CN105722203A - High energy efficiency power distribution method based on particle swarm optimization for large-scale antenna system - Google Patents

Abstract

The invention discloses a high energy efficiency power distribution method based on particle swarm optimization for a large-scale antenna system. The power configuration corresponding to the optimal energy efficiency is solved by a distributed planning method, a Lagrangian multiplier penalty function method, a particle swarm optimization method and the like. The method comprises the steps of firstly, acquiring a user transmission rate lower bound expression under maximal ratio combining beam forming according to the unique channel matrix characteristic of a large-scale antenna by using a Jensen inequality; secondly, converting the original constrained non-convex optimization problem into a non-constrained convex optimization problem by using a distributed planning thought and a Lagrangian multiplier method; and finally, obtaining an optimal solution by using the particle swarm optimization under the multi-variant coupling condition. The method solves the problem that the power distribution is difficult to solve under the condition of serious interference (multi-variant coupling), and effectively improves the energy efficiency of the system while ensuring the user transmission rate.

Description

Extensive antenna system is based on the high energy efficiency power distribution method of particle cluster algorithm

Technical field

The present invention relates to a kind of extensive antenna system high energy efficiency power distribution method based on particle cluster algorithm, belong to the resource allocation techniques in mobile communication system.

Background technology

Extensive antenna technology becomes the key technology in future mobile communications with its high spectrum effect, the feature of high energy efficiency.But it is as the increase of antenna amount, overall antenna constant power consumption also can sharply increase, power module needs to rebuild, simultaneously, the conventional beam shaping operation computation complexity such as the technology such as ZF, least mean-square error can steeply rise, but fortunately, the feature of extensive antenna system makes the simplest maximum mixing ratio beam shaping can realize excellent performance, but, maximum mixing ratio beam shaping but cannot suppress the interference between user, in turn results in the Multivariable Coupling problem in efficiency expression formula.

Although extensive antenna system has the potential advantages of high energy efficiency, but from the angle of green communications, the high energy efficiency resource allocation methods in extensive antenna system remains to be needed badly.After problems in considering extensive antenna system, the invention provides a kind of high energy efficiency resource allocation optimization method.

Summary of the invention

Goal of the invention: in order to overcome the deficiencies in the prior art, the present invention provides a kind of extensive antenna system based on the high energy efficiency power distribution method of particle cluster algorithm, under the premise ensureing user data transmission speed, it is achieved the improvement to base station end efficiency situation.

Technical scheme: for achieving the above object, the technical solution used in the present invention is:

Extensive antenna system is based on the high energy efficiency power distribution method of particle cluster algorithm, the power configuration that comprehensive fractional programming, Lagrange multiplier penalty function method are corresponding with PSO Algorithm optimum efficiency；Specifically include following steps:

(1) iteration mark k=0 is initialized；Given Initial Trans vector Representing the Initial Trans of user i, N represents total number of users, and the initial Lagrange coefficient that user i retrains about transfer rate is designated as

(2) energy valid value q during kth time iteration is calculated^(k):

$q^{\left(k\right)}=\frac{\underset{i=1}{\overset{N}{\Σ}}{\log }_2(1+\frac{N_t{p}_i^{\left(k\right)}}{\underset{j=1,j\≠i}{\overset{N}{\Σ}}{p}_j^{\left(k\right)}+{\mathrm{\σ}}^2/{\mathrm{\β}}_i})}{\underset{i=1}{\overset{N}{\Σ}}{p}_i^{\left(k\right)}+\underset{j=1}{\overset{N_t}{\Σ}}{p}_{c,j}}$

In formula: q^(k)Energy valid value q during for kth time iteration^(k),The transmitting power of user i during for kth time iteration,The transmitting power of user j during for kth time iteration；N_tFor base station line number, p_c,jFor the constant power consumption on antenna j；σ²For noise power spectral density, β_iChannel large scale fading parameter for user i；

(3) to q^(k)Adopt PSO Algorithm p^(k+1),It is transmitting power vector during (k+1) secondary iteration,The transmitting power of user i when being (k+1) secondary iteration；

(4) to p^(k+1)It is adjusted, ${\tilde{p}}^{(k+1)}=\[{\tilde{p}}_1^{(k+1)},{\tilde{p}}_2^{(k+1)},...,{\tilde{p}}_i^{(k+1)},...,{\tilde{p}}_N^{(k+1)}\],$ ${\tilde{p}}_i^{(k+1)}=min\{max(p_i^{(k+1)},0),p_{T,i}\},$ p_T,iMaximum transmission power for user i；

(5) utility function is calculated

$U_{EE}\left({\tilde{p}}^{\left(k+1\right)}\right)=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\stackrel{\‾}{r_i}}^{\left(k+1\right)}-q^{\left(k\right)}\left(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{p}_i^{\left(k+1\right)}+\underset{j=1}{\overset{N_t}{\mathrm{\Σ}}}{p}_{c,j}\right)$

In formula:For the user i real-time Transmission speed lower bound when+1 iteration of kth, receive in maximum mixing ratio, be derived under large-scale antenna array and Jensen's inequality synergy, be expressed as:

${\stackrel{\‾}{r_i}}^{(k+1)}=B{\log }_2\left(\frac{N_t{\mathrm{\β}}_i{p}_i^{(k+1)}}{\underset{j=1,j\≠i}{\overset{N}{\Σ}}{\mathrm{\β}}_j{p}_j^{(k+1)}+{\mathrm{BN}}_0}\right)$

Wherein, B is spectral bandwidth, N₀For channel noise power spectrum density；

To given normal number ε, (magnitude is 10^-1Within): ifThen iterative computation terminates；Otherwise, k=k+1；

(6) Lagrange coefficient that when calculating kth time iteration, user i retrains about transfer rate is designated as

${\mathrm{\ω}}_i^{\left(k\right)}=max\{0,{\mathrm{\ω}}_i^{\left(k+1\right)}-\mathrm{\θ}({\stackrel{\‾}{r_i}}^{\left(k\right)}-R_{T,i})\}$

In formula: θ is non-negative step-length, R_T,iMinimum transmission rate request for user i；

Return step (2).

Concrete, described step (3), according to fractional programming thought, solve p in combination with Lagrange multiplier penalty function method^(k+1)；Order Lagrange utility function Φ⁽ⁿ⁾For:

$\begin{array}{c}{\mathrm{\Φ}}^{\left(n\right)}=\mathrm{\Φ}\left(p^{\left(n\right)},q^{\left(k\right)}\right)=\mathrm{\Φ}\left(p^{\left(n\right)}\right)\\ =-\[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\stackrel{\‾}{r_i}}^{\left(n\right)}-q^{\left(k\right)}\left(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{p}_i^{\left(n\right)}+\underset{j=1}{\overset{N_t}{\mathrm{\Σ}}}{p}_{c,j}\right)\]-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ω}}_i^{\left(k\right)}\left({\stackrel{\‾}{r_i}}^{\left(n\right)}-R_{T,i}\right)\end{array}$

In formula: Φ⁽ⁿ⁾For Lagrange utility function；

Solve p^(k+1)Specifically include following steps:

(31) set up a population comprising M particle, initialize the velocity vector of iterations n=0 and each particle；M-th particle is designated asRepresent transmitting power vector during m-th particle nth iteration, time initialThe velocity vector of m-th particle is designated asRepresent velocity vector during m-th particle nth iteration, time initial $v_m^{\left(n\right)}=v_m^{\left(0\right)};$

(32) self optimal solution of m-th particle and globally optimal solution are calculated

Situation one: iterations n=0

The m-th particle solution when the 0th iteration is designated asThe m-th particle self optimal solution when the 0th iteration is designated as ${\mathrm{\Φ}}_m^{local}=\mathrm{\Φ}\left(p_m^{local}\right),p_m^{local}=p_m^{\left(0\right)};$

Globally optimal solution when the 0th iteration is designated as Φ^global=Φ (p^global),

Situation two: iterations n ≠ 0

The m-th particle solution when nth iteration is designated asIfThen update the m-th particle self optimal solution when nth iterationOtherwise, self optimal solution of m-th particle is maintained ${\mathrm{\Φ}}_m^{local}=\mathrm{\Φ}\left(p_m^{local}\right);$

IfThen update globally optimal solution Φ during nth iteration^global=Φ (p^global),Otherwise, globally optimal solution Φ is maintained^global=Φ (p^global)；

(33) velocity vector during m-th particle (n+1) secondary iteration is updated

$v_m^{(n+1)}=v_m^{\left(n\right)}+r_1{s}_1(p_m^{local}-p_m^{\left(n\right)})+r_2{s}_2(p^{global}-p_m^{\left(n\right)})$

In formula: r₁And r₂It is the stochastic variable between 0 to 1, s₁And s₂It it is non-negative step-length；

(34) adjust $p_m^{(n+1)}=p_m^{\left(n\right)}+v_m^{(n+1)};$

(35) judgeWhether restrain: if convergence, thenOtherwise, n=n+1, return step (32).

Beneficial effect: extensive antenna system provided by the invention is based on the high energy efficiency power distribution method of particle cluster algorithm, and utilizing Random Matrices Theory is a compact form relevant with antenna number and large scale fading factor by the signal to noise ratio abbreviation under the channel matrix vector effect of high dimension；Meanwhile, the present invention is by taking the operation of transfer rate lower limit in transfer rate in retraining so that former problem can be converted into unconfined convex optimization problem, greatly simplify the analysis and solution process of resource distribution；It addition, the invention solves the Solve problems of a class Multivariable Coupling, when the derivative method of object function solves difficulty, this efficiency prioritization scheme has avoided the infeasibility of calculating aspect dexterously.

Accompanying drawing explanation

Fig. 1 is extensive antenna scene schematic diagram；

Fig. 2 is extensive antenna system High-energy-efficienresource resource optimization method schematic diagram；

Fig. 3 is extensive antenna system High-energy-efficienresource resource optimization method algorithm flow schematic diagram；

Fig. 4 is particle cluster algorithm schematic diagram.

Detailed description of the invention

Below in conjunction with accompanying drawing, the present invention is further described.

Extensive antenna system is based on the high energy efficiency power distribution method of particle cluster algorithm, the power configuration that comprehensive fractional programming, Lagrange multiplier penalty function method are corresponding with PSO Algorithm optimum efficiency；Specifically include following steps:

(1) iteration mark k=0 is initialized；Given Initial Trans vector Representing the Initial Trans of user i, N represents total number of users, and the initial Lagrange coefficient that user i retrains about transfer rate is designated as

(2) energy valid value q during kth time iteration is calculated^(k):

$q^{\left(k\right)}=\frac{\underset{i=1}{\overset{N}{\Σ}}{\log }_2(1+\frac{N_t{p}_i^{\left(k\right)}}{\underset{j=1,j\≠i}{\overset{N}{\Σ}}{p}_j^{\left(k\right)}+{\mathrm{\σ}}^2/{\mathrm{\β}}_i})}{\underset{i=1}{\overset{N}{\Σ}}{p}_i^{\left(k\right)}+\underset{j=1}{\overset{N_t}{\Σ}}{p}_{c,j}}$

In formula: q^(k)Energy valid value q during for kth time iteration^(k),The transmitting power of user i during for kth time iteration,The transmitting power of user j during for kth time iteration；N_tFor base station line number, p_c,jFor the constant power consumption on antenna j；σ²For noise power spectral density, β_iChannel large scale fading parameter for user i；

(3) to q^(k)Adopt PSO Algorithm p^(k+1),It is transmitting power vector during (k+1) secondary iteration,The transmitting power of user i when being (k+1) secondary iteration；

(4) to p^(k+1)It is adjusted, ${\tilde{p}}^{(k+1)}=\[{\tilde{p}}_1^{(k+1)},{\tilde{p}}_2^{(k+1)},...,{\tilde{p}}_i^{(k+1)},...,{\tilde{p}}_N^{(k+1)}\],$ ${\tilde{p}}_i^{(k+1)}=min\{max(p_i^{(k+1)},0),p_{T,i}\},$ p_T,iMaximum transmission power for user i；

(5) utility function is calculated

$U_{EE}\left({\tilde{p}}^{(k+1)}\right)=\underset{i=1}{\overset{N}{\Σ}}{\stackrel{\‾}{r_i}}^{(k+1)}-q^{\left(k\right)}\left(\underset{i=1}{\overset{N}{\Σ}}{p}_i^{\left(k+1\right)}+\underset{j=1}{\overset{N_t}{\Σ}}{p}_{c,j}\right)$

In formula:For the user i real-time Transmission speed lower bound when+1 iteration of kth, receive in maximum mixing ratio, be derived under large-scale antenna array and Jensen's inequality synergy, be expressed as:

${\stackrel{\‾}{r_i}}^{(k+1)}=B{\log }_2\left(\frac{N_t{\mathrm{\β}}_i{p}_i^{(k+1)}}{\underset{j=1,j\≠i}{\overset{N}{\Σ}}{\mathrm{\β}}_j{p}_j^{(k+1)}+{\mathrm{BN}}_0}\right)$

Wherein, B is spectral bandwidth, N₀For channel noise power spectrum density；

To given normal number ε, (magnitude is 10^-1Within): ifThen iterative computation terminates；Otherwise, k=k+1；

(6) Lagrange coefficient that when calculating kth time iteration, user i retrains about transfer rate is designated as

${\mathrm{\ω}}_i^{\left(k\right)}=max\{0,{\mathrm{\ω}}_i^{\left(k-1\right)}-\mathrm{\θ}({\stackrel{\‾}{r_i}}^{\left(k\right)}-R_{T,i})\}$

In formula: θ is non-negative step-length, R_T,iMinimum transmission rate request for user i；

Return step (2).

Concrete, described step (3), according to fractional programming thought, solve p in combination with Lagrange multiplier penalty function method^(k+1)；Order Lagrange utility function Φ⁽ⁿ⁾For:

$\begin{array}{c}{\mathrm{\Φ}}^{\left(n\right)}=\mathrm{\Φ}\left(p^{\left(n\right)},q^{\left(k\right)}\right)=\mathrm{\Φ}\left(p^{\left(n\right)}\right)\\ =-\[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\stackrel{\‾}{r_i}}^{\left(n\right)}-q^{\left(k\right)}\left(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{p}_i^{\left(n\right)}+\underset{j=1}{\overset{N_t}{\mathrm{\Σ}}}{p}_{c,j}\right)\]-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ω}}_i^{\left(k\right)}\left({\stackrel{\‾}{r_i}}^{\left(n\right)}-R_{T,i}\right)\end{array}$

In formula: Φ⁽ⁿ⁾For Lagrange utility function；

Solve p^(k+1)Specifically include following steps:

(31) set up a population comprising M particle, initialize the velocity vector of iterations n=0 and each particle；M-th particle is designated asRepresent transmitting power vector during m-th particle nth iteration, time initialThe velocity vector of m-th particle is designated asRepresent velocity vector during m-th particle nth iteration, time initial $v_m^{\left(n\right)}=v_m^{\left(0\right)};$

(32) self optimal solution of m-th particle and globally optimal solution are calculated

Situation one: iterations n=0

The m-th particle solution when the 0th iteration is designated asThe m-th particle self optimal solution when the 0th iteration is designated as ${\mathrm{\Φ}}_m^{local}=\mathrm{\Φ}\left(p_m^{local}\right),p_m^{local}=p_m^{\left(0\right)};$

Globally optimal solution when the 0th iteration is designated as Φ^global=Φ (p^global),

Situation two: iterations n ≠ 0

The m-th particle solution when nth iteration is designated asIfThen update the m-th particle self optimal solution when nth iterationOtherwise, self optimal solution of m-th particle is maintained ${\mathrm{\Φ}}_m^{local}=\mathrm{\Φ}\left(p_m^{\mathrm{local}}\right);$

IfThen update globally optimal solution Φ during nth iteration^global=Φ (p^global),Otherwise, globally optimal solution Φ is maintained^global=Φ (p^global)；

(33) velocity vector during m-th particle (n+1) secondary iteration is updated

$v_m^{(n+1)}=v_m^{\left(n\right)}+r_1{s}_1(p_m^{local}-p_m^{\left(n\right)})+r_2{s}_2(p^{global}-p_m^{\left(n\right)})$

In formula: r₁And r₂It is the stochastic variable between 0 to 1, s₁And s₂It it is non-negative step-length；

(34) adjust $p_m^{(n+1)}=p_m^{\left(n\right)}+v_m^{(n+1)};$

(35) judgeWhether restrain: if convergence, thenOtherwise, n=n+1, return step (32).

The above is only the preferred embodiment of the present invention; it is noted that, for those skilled in the art; under the premise without departing from the principles of the invention, it is also possible to make some improvements and modifications, these improvements and modifications also should be regarded as protection scope of the present invention.

Claims (2)

1. extensive antenna system is based on the high energy efficiency power distribution method of particle cluster algorithm, it is characterised in that: the power configuration that comprehensive fractional programming, Lagrange multiplier penalty function method are corresponding with PSO Algorithm optimum efficiency；Specifically include following steps:

(1) iteration mark k=0 is initialized；Given Initial Trans vector Representing the Initial Trans of user i, N represents total number of users, and the initial Lagrange coefficient that user i retrains about transfer rate is designated as

(2) energy valid value q during kth time iteration is calculated^(k):

$q^{\left(k\right)}=\frac{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\log }_2(1+\frac{N_t{p}_i^{\left(k\right)}}{\underset{j=1,j\≠i}{\overset{N}{\mathrm{\Σ}}}{p}_j^{\left(k\right)}+{\mathrm{\σ}}^2/{\mathrm{\β}}_i})}{\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{p}_i^{\left(k\right)}+\underset{j=1}{\overset{N_t}{\mathrm{\Σ}}}{p}_{c,j}}$

In formula: q^(k)Energy valid value q during for kth time iteration^(k),The transmitting power of user i during for kth time iteration,The transmitting power of user j during for kth time iteration；N_tFor base station line number, p_c,jFor the constant power consumption on antenna j；σ²For noise power spectral density, β_iChannel large scale fading parameter for user i；

(3) to q^(k)Adopt PSO Algorithm p^(k+1),It is transmitting power vector during (k+1) secondary iteration,The transmitting power of user i when being (k+1) secondary iteration；

(4) to p^(k+1)It is adjusted, ${\tilde{p}}^{(k+1)}=\[{\tilde{p}}_1^{(k+1)},{\tilde{p}}_2^{(k+1)},...,{\tilde{p}}_i^{(k+1)},...,{\tilde{p}}_N^{(k+1)}\],$ ${\tilde{p}}_i^{(k+1)}=min\{max(p_i^{(k+1)},0),p_{T,i}\},$ p_{T, i}Maximum transmission power for user i；

(5) utility function is calculated

$U_{EE}\left({\tilde{p}}^{(k+1)}\right)=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\stackrel{\‾}{r_i}}^{(k+1)}-q^{\left(k\right)}(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{p}_i^{(k+1)}+\underset{j=1}{\overset{N_t}{\mathrm{\Σ}}}{p}_{c,j})$

In formula:For the user i real-time Transmission speed lower bound when+1 iteration of kth, receive in maximum mixing ratio, be derived under large-scale antenna array and Jensen's inequality synergy, be expressed as:

${\stackrel{\‾}{r_i}}^{(k+1)}=B{\log }_2\left(\frac{N_t{\mathrm{\β}}_i{p}_i^{(k+1)}}{\underset{j=1,j\≠i}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\β}}_j{p}_j^{(k+1)}+{\mathrm{BN}}_0}\right)$

Wherein, B is spectral bandwidth, N₀For channel noise power spectrum density；

To given normal number ε, (magnitude is 10^-1Within): ifThen iterative computation terminates；Otherwise, k=k+1；

(6) Lagrange coefficient that when calculating kth time iteration, user i retrains about transfer rate is designated as

${\mathrm{\ω}}_i^{\left(k\right)}=max\{0,{\mathrm{\ω}}_i^{(k-1)}-\mathrm{\θ}({\stackrel{\‾}{r_i}}^{\left(k\right)}-R_{T,i})\}$

In formula: θ is non-negative step-length, R_T,iMinimum transmission rate request for user i；

$p^{\left(k\right)}={\tilde{p}}^{\left(k\right)},$ Return step (2).

2. extensive antenna system according to claim 1 is based on the high energy efficiency power distribution method of particle cluster algorithm, it is characterised in that: described step (3), according to fractional programming thought, solve p in combination with Lagrange multiplier penalty function method^(k+1)；Order Lagrange utility function Φ⁽ⁿ⁾For:

$\begin{array}{c}{\mathrm{\Φ}}^{\left(n\right)}=\mathrm{\Φ}(p^{\left(n\right)},q^{\left(k\right)})=\mathrm{\Φ}\left(p^{\left(n\right)}\right)\\ =-\[\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\stackrel{\‾}{r_i}}^{\left(n\right)}-q^{\left(k\right)}(\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{p}_i^{\left(n\right)}+\underset{j=1}{\overset{N_t}{\mathrm{\Σ}}}{p}_{c,j})\]-\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ω}}_i^{\left(k\right)}({\stackrel{\‾}{r_i}}^{\left(n\right)}-R_{T,i})\end{array}$

In formula: Φ⁽ⁿ⁾For Lagrange utility function；

Solve p^(k+1)Specifically include following steps:

(31) set up a population comprising M particle, initialize the velocity vector of iterations n=0 and each particle；M-th particle is designated asRepresent transmitting power vector during m-th particle nth iteration, time initialThe velocity vector of m-th particle is designated asRepresent velocity vector during m-th particle nth iteration, time initial $v_m^{\left(n\right)}=v_m^{\left(0\right)};$

(32) self optimal solution of m-th particle and globally optimal solution are calculated

Situation one: iterations n=0

The m-th particle solution when the 0th iteration is designated asThe m-th particle self optimal solution when the 0th iteration is designated as ${\mathrm{\Φ}}_m^{local}=\mathrm{\Φ}\left(p_m^{local}\right),$ $p_m^{local}=p_m^{\left(0\right)};$

Globally optimal solution when the 0th iteration is designated as Φ^global=Φ (p^global),

Situation two: iterations n ≠ 0

The m-th particle solution when nth iteration is designated asIfThen update the m-th particle self optimal solution when nth iteration Otherwise, self optimal solution of m-th particle is maintained ${\mathrm{\Φ}}_m^{local}=\mathrm{\Φ}\left(p_m^{local}\right);$

IfThen update globally optimal solution Φ during nth iteration^global=Φ (p^global),Otherwise, globally optimal solution Φ is maintained^global=Φ (p^global)；

(33) velocity vector during m-th particle (n+1) secondary iteration is updated

$v_m^{(n+1)}=v_m^{\left(n\right)}+r_1{s}_1(p_m^{local}-p_m^{\left(n\right)})+r_2{s}_2(p^{global}-p_m^{\left(n\right)})$

In formula: r₁And r₂It is the stochastic variable between 0 to 1, s₁And s₂It it is non-negative step-length；

(34) adjust $p_m^{(n+1)}=p_m^{\left(n\right)}+v_m^{(n+1)};$

(35) judgeWhether restrain: if convergence, thenOtherwise, n=n+1, return step (32).