CN105071433A - Optimal configuration scheme of distributed power supply - Google Patents

Abstract

The invention discloses an optimal configuration scheme of distributed power supply. According to the method, the benefits of power distribution operators are mainly considered from a planning perspective; the randomness of wind power generation is considered; a multi-target optimal configuration model, with the minimal active electrical energy loss, the minimal total voltage deviation and the minimal risk cost, of a power distribution network is built; multi-target normalization is realized by a fuzzy set theory; the problem that various sub-targets are different in dimension is not generated any more; the built optimal configuration model is determined by an adaptive mutation particle swarm optimization, so that a mutation operation is introduced; the condition that basic particle swarm optimization easily falls into local optimum is improved; finally, as an example, an IEEE 33 node power distribution system validates that the optimal configuration model of DG and selected adaptive particle swarm optimization (AMPSO) disclosed by the invention; and the obtained simulation result shows that the model and the adaptive particle swarm optimization adopted by the scheme are feasible and effective.

Description

A kind of configuration scheme of distributed power source

Technical field

The present invention relates to a kind of configuration scheme of distributed power source, belong to technical field of power systems.

Background technology

Because it has, efficiency is high, scale of investment is little, reduce more and more many being linked in power distribution network of the feature such as network loss, generation mode is flexible, variety of energy sources is various, environmental protection is incorporated into the power networks distributed power source.Distributed power source can be arranged on load center, so that the change of follow load in good time, compare concentration supply power more economical in peak times of power consumption, distributed power source and concentration supply power cooperation, can save power transmission and transformation investment, burning voltage, reduction energy consumption, the quality of power supply improving power system operation, power supply reliability, flexibility and fail safe.

DG accesses the position of power distribution network and access power capacity can plan it and operation produces certain impact, such as have influence on system voltage distribution, voltage stability, the size of short circuit current, running status, relaying protection etc., rationally determine that on-position and the access power capacity of DG have very important effect to the benefit and safe and stable operation that improve power distribution network, according to the feature of power distribution network, meeting under relevant electric power network technique constraints, find scientific and reasonable DG on-position and access power capacity, and the access as far as possible reduced because of DG normally runs brought impact to power distribution network is exactly the main contents that DG distributes rationally.

So traditional distributed power source distributes general only consideration economy or the single aspect of technology rationally, cannot consider multiple target, have certain limitation.

Summary of the invention

The object of the invention is: for the defect of prior art, a kind of configuration scheme of distributed power source is provided, for existing distributed power source access power distribution network provides a kind of new method, to overcome the deficiencies in the prior art.

Technical scheme of the present invention

A kind of configuration scheme of distributed power source, the method is from the angle of planning, the interests of main consideration distribution operator, and the randomness of distributed power source generated output, distributing rationally of distributed power source is studied, namely online position and the capacity of distribution type electric energy is solved, establish minimum by the active energy loss of power distribution network, total voltage deviation is minimum and the minimum multiple-objection optimization allocation models formed of risk cost, arranges constraints and optimization process to above-mentioned multiple-objection optimization allocation models; Adopt fuzzy set theory to achieve multiobject normalization, coordinate the relation between each sub-goal well, and then reach the effect of global optimization; Solved by TSP question particle cluster algorithm.

In the configuration scheme of above-mentioned a kind of distributed power source, described multiple-objection optimization allocation models specifically comprises as follows:

One, active energy loss is minimum;

Distribution network voltage lower grade, R/X value is just larger, so in Load flow calculation, via net loss is just larger, rational access distribution type electric energy can reduce via net loss, therefore, good economic benefit can be brought minimum for via net loss as target function, owing to being the angle from planning herein, then minimum for optimization aim with active energy loss, the minimum target function of active energy loss is:

$\min W_N=\underset{i=1}{\overset{N}{\Σ}}|I_i{|}^2{R}_i\×h$

In formula: I _ifor the electric current of branch road i; N is total circuitry number of system; R _ifor branch resistance, h is the hourage of project period.

Two, the optimization aim model that total voltage deviation is minimum is;

Node voltage amplitude out-of-limit normally will work affecting user and the safety of system, therefore, using total node voltage deviation as target function, node voltage can be made closer to fiducial value, ensure the voltage levvl of power distribution network, reach the effect improving quality of voltage.The minimum target function of total voltage deviation is:

$DetV=\underset{i=1}{\overset{N}{\Σ}}|U_b-U_i|$

In formula: DetV is total voltage deviation; N is number of network node; U _bfor reference voltage; U _iit is the magnitude of voltage of i-th node.

Three, risk cost is minimum;

The effect that the distributed power source of the renewable energy resources such as wind power generation, photovoltaic generation is subject to natural environment or other factors makes it exert oneself to have randomness, fluctuation, probabilistic feature, so herein by risk cost embody distributed power source exert oneself exist randomness, characterized by the expense of electrical network and higher level's Power supply when the energy output of distributed power source does not reach desired value, the target function of risk cost is:

C _risk＝P·C·t·n

In formula: P is its probability being less than and estimating to exert oneself of exerting oneself of DG with randomness; C for when DG exert oneself be less than expectation exert oneself time, by the expense of higher level's Power supply; T is the running time in year of this DG; N is the planning time limit.

In the configuration scheme of above-mentioned a kind of distributed power source, the constraints of described multiple-objection optimization allocation models is inequality constraints, and comprises node voltage constraint, distribution type electric energy access constraint and branch current constraint.

One, node voltage constraint

According to China's operation of power networks engineering philosophy, the voltage bound of 10kV electrical network interior joint must between 1.07p.u. and 0.93p.u., and after access distribution type electric energy, local voltage may be out-of-limit, therefore will ensure that node voltage meets corresponding restrictive condition, node voltage is constrained to:

$U_i^{\min }\≤U_i\≤U_i^{\max },(i=1,2,3...\mathrm{..}N)$

In formula: be respectively the minimum of node i and ceiling voltage, N is number of network node.

Two, the capacity-constrained of distributed power source access electric energy

Distributed power source is exerted oneself because of it and start and stop are not all subject to dispatching of power netwoks, power distribution network can cause very large impact to user when accessing the distribution type electric energy of excess capacity, in order to control because the access of distribution type electric energy is on the impact of power distribution network, need to use restraint to the capacity of distribution type electric energy access, specific as follows:

ΣS _DG≤ηΣS _LD

In formula: S _dGfor accessing the total capacity of the distribution type electric energy of power distribution network; S _lDfor system loading total amount; η is the ratio upper limit that distribution type electric energy total capacity accounts for system loading total amount, and the value of η is 0.4 herein.

Three, restriction of current

$I_i\≤I_i^{\max }$

In formula: it is the upper current limit that i-th branch road allows process.

Equality constraint is node power flow equation:

$\left\{\begin{array}{c}{P}_i-U_i\underset{j=1}{\overset{n}{\Σ}}{U}_j(G_{ij}{\mathrm{cos\θ}}_{ij}+B_{ij}{\mathrm{sin\θ}}_{ij})=0\\ Q_i-U_i\underset{j=1}{\overset{n}{\Σ}}{U}_j(G_{ij}{\mathrm{sin\θ}}_{ij}-B_{ij}{\mathrm{cos\θ}}_{ij})=0\end{array}\right.$

In formula: P _ifor the injection active power of node i; Q _ifor the injection reactive power of node i; G _ijfor the conductance between node i and j; B _ijfor the susceptance between node i and j; U _i, U _jbe respectively the voltage magnitude of node i and node j.

In the configuration scheme of above-mentioned a kind of distributed power source, described employing fuzzy set theory achieves multiobject normalized method and first improves poor index by carrying out maximizing process to the minimum value of multiobject degree of membership, reach the object improving entire system performance, and the membership function adopted is disjunction line function, for the optimization containing multiple target function, each optimization aim conflicting and very difficultly reach optimum often simultaneously, in the importance of each optimization aim all under same case, the maximum satisfaction degree method of usual employing solves, more large population's satisfaction is larger for the membership function minimum value of i.e. optimization aim.

Concrete multiple target method for normalizing is: because active energy loss, total voltage deviation and risk cost dimension are different, when adopting Exchanger Efficiency with Weight Coefficient Method, weight coefficient is not easily determined, and make up poor index by good index obtain by optimizing the possibility of result that obtains, all effects optimized cannot be reached.Therefore, adopt fuzzy set theory to realize multiobject normalization herein, improve poor index by carrying out maximizing process to the minimum value of degree of membership, reach the object improving entire system performance.

One, the membership function of optimization aim

Policymaker can be embodied by the difference of membership function the requirement of target function, and relatively conventional membership function mainly contains: piecewise linear function, power exponent, hyperbola, linear and inverse hyperbolic etc.For above-mentioned several function, linear and piecewise linear function is more conventional by contrast, mainly because inverse hyperbolic, hyperbola and power exponent etc. can increase the non-linear behavior of decision process, make to solve more difficult, select piecewise linear function to be used as membership function to be in this article optimized each sub-goal, membership function is as accompanying drawing 1, and concrete functional form is as follows:

${\mathrm{\&mu;}}_i=\left\{\begin{array}{cc}1& f_i\&le;f_i^{*}\\ 1-\frac{f_i-f_i^{*}}{f_i^{max}-f_i^{*}}& f_i^{*}<f_i<f_i^{max}\\ 0& f_i>f_i^{max}\end{array}\right.$

In formula: μ _ifor target function f _idegree of membership, value, in [0,1], when degree of membership is 1, means that policymaker is to optimum results satisfaction completely; When degree of membership is 0, mean that policymaker is thoroughly unsatisfied with optimum results; f _i ^*for the desired value of target function; f _i ^maxfor the limiting value of target function.

Two, multiple target normalization

For the optimization containing multiple target function, each optimization aim conflicting and very difficultly reach optimum often simultaneously, therefore, generally only requires that each optimization aim reaches optimal value as far as possible in actual applications.All under same case, maximum satisfaction degree method [62] is usually adopted to solve in the importance of each optimization aim.Total satisfaction is defined as:

M＝min{μ ₁,μ ₂,μ ₃}

As can be seen from the above equation, more large population's satisfaction is larger for the membership function minimum value of optimization aim, no matter how the membership function value of two other optimization aim changes, but their membership function value scarcely can be also less than minimum membership function value, thus overall performance can be improved by improving minimum membership function value, entirety can be optimized.So, meeting under corresponding constraints, multi-objective optimization question originally is just converted into the maximized problem of total satisfaction M value:

maxM＝min{μ ₁,μ ₂,μ ₃}。

In the configuration scheme of above-mentioned a kind of distributed power source, the optimal configuration algorithm of distributed power source is solved by TSP question particle cluster algorithm, specific as follows:

One, basic particle cluster algorithm

Assuming that particle cluster algorithm optimizing search in the space that a N ties up, after initialization, algorithm will produce a group random particles, and each particle has corresponding N dimensional vector X _i=(x ₁, x ₂, x ₃..., x _iN) and V _i=(v ₁, v ₂, v ₃..., v _iN) position that represents and speed.Each particle by comparing the quality judging particle by calculating the fitness function value obtained, can be evolved, until find out globally optimal solution by the position and speed constantly changing self.The optimal location that particle searches in iteration each time, is denoted as p _best, the optimal location that whole population searches, is denoted as g _best.

For each iteration, in population, the speed of particle and position are upgraded by following equation:

$\left\{\begin{array}{c}{v}_{in}^{k+1}={\mathrm{\&omega;v}}_{in}^k+c_1{r}_1\&lsqb;p_{best}-x_{in}^k\&rsqb;+c_2{r}_2\&lsqb;g_{best}-x_{in}^k\&rsqb;\\ x_{in}^{k+1}=x_{in}^k+v_{in}^{k+1}\end{array}\right.$

(4-1)

In formula: n represents that N ties up the n-th dimension of solution space, n=l, 2,3 ..., N; K is iterations; ω is inertia weight, is a nonnegative constant; r ₁and r ₂it is the random number produced between [0,1]; c ₁and c ₂for accelerated factor, usual c ₁=c ₂=2.

Can find out from formula (4-1), the velocity variations of particle is divided into three pieces: front portion represent that the speed at a moment on particle is on the impact of present speed, have the balanced capacity to local and global search, be called motional inertia; Part II represent the self thought of particle in search procedure, guiding particle to move towards personal best particle, is the autognosis part of particle; Part III represent and carry out sharing of information between particle, guiding particle to move towards the direction of colony's optimal location, is the social recognition part of particle.By this three part, particle, according to experience and population experience, constantly changes the position of self.

Inertia weight ω, as the parameter of in PSO, has very important effect, and the inertia of its main particle, to the capability of influence of speed, is introduced into the overall situation and local search ability that can control algorithm.ω value is larger, and the search volume of particle will become greatly, and likely expand to new region and search for, ability of searching optimum is stronger; ω value is less, represents that the local search ability of algorithm is comparatively strong, can carry out more detailed search around the current solution searched out; When ω is 0, particle just loses memory function, and the speed with upper a moment has nothing to do by current speed, and so the renewal of particle position is just only by the determining positions of individual pole optimal value and global optimum.

Therefore, allow ω with algorithm iteration process carry out and adaptive line successively decrease, thus significantly improve convergence of algorithm performance, adopt linear decrease weights that inertia weight can be made to be well controlled, the ω of kth time iteration is:

${\mathrm{\&omega;}}_k={\mathrm{\&omega;}}_{\min }+\frac{{\mathrm{\&omega;}}_{max}-{\mathrm{\&omega;}}_{min}}{N}$

(4-2)

In formula: ω _minand ω _maxbe respectively minimum and maximum Inertia Weight; N is maximum iteration time.

Particle cluster algorithm mainly have followed following five basic principles:

L) principle (proximity) is closed on: population must can carry out simple room and time calculating;

2) quality principle (quality): population must be reacted to some extent to the quality factor of surrounding environment;

3) diversity principle (diverseresP0nse): population should behavior in too narrow scope;

4) qualitative principle (stability): population all should not change self behavior when each environment change;

5) adaptability principle (adaptability): under receptible amount of calculation, population needs the behavior that can change them in due course.

Basic particle cluster algorithm step is as follows:

1) position and the speed of each particle in population is produced in an n-dimensional space at random.

2) calculate, at p according to the fitness function value of fitness function to each particle _bestthe current location of each particle of middle storage and fitness function value, at g _bestmiddle position and the fitness function value storing optimal particle in population.

3) speed and the position of each particle is upgraded, according to formula (4-1).

4) by the fitness value of each particle and the p of particle _bestcompare, if be better than p _best, then p is replaced by current particle value _best, otherwise remain unchanged; More current p _bestwith population optimal value g _bestcompare, if be better than g _best, then p is used _bestreplace g _best, otherwise remain unchanged.

5) just stop search if meet the condition of convergence, export globally optimal solution, otherwise turn to step 2).

Two, TSP question particle cluster algorithm

If when algorithm is absorbed in local optimum, can carry out mutation operation, optimal solution is found in other space of removal search, likely can find new p in other space _bestand g _bestcirculate down until find globally optimal solution with this, TSP question particle cluster algorithm (AdaptiveMutationParticleSwarmOptimization, AMPSO) on the basis of this thought, the fitness variance of population is utilized to judge convergence of algorithm degree, thus adaptively changing mutation probability, and by the method for random perturbation, mutation operation is carried out to the extreme value of the overall situation.

(1), Colony fitness variance

The fitness function value of each particle is determined by its position, and so the aggregation extent of each particle just can be reacted by the overall variation of the fitness function value of all particles.When the fitness function value of each particle is close to time consistent, so the fitness variance of population is then tending towards 0.The fitness variance δ of definition population ²for:

${\mathrm{\&delta;}}^2=\underset{k=1}{\overset{n}{\&Sigma;}}{\left(\frac{f_k-f_a}{f}\right)}^2$

(4-3)

In formula: n is population scale; f _kfor the adaptive value of particle k; F is the echo cancellation factor, can limit δ ²size, determined by following formula:

f＝max{max(|f _k-f _a|),1}

(4-4)

F _aaverage fitness for population:

$f_a=\frac{1}{n}\underset{k=1}{\overset{n}{\&Sigma;}}{f}_k$

(4-5)

Can be found out by formula (4-4), the aggregation extent of population is by the fitness variance δ of population ²size react, δ ²numerical value is less, and population is more tending towards convergence; δ ²numerical value is larger, just represents that population is still entering row stochastic search.

(2), mutation operation

As the fitness variance δ of population ²time less, perhaps population has just been absorbed in local optimum, if now carry out TSP question operation according to globally optimal solution, the ability that so particle jumps out locally optimal solution will strengthen, and in other words, mutation probability should with fitness variance δ ²the size of value and changing.Mutation probability is determined by following formula:

P _k＝(P _max-P _min)(δ ²/n) ²+(P _min-P _max)(2δ ²/n)+P _max

(4-6)

In formula: P _max, P _minfor the bound of mutation probability.As can be seen from formula 4-6, δ ²less, P _klarger.

Select the mode of random perturbation to g _bestcarry out mutation operation, α is the stochastic variable that obedience Guass (0,1) distributes, and produces random number r ∈ [0,1], as r < P _ktime just carry out Gaussian mutation:

g _best'＝g _best(1+0.5α)

(4-7)。

Three, solution procedure is distributed rationally based on the distributed power source of AMPSO as follows:

1) initialization; The various parameter informations of input distribution system, comprise line parameter circuit value and load power data, determine corresponding constraints in chapter 3.Initialization PSO parameter, comprises population scale m, particle dimension n, accelerated factor c ₁and c ₂, inertia weight minimum and maximum ω _minand ω _max, and the maximum iteration time k of algorithm;

2) initialization population; The position of stochastic generation particle and velocity original value, herein the representation node position, position of particle, i.e. the position of distribution type electric energy access, speed representation accesses the meritorious and idle of respective nodes;

3) carry out Load flow calculation, calculate the fitness function value of each particle, after fitness function value herein and multiple target change into single goal, be satisfied with angle value;

4) individual and population optimal location is upgraded; By each particle and the p of oneself _bestcontrast, if compare p _bestgood, just replace p _best, otherwise p _bestremain unchanged; By adaptive value p in current population _bestwith g _bestrelatively, if be better than g _best, then g is replaced with it _best, otherwise keep g _bestconstant;

5) fitness variance and the mutation probability of current population is calculated by formula (4-3) and formula (4-6);

6) random generation r ∈ [0,1], as r < P _ktime carry out mutation operation by formula (4-7), otherwise proceed to 7);

7) according to formula (4-1) the more position of new particle and speed;

8) if meet the condition of convergence, stop search, export globally optimal solution, not so turn to step 3);

Distributed power source based on AMPSO is distributed rationally and is solved flow process as shown in Figure 2.

Owing to have employed technique scheme, compared with prior art, the present invention is from the angle of planning, the interests of main consideration distribution operator, taken into account the randomness of wind power generation, establish herein minimum by the active energy loss of power distribution network, total voltage deviation is minimum and the minimum multiple-objection optimization allocation models formed of risk cost, adopt fuzzy set theory to achieve multiobject normalization, no longer include the different problem of each sub-goal dimension, to the Optimal Allocation Model set up herein, adopt TSP question particle cluster algorithm (AdaptiveMutationParticleSwarmOptimization, AMPSO) solve, because which introducing mutation operation, make basic particle group algorithm (ParticleSwarmOptimization, PSO) situation being easily absorbed in local optimum is improved, finally, as example is IEEE33 Node power distribution system, the Optimal Allocation Model demonstrating DG in this paper and the AMPSO selected, the simulation result obtained shows that the model and algorithm adopted is feasible validity herein.

Accompanying drawing explanation

Accompanying drawing 1 is membership function schematic diagram;

Accompanying drawing 2 distributes rationally in the distributed power source of AMPSO to solve schematic flow sheet;

Accompanying drawing 3 is IEEE33 Node power distribution system structural representations;

Convergence property curve when accompanying drawing 4 is blower fan random generating optimization;

Accompanying drawing 5 is node voltage comparison diagrams before and after blower fan is optimized;

Accompanying drawing 6 is convergence property curve synoptic diagrams when adopting PSO algorithm optimization;

Accompanying drawing 7 is convergence property curve synoptic diagrams when adopting AMPSO algorithm optimization.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in further detail, but not as any limitation of the invention.

Embodiments of the invention: in order to validity and the feasibility of the Optimal Allocation Model and TSP question particle cluster algorithm of verifying the distributed power source of proposition, the present embodiment selects IEEE33 Node power distribution system as example, use the programmed environment of MatlabR2009a, program calculation is carried out to the trend of power distribution network and optimized algorithm, makes a concrete analysis of as follows:

One, parameter choose

(1) power distribution network parameter

The present embodiment selects IEEE33 Node power distribution system as tested object, test and access the distribution type electric energy of different capabilities to the impact of distribution system at diverse location, and analysis and summary have been carried out to simulation result, IEEE33 Node power distribution system structural representation as shown in Figure 3, the detailed data of branch impedance and power load distributing is in table 1.1 and table 1.2, this distribution system has 1 power supply point and node 0, totally 33 nodes, article 32, branch road, node 0 is balance node, all the other 32 nodes are load bus, can be used as distributed power source generate electricity can online position, network head end reference voltage is 12.66kV, before not accessing distribution type electric energy, the burden with power of whole network is 3715.0kW, load or burden without work is 2300.0kvar.

Table 1.1 line parameter circuit value table

Table 1.2 load power table

Continued 1.2 load power table

The then capacity-constrained restriction of distributed power source access: the heap(ed) capacity of the distribution type electric energy of distribution system to be accessed should be less than 40% of system loading total amount, the maximum active power so in DG access electrical network is:

P _max＝P _total×40％＝1486kW

(5-1)

(2), algorithm parameter

Population scale m=50 in particle cluster algorithm, particle dimension n=32, maximum iteration time is K=300, accelerated factor c ₁=c ₂=2, the minimum and maximum value of Inertia Weight is respectively 0.9 and 0.4, and in Load flow calculation, blower fan is considered as the PQ node with firm power factor, and that is the DG of connecting system regards the load of " bearing " as in Load flow calculation, and power factor is 0.85.

(3), the parameter of distributed power source

The wind-driven generator of the domestic FD20-200kW model that this example adopts, rated power is 200kW, and incision wind speed Vci is 3m/s, and rated wind speed Vr is 13.8m/s, and cut-out wind speed Vco is 25m/s.

Two, analysis of simulation result

The tidal current computing method in generation is pushed back herein before adopting, when not being optimized, namely during the non-connecting system of distributed power source, the active energy loss being carried out the IEEE33 Node power distribution system that Load flow calculation obtains by matlab is 175.6542kW/h, total voltage deviation is 0.07673, voltage minimum point appears at 17 nodes, and minimum voltage is 0.9219p.u..

(1), the single simulation analysis of exerting oneself at random

Suppose that certain wind energy turbine set has the wind-driven generator of 8 FD20-200kW models, this wind energy turbine set has four to return back out line to be connected with power distribution network, and blower fan is sent out how many electric energy all connecting systems, not by the restriction of distributed power source access power capacity constraints, the optimizing adopting AMPSO to carry out once solves, matlab is utilized to programme, under the wind field environment of locality, by calculating, the form parameter k=2 of the Weibull distribution of the wind speed obedience of this wind field, scale parameter c=15.5, adopt blower fan exert oneself simulation method a stochastic simulation is carried out to exerting oneself of 8 Fans, the result obtained is respectively: 200kW, 184.55kW, 200kW, 182kW, 175.76kW, 148kW, 200kW, 200kW, amount to 1490.31kW.So, when building multiple-objection optimization allocation models, owing to only carrying out an optimizing,, so just there is not risk cost in whole connecting system of exerting oneself of blower fan, it is minimum minimum with total voltage deviation that optimization aim will become active energy loss, optimum results is in table 2.1, the membership function that each sub-goal is corresponding changes in table 2.2, and convergence property curve during algorithm optimization is as accompanying drawing 4, and the node voltage comparing result before optimizing and after optimizing is as accompanying drawing 5.

The random generating optimization result of table 2.1

Note: the front node serial number for distribution type electric energy access of the bracket of program results item in table, the numeral in bracket is the power capacity of this node of access, and unit is: kW.

Table 2.2 sub-goal membership function value changes

As can be seen from table 2.1, after access distribution type electric energy, active energy loss and the total voltage deviation of distribution system obviously decline, and active energy loss drops to 50.8484kW/h from 175.6542kW/h, and total voltage deviation drops to 0.01158 from 0.07673; As can be seen from table 2.2 and accompanying drawing 4, when iterating to the 117th time, maximum satisfaction M value reaches 0.9848, and maximum satisfaction rises to 0.9848 from 0.9270, the membership function of sub-goal is also improving gradually, illustrate and use maximum satisfaction degree method to improve overall performance, and each sub-goal all obtains reasonable optimum results; The comparison diagram of front and back voltage is optimized as can be seen from accompanying drawing 5, node voltage level after optimization significantly improves, more close to fiducial value, the magnitude of voltage of 17 nodes has brought up to 0.9797p.u by 0.9219p.u, improve 6.26%, lowest section point voltage then appears at 32 Nodes, for 0.9636p.u, the numerical value at the lowest section point voltage place before comparatively optimizing also increases, improve the voltage levvl of distribution system entirety, improve economy, in sum, the model and algorithm adopted herein is feasible.

(2) simulation analysis of planned capacity

This section is from the angle of planning, according to the capacity limit of distribution type electric energy access power distribution network, suppose that certain wind energy turbine set only provides the meritorious capacity of 1200kW to system, namely this wind energy turbine set has the wind-driven generator of 6 FD20-200kW models, this wind energy turbine set has three to return back out line to be connected with power distribution network, distributing rationally by distributed power source, determines online position and the access capacity of distribution type electric energy.Expectation so in target function risk cost is exerted oneself then for 200kW, if blower fan is exerted oneself when not meeting 200kW, and will by higher level's Power supply.

1), the contrast of PSO and AMPSO

Under these conditions, carry out the planning of 10 years by a definite date, adopt PSO and AMPSO method to solve the distributed power source Optimal Allocation Model set up before respectively, optimum results only outputs active energy loss and total voltage deviation, and optimum results is in table 2.3.Data in corresponding table 2.3, be optimized the data of rear each node voltage in table 2.4 by PSO and AMPSO, convergence property curve when corresponding PSO and AMPSO is optimized as shown in Figure 6,7.

Can be found out by table 2.3 and 2.4, although total voltage deviation is less than TSP question particle cluster algorithm numerical value in the optimum results of particle cluster algorithm, but difference is little, and in figure 6, time particle cluster algorithm iterates to 33 times, maximum satisfaction M reaches 0.9073, do not change ever since, algorithm has been absorbed in local optimum, does not reach globally optimal solution, and effect of optimization is just poor in view of this, and when adopting AMPSO to be optimized, iterate to the 254th maximum satisfaction M value and just reach optimum 0.9375, be iterations should maximum satisfaction all than employing particle cluster algorithm will, this is the mutation probability owing to introducing the adaptively changing according to Colony fitness variance, (fluctuation repeatedly when also can see that AMPSO optimizes from convergence graph, is had) and changes speed and the position of particle when algorithm is absorbed in locally optimal solution, maintain the diversity of colony, overcome Premature Convergence to a certain extent, obtain and restrain result preferably, describe and adopt AMPSO to be optimized configuration to DG and to have feasibility and validity herein.

The optimum results of table 2.3PSO and AMPSO

Note: the front node serial number for distribution type electric energy access of the bracket of program results item in table, the numeral in bracket is the power capacity of this node of access, and unit is: kW.

Each node voltage (p.u) in table 2.4 system

Each node voltage (p.u) in continued 2.4 system

2), the impact of risk cost

Under the condition that (two) provide, carry out the planning of 10 years by a definite date, respectively with active energy loss, total voltage deviation and with active energy loss, total voltage deviation, risk cost two kinds of situations for target function, namely be divided into and consider risk cost and do not consider that risk cost two kinds of situations are analyzed, adopt AMPSO to be optimized to solve, wherein the optimum results of three target functions is 1) in the optimal result that obtains of AMPSO, the optimum results of Bi-objective and three targets is in table 2.5.

The optimum results of table 2.5 Bi-objective and three targets

As can be seen from table 2.5, Bi-objective is adopted to be optimized to be optimized than three targets distribution system can be made to obtain better effect of optimization, namely the loss of distribution system active energy and the total voltage amount of deflection little, the existence of risk cost is described, namely the randomness that blower fan is exerted oneself is considered, the interests of distribution operator are reduced, but the effect that the distributed power source reflecting the renewable energy resources is really subject to the factors such as natural environment makes it exert oneself have the feature of randomness.

Claims (4)

1. the configuration scheme of a distributed power source, it is characterized in that: the method is from the angle of planning, the interests of main consideration distribution operator, and the randomness of distributed power source generated output, distributing rationally of distributed power source is studied, namely online position and the capacity of distribution type electric energy is solved, establish minimum by the active energy loss of power distribution network, total voltage deviation is minimum and the minimum multiple-objection optimization allocation models formed of risk cost, arranges constraints and optimization process to above-mentioned multiple-objection optimization allocation models; Adopt fuzzy set theory to achieve multiobject normalization, coordinate the relation between each sub-goal well, and then reach the effect of global optimization; Solved by TSP question particle cluster algorithm.

2. the configuration scheme of a kind of distributed power source according to claim 1, is characterized in that: described multiple-objection optimization allocation models specifically comprises as follows:

One, active energy loss is minimum;

Distribution network voltage lower grade, R/X value is just larger, so in Load flow calculation, via net loss is just larger, rational access distribution type electric energy can reduce via net loss, therefore, good economic benefit can be brought minimum for via net loss as target function, owing to being the angle from planning herein, then minimum for optimization aim with active energy loss, the minimum target function of active energy loss is:

$\min W_N=\underset{i=1}{\overset{N}{\&Sigma;}}|I_i{|}^2{R}_i\&times;h$

In formula: I _ifor the electric current of branch road i; N is total circuitry number of system; R _ifor branch resistance, h is the hourage of project period.

Two, the optimization aim model that total voltage deviation is minimum is;

Node voltage amplitude out-of-limit normally will work affecting user and the safety of system, therefore, using total node voltage deviation as target function, node voltage can be made closer to fiducial value, ensure the voltage levvl of power distribution network, reach the effect improving quality of voltage.The minimum target function of total voltage deviation is:

$DetV=\underset{i=1}{\overset{N}{\&Sigma;}}|U_b-U_i|$

In formula: DetV is total voltage deviation; N is number of network node; U _bfor reference voltage; U _iit is the magnitude of voltage of i-th node.

Three, risk cost is minimum;

The effect that the distributed power source of the renewable energy resources such as wind power generation, photovoltaic generation is subject to natural environment or other factors makes it exert oneself to have randomness, fluctuation, probabilistic feature, so herein by risk cost embody distributed power source exert oneself exist randomness, characterized by the expense of electrical network and higher level's Power supply when the energy output of distributed power source does not reach desired value, the target function of risk cost is:

C _risk＝P·C·t·n

In formula: P is its probability being less than and estimating to exert oneself of exerting oneself of DG with randomness; C for when DG exert oneself be less than expectation exert oneself time, by the expense of higher level's Power supply; T is the running time in year of this DG; N is the planning time limit.

3. the configuration scheme of a kind of distributed power source according to claim 1, it is characterized in that: the constraints of described multiple-objection optimization allocation models is inequality constraints, and comprise node voltage constraint, distribution type electric energy access constraint and branch current constraint.

4. the configuration scheme of a kind of distributed power source according to claim 1, it is characterized in that: described employing fuzzy set theory achieves multiobject normalized method and first improves poor index by carrying out maximizing process to the minimum value of multiobject degree of membership, reach the object improving entire system performance, and the membership function adopted is disjunction line function, for the optimization containing multiple target function, each optimization aim conflicting and very difficultly reach optimum often simultaneously, in the importance of each optimization aim all under same case, the maximum satisfaction degree method of usual employing solves, more large population's satisfaction is larger for the membership function minimum value of i.e. optimization aim.