Design method of ultra-wideband antenna based on PGP | |
| Benmus et al. | Neural network approach to model the propagation path loss for great Tripoli area at 900, 1800, and 2100 MHz bands | |
| CN107636656A (en) | GaN device technological parameter statistical analysis technique based on large signal equivalent circuit model | |
| CN108768550A (en) | Wide-band transmitter nonlinear modeling method based on dynamic multi-core bandwidth generalized regression neural network algorithm | |
| Xu et al. | Application of deep learning for high-speed differential via TDR impedance fast prediction | |
| CN109697302A (en) | The behavior modeling method of radio-frequency power amplifier based on OP-ELM | |
| CN111539165B (en) | Chip design method and system based on yield load traction system | |
| CN107085646A (en) | A kind of field strength measurement method in ground under transmission line of electricity | |
| Sotiroudis et al. | Ensemble learning for 5G flying base station path loss modelling | |
| CN111898320A (en) | Radio frequency power amplifier scattering parameter extraction method and device based on neural network | |
| CN102692241B (en) | Error compensation method of wireless positioning system | |
| CN114970419B (en) | InP terahertz HEMT positive and negative grid voltage small-signal model | |
| CN104573216A (en) | Antenna performance optimizing method and device | |
| CN103414671A (en) | 2D-DPD iteration reduction algorithm and application based on unit primary function | |
| Sofos et al. | Propagation model for vegetation effects in terrestrial and satellite mobile systems | |
| CN106055761A (en) | Load model parameter verification method based on decision tree classification | |
| Singh et al. | Design anylsis of an E-shaped slit loaded MPA and parameters estimation using ANN | |
| CN113852433B (en) | Wireless channel shadow fading model prediction method based on computer vision | |
| CN108268700A (en) | A kind of RF power amplification temperature characterisitic modeling method based on BPNN | |
| Caddemi et al. | Advanced simulation of semiconductor devices by artificial neural networks | |
| CN107729694B (en) | Multi-parameter electromagnetic field modeling simulation method based on neural network | |
| Li et al. | Microwave nonlinear device modelling by using an artificial neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190430 |