CN106102095A - Double yardstick dynamic resource Optimal Configuration Methods under collaborative multipoint communication system - Google Patents

Abstract

The invention belongs to the most optimum distribution of resources technical field of wireless communication system, a kind of by the double yardstick dynamic resource Optimal Configuration Methods under the collaborative multipoint communication system of intelligent grid energy supply.In the present invention: in big time scale (period), base station determines, by the regenerative resource of forward purchasing market power price and collection, the energy size concluded the business in advance with electrical network, and the energy having is averagely allocated to this period；In little time scale (time slot), base station carries out real-time power transaction based on Vehicles Collected from Market power price, channel information and forward purchasing energy with electrical network, and design collaboration launching beam shapes, and battery is carried out discharge and recharge.Present invention introduces a virtual energy sequence, and real-time update, obtain the online resource Optimal Configuration Method of optimum in combination with Lyapunov optimization method.The inventive method can be in the case of meeting certain user's qos requirement, the purpose of the average total electricity charge of minimizing.

Description

Double yardstick dynamic resource Optimal Configuration Methods under collaborative multipoint communication system

Technical field

The present invention relates to the most optimum distribution of resources field of wireless communication system, powered by intelligent grid particularly to a kind of Double yardstick dynamic resource Optimal Configuration Methods under collaborative multipoint communication system.

Background technology

Heterogeneous network (HetNets) is the developing direction of next generation cellular network.Do for the minizone in heterogeneous network Disturbing, collaborative multiple spot (coordinated multi-point, CoMP) technology is a kind of effective administrative mechanism.In CoMP system In, base station is divided into multiple group, and the base station in each group can be come by launching collaborative beam shaping (beamforming) Service user.Along with the increase of base stations in heterogeneous network number, they create the higher electricity charge and more CO2 emission Amount.The high efficiency of transmission of energy and economic and environment-friendly base station operational mode are pursued in green communication for coordination, are that to realize efficiency excellent with resource The important means changed.Along with the development of current intelligent power grid technology, combined with intelligent electric power network technique carries out green communication for coordination resource Distribute the new method being an up performance for wireless communications rationally.Compared to tradition electrical network, intelligent grid has following feature: can The utilization of the renewable sources of energy (renewable energy sources, RES), dynamic electricity price, bidirectional energy transaction and Demand-side ring Answer etc..

Fig. 1 is the CoMP system powered by intelligent grid, and shown all base stations are in a group.At its downlink In, total I:={1 ..., I} base station, each base station is furnished with M >=1 piece transmitting antenna, and they common service are in K:= 1 ..., K} user, each user is furnished with a reception antenna.Base station is except can obtain energy supply at intelligent grid Outside, it is also possible to gather regenerative resource by solar panels, wind turbine etc..Meanwhile, base station is equipped with battery, can store collection The energy arrived.Base station can also carry out the dealing of energy with the market price of dynamically change and electrical network.In this CoMP group, in Centre controller is connected to by the feedback network of a low latency with base station, plays coordination power exchange and the work of communication for coordination With.Central controller can collect communication data (transmission information and channel information) by honeycomb feedback link from base station, simultaneously By the intelligent electric meter installed in base station and the communication and control link collection energy information (the power exchange price that connect both With energy sequence size).

In practice, RES is typically to carry out in different time scales from the dynamically change of wireless channel.Time quilt Being divided into different " time slot (slot) ", its length, less than the coherence time of wireless channel, is normalized as unit for the sake of convenience “1”.On the other hand, the collection yardstick of note RES is " period (interval) ", and each period is made up of T time slot.

Remember in each start time period, t=nT, n=1,2 ..., A_i,n(A_n:=[A_1,n,...,A_I,n] ') is base station i The regenerative resource collected, E_i[n] is the energy expenditure estimated at next T time slot, and the energy concluded the business the most in advance is [E_i[n]- A_i,n]⁺Or [A_i,n-E_i[n]]⁺([a]⁺:=max{a, 0}), the former needs the energy bought in from electrical network for base station, and the latter is base station The unnecessary energy selling electrical network.Then for the i of base station, power exchange cost (i.e. the electricity charge) in advance is:

$G^{lt}(E_i\[n\]):={\mathrm{\α}}_n^{lt}{\[E_i\[n\]-A_{i,n}\]}^{+}-{\mathrm{\β}}_n^{lt}{\[A_{i,n}-E_i\[n\]\]}^{+}---\left(1\right)$

Wherein,For energy buying price in advance,For energy selling price in advance, it is clear thatNoteFor the stochastic variable under big time scale.

Remember for each time slot t,For the channel vector of base station i to user k, h_k,t:=[h_1k,t′,..., h_Ik,t′] ' for the channel vector of all base stations to user k, H_t:=[h_1,t....,h_K,t].WillIt is set to act on use The beam forming vector of family k, s_kT () is the transmitting signal with unit energy, then the vector signal being sent to user k is expressed as q_k(t)=w_k(t)s_k(t).Signal to Interference plus Noise Ratio at user k (signal-to-interference-plus-noise ratio, SINR) it is:

${\mathrm{SINR}}_k\left(\right\{w_k\left(t\right)\left\}\right)=\frac{|h_{k,t}^H{w}_k\left(t\right){|}^2}{\underset{l\≠k}{\Σ}\left(\right|h_{k,t}^H{w}_l\left(t\right){|}^2)+{\mathrm{\σ}}_k^2}---\left(2\right)$

Wherein,Multiple Gaussian noise (the circularly symmetric of additivity Cyclic Symmetry for being present in down channel Complex Gaussian, CSCG) variance.Then the transmission energy of base station i isWhereinFromThe row that middle selection is corresponding, constructs sending out of base station i Penetrate beam forming vector.For guarantee QoS of customer (quality ofservice, QoS), need to meet:

${\mathrm{SINR}}_k\left(\right\{w_k\left(t\right)\left\}\right)\≥{\mathrm{\γ}}_k,\∀k---\left(3\right)$

Wherein, γ_kRepresent the target sinr values of user k.

Note is for each time slot t, P_g,iT () is the gross energy that base station i consumes, it is by transmitting energy P_x,i(t) and constant P_c ＞ 0 (circuit, data processor etc. consumes) composition.P_g,iT the upper limit of () isThat is:

Real-time power exchange can be carried out with electrical network at each time slot t, base station i and carry out its demand supplies equalize.Note P_i T () is the energy of base station i and electrical network real-time deal,WithIt is respectively the buying price of real-time power and sells Going out price, the most real-time power exchange spends and is:

$G^{rt}\left(P_i\right(t\left)\right):={\mathrm{\α}}_t^{rt}{\[P_i\left(t\right)\]}^{+}-{\mathrm{\β}}_t^{rt}{\[-P_i\left(t\right)\]}^{+}---\left(5\right)$

NoteFor the stochastic variable under little time scale.

Note C_i(0) it is the initial value of battery energy storage, C in the i of base station_iT () is the battery energy storage capacity in time slot t start time, electricity Minima and the maximum of tankage are respectively C^minAnd C^max.Note P_b,iT () is the electricity of battery charge or discharge in time slot t, Then energy storage capacity follows following dynamic change:

$C_i(t+1)={\mathrm{\ηC}}_i\left(t\right)+P_{b,i}\left(t\right),C^{\min }\≤C_i\left(t\right)\≤C^{max},\∀i---\left(6\right)$

Wherein, η ∈ (0,1] represent energy storage efficiency.The size of charge/discharge electricity amount is the most restricted every time:

$P_b^{min}\≤P_{b,i}\left(t\right)\≤P_b^{max},\∀i---\left(7\right)$

Note Represent and round downwards, then can obtain following demand-supply balance equation:

It should be noted that the CoMP downlink powered by intelligent grid is a stochastic system.Target is design one Online resource allocation strategy, the energy that this strategy can pre-purchase in each period t=nT decisionAnd Each time slot t determines the energy implementing transactionThe electricity of battery charging and dischargingBeam shaping with CoMP VectorPurpose with the average total electricity charge of minimizing.The premise of this strategy is all stochastic processs in system Distribution is all the most unknowable.

According to formula (1) and formula (5), remember that the electricity charge of every time slot base station i are:

${\mathrm{\Φ}}_i\left(t\right):=\frac{1}{T}{G}^{lt}(E_i\[n_t\])+G^{rt}\left(P_i\right(t\left)\right)---\left(9\right)$

OrderThen optimization problem is as follows:

Subject to (3), (4), (6), (7), (8),

Summary of the invention

The present invention provides the double yardstick dynamic resource optimizations in a kind of collaborative multi-point communication network powered by intelligent grid Collocation method, in order in the case of meeting certain user's qos requirement, the purpose of the average total electricity charge of minimizing.

Double yardstick dynamic resource optimizations in the collaborative multi-point communication network powered by intelligent grid that the present invention provides are joined Put method, including:

Energy method of commerce in advance in big time scale (period)；

Real-time power equilibrium in little time scale (time slot) and beam-forming method.

Wherein, the method for commerce in advance of the energy in big time scale, concretely comprise the following steps:

(1) when each period starts, base station by forward purchasing market power price and the regenerative resource of collection determine with The energy size that electrical network is concluded the business in advance；

(2) base station and electrical network do power exchange, and all time slots being averagely allocated in this period by the energy having.

Wherein, the real-time power equilibrium in little time scale and beam-forming method, concretely comprise the following steps:

(1) at each time slot, base station carries out reality based on Vehicles Collected from Market power price, channel information and forward purchasing energy with electrical network Time power exchange；

(2) Base Transmitter collaborative beam forming, makes user QoS reach certain requirement；

(3) battery is carried out discharge and recharge.

In the present invention, when each period (each period is made up of T time slot) starts, base station is handed in advance with intelligent grid Easy energy expenditure (i.e. the electricity charge) is as shown in above-mentioned formula (1):

$G^{lt}(E_i\[n\]):={\mathrm{\α}}_n^{lt}{\[E_i\[n\]-A_{i,n}\]}^{+}-{\mathrm{\β}}_n^{lt}{\[A_{i,n}-E_i\[n\]\]}^{+},---\left(1\right)$

Wherein, A_i,n(A_n:=[A_1,n,…,A_I,n] ') is the regenerative resource that base station i collects, E_i[n] is when next The energy expenditure that in section, (T time slot) is anticipated,WithIt is respectively energy current bid and ask prices in advance.

Each time slot (t), base station is shown with the most above-mentioned formula of the energy expenditure (5) of intelligent grid real-time deal:

$G^{rt}\left(P_i\right(t\left)\right):={\mathrm{\α}}_t^{rt}{\[P_i\left(t\right)\]}^{+}-{\mathrm{\β}}_t^{rt}{\[-P_i\left(t\right)\]}^{+},---\left(5\right)$

Wherein, P_iT () is the energy of base station i and electrical network real-time deal,WithIt is respectively real-time power Current bid and ask prices.

According to formula (1) and formula (5), remember that the electricity charge of every time slot base station i are:

${\mathrm{\Φ}}_i\left(t\right):=\frac{1}{T}{G}^{lt}(E_i\[n_t\])+G^{rt}\left(P_i\right(t\left)\right)---\left(9\right)$

OrderThen optimization problem is as follows:

Subject to formula (3), (4), (6), (7), (8),

Here formula (3), (4), (6), (7), (8), all see illustrated above.

The present invention will obtain double yardstick On-line Control (two scale online by Lyapunov optimization method Control, TS-OC) algorithm.

Present invention introduces a virtual energy sequence, and real-time update, obtain in combination with Lyapunov optimization method Optimum online resource is distributed rationally.

By the dynamic resource Optimal Configuration Method on double yardsticks that the present invention provides, certain user QoS can met In the case of requirement, the purpose of the average total electricity charge of minimizing.

Accompanying drawing explanation

Fig. 1 is the collaborative multi-point communication network configuration diagram powered by intelligent grid.

The schematic flow sheet of double yardstick dynamic resource Optimal Configuration Methods that Fig. 2 provides for the present invention.

Detailed description of the invention

The schematic flow sheet of double yardstick dynamic resource Optimal Configuration Methods that Fig. 2 provides for the present invention.As in figure 2 it is shown, should Method may include that

Step 210, initializes, introduces virtual energy sequence.

Step 220, when each period starts, base station is true by the regenerative resource of forward purchasing market power price and collection The fixed energy size concluded the business in advance with electrical network, base station and electrical network do power exchange, and when the energy having is averagely allocated to this All time slots in Duan.

Step 230, at each time slot, base station is based on Vehicles Collected from Market power price, channel information and forward purchasing energy and electrical network Carry out real-time power transaction, base station design coordinated emission beam shaping, make user QoS reach certain requirement, battery is entered simultaneously Row discharge and recharge.

Step 240, updates virtual energy sequence.

Specific as follows:

Problem (10) is a stochastic optimization problems.The present invention will obtain double yardstick by Lyapunov optimization method and exist Line traffic control (two scale online control, TS-OC) algorithm.First, it is assumed that it is relatively wide to there is following two in system The condition of pine:

$P_b^{max}\≥(1-\mathrm{\η})C^{\min }---\left(11\right)$

$C^{max}-C^{\min }\≥\frac{1-{\mathrm{\η}}^T}{1-\mathrm{\η}}(P_b^{max}-P_b^{min})---\left(12\right)$

Condition (11) represents that the energy leakage of battery can be compensated by charging；Condition (12) represents the energy storage scope of battery Sufficiently large, with reply maximum possible charge or discharge amount in T time slot.

TS-OC algorithm upsets (queue perturbation) parameter Γ and weight parameter V based on two parameters, sequence. NoteBeing required to obtain by the feasibility of algorithm, any pair of (Γ, V) all needs full Foot condition once:

Γ^min≤Γ≤Γ^max, 0 ＜ V≤V^max (13)

Wherein:

${\mathrm{\Γ}}^{\min }:={\max }_{\mathrm{\τ}=1,...,T}\{\frac{1}{{\mathrm{\η}}^{\mathrm{\τ}}}(\frac{1-{\mathrm{\η}}^{\mathrm{\τ}}}{1-\mathrm{\η}}{P}_b^{max}-C^{max})-V\underset{\‾}{\mathrm{\β}}\}---\left(14\right)$

${\mathrm{\Γ}}^{max}:={\min }_{\mathrm{\τ}=1,...,T}\{\frac{1}{{\mathrm{\η}}^{\mathrm{\τ}}}(\frac{1-{\mathrm{\η}}^{\mathrm{\τ}}}{1-\mathrm{\η}}{P}_b^{\min }-C^{\min })-V\stackrel{\‾}{\mathrm{\α}}\}---\left(15\right)$

$V^{max}:={\min }_{\mathrm{\τ}=1,...,T}\left\{\frac{C^{max}-C^{\min }-\frac{1-{\mathrm{\η}}^{\mathrm{\τ}}}{1-\mathrm{\η}}(P_b^{\max }-P_b^{\min })}{{\mathrm{\η}}^{\mathrm{\τ}}(\stackrel{\‾}{\mathrm{\α}}-\underset{\‾}{\mathrm{\β}})}\right\}---\left(16\right)$

TS-OC algorithm performs as follows:

(1) initialize: select suitable Γ and V, introduce a virtual energy sequence

(2) energy forward purchasing: at each period τ=nT, it was observed that stochastic variableAnd determine the size of forward purchasing energyThe size of forward purchasing energy obtains by solving following formula:

Subject to (3), (4), (7), (8),

Wherein, it is desirable to act onOn.Then base station according toElectricity transaction is carried out, it is desirable to electrical network with electrical network To ensuing time slot t=τ ..., τ+T-1 provides averageEnergy.

(3) real-time power equilibrium and beam shaping: at each time slot t ∈ [nT, (n+1) T-1], it is known that AndCan obtain by solving following formula

${\min }_{\{P_i^{*}\left(t\right),P_{b,i}^{*}\left(t\right),w_k^{*}\left(t\right)\}}\underset{i\∈I}{\Σ}\{{\mathrm{VG}}^{rt}\left(P_i\right(t\left)\right)+Q_i\left(nT\right)P_{b,i}\left(t\right)\}---\left(18\right)$

Subject to (3), (4), (7), (8)

Based on obtainBase station and electrical network carry out real-time power exchange；Based onBase station is sent out Penetrate collaborative beam forming；Based onValue carrys out discharge and recharge, then the energy storage in battery is

(4) sequence updates: at each time slot t, updates virtual energy sequence:

It follows that solution is found in respectively problem (17) and (18) by the present invention, obtain TS-OC algorithm.

The target formula of problem (18) is a convex problem.The restrictive condition (3) of SINR can also be write as convex second order cone again (second-order cone, SOC) condition, it may be assumed that

$\begin{array}{c}\sqrt{\underset{l\≠k}{\Σ}|h_{k,t}^H{w}_l\left(t\right){|}^2+{\mathrm{\σ}}_k^2}\≤\frac{1}{\sqrt{{\mathrm{\γ}}_k}}\mathrm{Re}\left\{h_{k,t}^H{w}_k\left(t\right)\right\},\\ \mathrm{Im}\left\{h_{k,t}^H{w}_k\left(t\right)\right\}=0,\∀k\end{array}---\left(19\right)$

Then problem (18) can be re-written as:

$\begin{array}{cc}\min & \underset{i\∈I}{\Σ}\{{\mathrm{VG}}^{rt}(P_c+\underset{k\∈K}{\Σ}{w}_k^H(t\left)B_i{w}_k\right(k)+P_{b,i}(t)-\frac{E_i^{*}\[n_t\]}{T})+Q_i\left(n_{t}T\right)P_{b,i}\left(t\right)\}\}\end{array}---\left(20\right)$

Subject to (4), (7), (19)

Due to G^rt() has convexity and is incremented by, it can be deduced that (P_b,i(t),{w_k(t) })In there is associating convexity.Therefore deduce that problem (20) is individual convex Optimization problem, available common interior point method solution.

For solving problem (17), the present invention proposes a kind of random subgradient method.AssumeChange is independent same over time (independent and identically distributed, the i.i.d.) of distribution, can be in all parameters optimization Index t removes, and problem (17) is re-written as: (wherein Q_i[n] :=Q_i(nT))

$subjectto\begin{array}{c}\sqrt{\underset{l\≠k}{\Σ}|h_k^H{w}_l\left({\mathrm{\ξ}}_t^{rt}\right){|}^2+{\mathrm{\σ}}_k^2}\≤\frac{1}{\sqrt{{\mathrm{\γ}}_k}}\mathrm{Re}\left\{h_k^H{w}_k\left({\mathrm{\ξ}}_t^{rt}\right)\right\},\\ \mathrm{Im}\left\{h_k^H{w}_k\left({\mathrm{\ξ}}_t^{rt}\right)\right\}=0,\∀k,{\mathrm{\ξ}}_t^{rt}\end{array}---\left(21a\right)$

$P_b^{\min }\≤P_{b,i}\left({\mathrm{\ξ}}_t^{rt}\right)\≤P_b^{max},\∀i,{\mathrm{\ξ}}_t^{rt}---\left(21b\right)$

$P_c+\underset{k\∈K}{\Σ}{w}_k^H\left({\mathrm{\ξ}}_t^{rt}\right)B_i{w}_k\left({\mathrm{\ξ}}_t^{rt}\right)\≤P_g^{max},\∀i,{\mathrm{\ξ}}_t^{rt}---\left(21c\right)$

$P_c+\underset{k\∈K}{\Σ}{w}_k^H\left({\mathrm{\ξ}}_t^{rt}\right)B_i{w}_k\left({\mathrm{\ξ}}_t^{rt}\right)+P_{b,i}\left({\mathrm{\ξ}}_t^{rt}\right)=\frac{E_i\[n\]}{T}+P_i\left({\mathrm{\ξ}}_t^{rt}\right),\∀i,{\mathrm{\ξ}}_t^{rt}---\left(21d\right)$

Energy forward purchasing problem (17) only determine optimum energy is pre-purchaseVariable P can be removed further_i, Problem (21) is rewritten as following about E_iThe formula not having restrictive condition of [n]:

${\min }_{\{E_i\[n\]\}}\underset{i\∈I}{\Σ}\[{\mathrm{VG}}^{lt}(E_i\[n\])+T{\stackrel{\‾}{G}}^{rt}\left(\right\{E_i\[n\]\left\}\right)\]---\left(22\right)$

Wherein,

Subject to (21a), (21b), (21c)

For sake of simplicity, definitionFormula (22) is actually One about E_iThe convex problem of the rough and unrestricted condition of [n], it can be by following random subgradient method solution.

First G^lt(E_i[n]) subgradient can be written as:

$\&part;G^{lt}(E_i\&lsqb;n\&rsqb;)=\left\{\begin{array}{ccc}{\mathrm{\&alpha;}}_n^{lt},& if& E_i\&lsqb;n\&rsqb;>L_{i,n}\\ {\mathrm{\&beta;}}_n^{lt},& if& E_i\&lsqb;n\&rsqb;<A_{i,n}\\ anyx\&Element;\&lsqb;{\mathrm{\&beta;}}_n^{lt},{\mathrm{\&alpha;}}_n^{lt}\&rsqb;,& if& E_i\&lsqb;n\&rsqb;=A_{i,n}.\end{array}\right.$

For the optimal solution of problem (23), then about E_i[n],Part subgradient beWherein:

$\&part;{\mathrm{\&Psi;}}^{rt}(E_i\&lsqb;n\&rsqb;,P_{b,i}^E({\mathrm{\&xi;}}_t^{rt}),\{w_k^E\left({\mathrm{\&xi;}}_t^{rt}\right)\left\}\right)=\left\{\begin{array}{cc}\frac{-{\mathrm{\&beta;}}_t^{rt}}{T},& \begin{array}{cc}if& \frac{E_i\&lsqb;n\&rsqb;}{T}>\mathrm{\&Delta;}\end{array}\\ \frac{-{\mathrm{\&alpha;}}_t^{rt}}{T},& \begin{array}{cc}if& \frac{E_i\&lsqb;n\&rsqb;}{T}<\mathrm{\&Delta;}\end{array}\\ x\&Element;\&lsqb;\frac{-{\mathrm{\&alpha;}}_t^{rt}}{T},\frac{-{\mathrm{\&beta;}}_t^{rt}}{T}\&rsqb;,& else\end{array}\right.$

And

DefinitionThen iteration can be declined by the subgradient of a standard Seek the optimal solution of problem (22)

$E_i^{(j+1)}\&lsqb;n\&rsqb;={\&lsqb;E_i^{\left(j\right)}\&lsqb;n\&rsqb;-{\mathrm{\&mu;}}^{\left(j\right)}{\stackrel{\&OverBar;}{g}}_i(E_i^{\left(j\right)}\&lsqb;n\&rsqb;)\&rsqb;}^{+},\&ForAll;i---\left(24\right)$

Wherein, j is iteration mark, { μ^(j)It it is step series.

Realizing formula (24) and need to carry out the computing of higher-dimension, in order to reduce complexity, the present invention proposes a kind of random ladder Degree algorithm: when every time iterating to j, at random from occurring beforeIn choose oneAnd carry out with Lower iteration:

$E_i^{(j+1)}\&lsqb;n\&rsqb;={\&lsqb;E_i^{\left(j\right)}\&lsqb;n\&rsqb;-{\mathrm{\&mu;}}^{\left(j\right)}{g}_i(E_i^{\left(j\right)}\&lsqb;n\&rsqb;)\&rsqb;}^{+},\&ForAll;i---\left(25\right)$

Wherein,And It is that convex problem (23) existsTime the solution that obtains.

It follows that the present invention will carry out performance evaluation, including its feasibility and gradual optimization to the algorithm proposed.

Feasibility:

Notice that restrictive condition (6) is left in the basket in problem (17) and (18).But by rationally selecting in (13) (Γ, V) is right, present invention can ensure thatThe On-line Control algorithm i.e. produced by TS-OC is feasible 's.

If feasibility based on the fact thatThen obtained by TS-OC algorithm Battery charging and discharging electricityMeet:

(1) if C_i(n_tT) ＞-Vβ-Γ, then

(2) ifThen

The above-mentioned fact discloses the part of properties of dynamic TS-OC algorithm.When energy sequence (i.e. battery energy storage) is sufficiently large Time, battery must discharge completelyContrary when energy sequence is sufficiently small, battery is by fully charged

Thus may certify that the feasibility of algorithm: under condition (11)-(12), choose defined in any pair formula (13) (Γ, V), TS-OC algorithm can ensure

Gradual optimization:

For helping to analyze, it is assumed that stochastic processWithIt large and small time scale is the most respectively independent same distribution 's.Definition:

Due toAnd C_i(t+1)=η C_i(t)+P_b,iT (), can obtain:

Again because ofTherefore can be obtained by (26):

$(1-\mathrm{\&eta;})C^{min}\&le;{\stackrel{\&OverBar;}{P}}_{b,i}\&le;(1-\mathrm{\&eta;})C^{max},\&ForAll;i---\left(27\right)$

Problem below considering now:

Subject to (3), (4), (7), (8), (27),

Notice that the restrictive condition in (6) is replaced by (27).Problem (28) is therefore the lax form of problem (10).Tool For body, any feasible solution of problem (10) is the most all the solution of problem (28), i.e.Shifting due to restrictive condition (6) Remove, variable { C_i(t) } it is no longer present in problem (28), other optimized variables " are decoupled " in time.

IfWithIt is independent identically distributed, the control strategy P of a stable state will be there is^stat, it be one only about work as BeforeFunction, meet condition (3), (4), (7) and (8) simultaneously, and ensure to meet at each time slot t:

Wherein,Represent strategy P^statThe charge/discharge electricity amount determined,Represent by strategy P^statThe electricity consumed Take.

Thus may certify that the gradual optimization of algorithm: assumed condition (11)-(13) are set up, ifWithIt is independent with dividing Cloth, then the average electricity charge that TS-OC algorithm produces meet:

Wherein

$M_1:=\frac{IT(1-\mathrm{\&eta;})}{2\mathrm{\&eta;}(1-{\mathrm{\&eta;}}^T)}{M}_B$

$M_2:=\frac{I\&lsqb;T(1-\mathrm{\&eta;})-(1-{\mathrm{\&eta;}}^T)\&rsqb;}{(1-\mathrm{\&eta;})(1-{\mathrm{\&eta;}}^T)}{M}_B$

M₃:=I (1-η) M_C

AndM_C:=max{ (Γ+C^min)²,(Γ+C^max )²}。Represent the electricity charge obtained by TS-OC, Φ^optThen the problem (10) optimal value in all possible strategy, including from Line algorithm.It can be seen that the optimal value that the average electricity charge of TS-OC algorithm generation and off-line algorithm obtain has one to be less thanOptimization spacing.

Feasibility according to algorithm and gradual optimization, may finally be concluded that

Assumed condition (11)-(13) are set up,WithIt is independent identically distributed.The TS-OC algorithm so proposed is permissible Feasible Dynamic Control Strategy, and asymptotic optimization in following degree is produced for problem (10):

Wherein, M:=M₁+M₂+M₃。

Carry out minimizing the analysis optimizing spacing below:

(1) when η=1 (perfect battery), the optimization spacing between TS-OC and off-line algorithm isWhen basis of price spacingThe least, or battery capacity C^maxVery Time big, V^max→ ∞, the optimization spacing that can minimize；

(2) when η ∈ (0,1), constant M₁、M₂And M₃It it is the function about Γ.A given V^max, minimum optimization spacing G^min(V^max) can solve from following point:

${\min }_{(V,\mathrm{\&Gamma;})}\frac{M}{V}=\frac{M_1\left(\mathrm{\&Gamma;}\right)}{V}+\frac{M_2\left(\mathrm{\&Gamma;}\right)}{V}+\frac{M_3\left(\mathrm{\&Gamma;}\right)}{V},s.t.\left(15\right)---\left(30\right)$

It is a convex problem that proof can obtain problem (30), can be solved efficiently by conventional interior point method.Notice G^min(V^max) No longer along with V^max(or C^max) monotone decreasing.The optimization spacing that minimum is likely to be obtained can be by G^min(V^maxTo V on)^max Carry out linear search to calculate.

Through the above description of the embodiments, those skilled in the art is it can be understood that arrive, and the present invention is permissible Realize by the way of add the hardware of necessity by software.Based on such understanding, technical scheme can be with soft The form of part product embodies, and it (can be CD-ROM, U that this software product can be stored in a non-volatile memory medium Dish, portable hard drive etc.) in, including some instructions with so that computer equipment (can be personal computer, server, Or the network equipment etc.) perform method of the present invention.

These are only presently preferred embodiments of the present invention, be not intended to limit protection scope of the present invention.All at this Within bright spirit and principle, any modification, equivalent substitution and improvement etc. made, should be included in protection scope of the present invention Within.

Claims (6)

1. the double yardstick dynamic resource Optimal Configuration Methods in the collaborative multi-point communication network powered by intelligent grid, full In the case of the certain QoS of customer of foot requires, the purpose of the average total electricity charge of minimizing, it is characterised in that including:

Energy method of commerce in advance on the big time scale i.e. period；

Real-time power equilibrium on little time scale i.e. time slot and beam-forming method；

Each period is made up of T time slot；

Energy method of commerce in advance on the described big time scale i.e. period, concretely comprises the following steps:

(1) when each period starts, base station is determined and electrical network by the regenerative resource of forward purchasing market power price and collection The energy size concluded the business in advance；

(2) base station and electrical network do power exchange, and all time slots being averagely allocated in this period by the energy having；

Real-time power equilibrium on the described i.e. time slot of little time scale and beam-forming method, concretely comprise the following steps:

(1) at each time slot, base station carries out real-time energy based on Vehicles Collected from Market power price, channel information and forward purchasing energy with electrical network Amount transaction；

(2) Base Transmitter collaborative beam forming, makes user QoS reach certain requirement；

(3) battery is carried out discharge and recharge.

Method the most according to claim 1, it is characterised in that when each period starts, base station is with intelligent grid in advance The energy expenditure of transaction, i.e. the electricity charge are:

(1)

Wherein() it is base stationiThe regenerative resource collected,For pre-within the next period The energy expenditure of meter,WithIt is respectively energy current bid and ask prices in advance,；

At each time slott, base station with the energy expenditure of intelligent grid real-time deal is:

(5)

Wherein,For base stationiWith the energy of electrical network real-time deal,WithIt is respectively real-time power current bid and ask prices,；

According to formula (1) and formula (5), remember every time slot base stationiThe electricity charge be:

(9)

Order, then optimization problem is as follows:

(10)

Subject to formula (3), (4), (6), (7), (8),

Wherein, formula (3), (4), (6), (7), (8) form are as follows:

Note is for each time slott,For base stationiTo userkChannel vector,For institute Base station is had to arrive userkChannel vector,；WillIt is set to act on userkWave beam Forming vector,For having the transmitting signal of unit energy, then it is sent to userkVector signal be expressed as；UserkThe Signal to Interference plus Noise Ratio (SINR) at place is:

(2)

Wherein,The variance of the multiple Gaussian noise (CSCG) of additivity Cyclic Symmetry for being present in down channel；Then base stationiBiography Delivery of energy is, whereinFromThe row that middle selection is corresponding, constructs base stationiLaunching beam forming vector；In order to ensure user's Service Quality Amount (QoS), meets:

(3)

Wherein,Represent userkTarget sinr values；

Note is for each time slott,For base stationiThe gross energy consumed, it is by transmitting energyAnd constantGroup Become；The upper limit be, it may be assumed that

(4)

NoteFor the stochastic variable under little time scale；

NoteFor base stationiThe initial value of middle battery energy storage,For battery at time slottThe energy storage capacity of start time, battery holds The minima of amount and maximum are respectivelyWith；NoteFor time slottThe electricity of middle battery charge or discharge Amount, then energy storage capacity follows following dynamic change:

(6)

Wherein,Represent energy storage efficiency；The size of charge/discharge electricity amount is limited to every time:

(7)

Note,Represent and round downwards, then have following demand-supply balance equation:

(8).

Method the most according to claim 2, it is characterised in that stochastic optimization problems (10) is used Lyapunov optimization side Method, it is achieved double yardstick On-line Control (TS-OC) algorithms:

Firstly it is assumed that there exists the condition of following two relative loose:

(11)

(12)

Condition (11) represents that the energy leakage of battery can be compensated by charging；Condition (12) represents that the energy storage scope of battery wants foot Enough big, exist with replyTMaximum possible charge or discharge amount in time slot；

TS-OC algorithm upsets parameter Γ and weight parameter based on two parameters, sequenceV, note,, the feasibility of algorithm require to obtain, any pair ofAll need to meet condition:

(13)

Wherein:

(14)

(15)

(16)

TS-OC algorithm performs as follows:

(1) initialize: select suitable Γ andV, introduce a virtual energy sequence；

(2) energy forward purchasing: in each period, it was observed that stochastic variable, and determine the size of forward purchasing energy；The size of forward purchasing energy obtains by solving following formula:

(17)

Subject to(3), (4), (7), (8),；

Wherein, it is desirable to act onOn；Then base station according toElectricity transaction is carried out, it is desirable to electrical network is given with electrical network Ensuing time slotThere is provided averageEnergy；

(3) real-time power equilibrium and beam shaping: at each time slot, it is known that And, obtain by solving following formula:

(18)

Subject to(3), (4), (7), (8)

Based on obtain, base station and electrical network carry out real-time power exchange；Based on, Base Transmitter Collaborative beam forming；Based onValue carrys out discharge and recharge, then the energy storage in battery is；

(4) sequence updates: at each time slott, update virtual energy sequence:。

Method the most according to claim 3, it is characterised in that the target formula of problem (18) is a convex problem, by SINR Restrictive condition (3) write as convex second order cone (SOC) condition, it may be assumed that

(19)

Then problem (18) is written as:

(20)

Subject to(4), (7), (19)

Due toThere is convexity and be incremented by,? In there is associating convexity, thus can go wrong (20) are convex optimization problems, use common interior point method solution.

Method the most according to claim 3, it is characterised in that to problem (17), uses random subgradient method to solve；False IfChange is independent identically distributed over time, thus the index in all parameters optimizationtRemove, problem (17) is write as Following formula:

subject to(21a)

(21b)

(21c)

(21d)

Wherein,；What energy forward purchasing problem (17) only decision was optimum pre-purchase energy, then enter One step removes variable, problem (21) be written as following aboutThe formula not having restrictive condition:

(22)

Wherein,

(23)

Subject to(21a), (21b), (21c)

Definition:；

Formula (22) be one aboutThe rough and convex problem of unrestricted condition, solution is as follows:

First,Subgradient be written as:

For the optimal solution of problem (23), then about,Part subgradient be, wherein:

And；

Definition, then decline iteration by the subgradient of a standard and come Seek the optimal solution of problem (22):

(24)

Wherein,jFor iteration mark,For step series.

Method the most according to claim 5, it is characterised in that solving of formula (24) uses random Subgradient Algorithm: every time Iterate tojTime, at random from occurring beforeIn choose one, and carry out following iteration:

(25)

Wherein,, andIt is that convex problem (23) existsTime the solution that obtains.