CN106033887A - Power distribution network reconstruction method based on improved PSO-DE hybrid algorithm - Google Patents

Power distribution network reconstruction method based on improved PSO-DE hybrid algorithm Download PDF

Info

Publication number
CN106033887A
CN106033887A CN201510119921.4A CN201510119921A CN106033887A CN 106033887 A CN106033887 A CN 106033887A CN 201510119921 A CN201510119921 A CN 201510119921A CN 106033887 A CN106033887 A CN 106033887A
Authority
CN
China
Prior art keywords
distribution network
population
step
power distribution
de
Prior art date
Application number
CN201510119921.4A
Other languages
Chinese (zh)
Inventor
邵雯
张俊芳
林莎
史媛
褚智亮
毕月
许辉
Original Assignee
南京理工大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 南京理工大学 filed Critical 南京理工大学
Priority to CN201510119921.4A priority Critical patent/CN106033887A/en
Publication of CN106033887A publication Critical patent/CN106033887A/en

Links

Abstract

The invention provides a power distribution network reconstruction method based on an improved PSO-DE hybrid algorithm. A generated initial group is evenly divided into two populations used for particle sward optimization and difference evolution respectively; the speed and position of individuals in the PSO population are updated, and variation, hybridization and selection are carried out on individuals in the DE population; an optimal binary code of a multi-target function is obtained from a load flow calculation result of the corresponding power distribution network of the individuals of the two populations, and an optimal network is obtained according to the code. According to the invention, an optimal solution including construction of the distribution power distribution network can be obtained effectively, and the convergence speed is improved greatly.

Description

A kind of based on the reconstruction method of power distribution network improving PSO-DE hybrid algorithm

Technical field

The invention belongs to power system automatic field, be specifically related to joining of a kind of PSO-DE hybrid algorithm based on improvement Reconfiguration of electric networks method.

Background technology

Due to distributed power source have economy, flexibly, the advantage such as high efficiency, distributed generation technology quickly grow and also by Walk the middle operation that is connected to the grid.Distributed power source accesses in power distribution network, some will certainly be caused to affect the properly functioning of electrical network. And power distribution network operation now tends to automatization, power distribution network reconfiguration is exactly one of its important content.Distributed power source accesses distribution Also will the reconstruct of power distribution network impact, so the power distribution network containing distributed power source has been reconstructed important after net Meaning.

Power distribution network generally is closed-loop structure design, is the state of open loop during operation.Power distribution network comprises a large amount of interconnection switch And block switch, can be by the structure of the disconnected incompatible change network of switch.Power distribution network reconfiguration is i.e. ensureing that whole network is On the premise of radial operation, also need to meet the constraint of quality of voltage requirement and transformer capacity, by changing simultaneously Switch and the folding condition of block switch on interconnection change topology of networks so that it is network loss, balancing the load Index or quality of voltage index are in optimum.Existing reconstruction method of power distribution network has convergence rate slowly, easily fall into mostly The shortcoming entering locally optimal solution.

Summary of the invention

The purpose of the present invention is proposing a kind of reconstruction method of power distribution network based on the PSO-DE hybrid algorithm improved, and it is with one Kind novel, fast convergence rate, the intelligent algorithm that can effectively obtain optimal solution carry out weight to the power distribution network containing distributed power source Structure, is possible not only to effectively obtain the optimal solution containing distributed power source power distribution network reconfiguration, and substantially increases convergence rate

In order to solve above-mentioned technical problem, the present invention proposes a kind of power distribution network weight based on the PSO-DE hybrid algorithm improved Structure method, comprises the following steps:

Step one, initializing and arrange basic parameter, basic parameter includes: parameter population size M in particle cluster algorithm, Maximum iteration time Tmax, maximum inertial factor wmax, minimum inertial factor wmin, controlling elements λ, accelerated factor c1 And c2, parameter maximum zoom factor F in differential evolution algorithmmax, minimum zoom factor Fmin, the mutation probability upper limit CRmaxAnd mutation probability lower limit CRmin

Step 2, generation initial mixing population, be divided into two populations, one of them population by the initial population of generation As the population of particle cluster algorithm, another population is as the population of differential evolution algorithm;

Step 3, the individuality in particle cluster algorithm population is carried out the renewal of speed and position, to differential evolution algorithm population In individuality carry out making a variation, hybridize and select operation;

Step 4, judge whether the distribution network corresponding to individuality in two populations is radial networks respectively, if spoke Penetrate shape net and then proceed step 6, if not radial net then returns step 3;

Step 5, radial networks is carried out Load flow calculation, calculate multiple objective function according to the result of Load flow calculation, according to Multiple objective function obtains adaptive value;

Step 6, in two populations, choose the maximum individuality of adaptive value as optimized individual in mixed population;

Step 7, judging whether to reach maximum iteration time restrictive condition, if reaching, stopping search, export optimum individual And the distribution network parameter that individuality is corresponding, otherwise return step 3.

Further, in step 3, the inertia power used when the individuality in particle cluster algorithm population being carried out speed and updating Shown in weight w such as formula (1):

w = w min + ( w max - w min ) · exp [ - λ ( T T max ) 2 ] - - - ( 1 )

In formula (1), λ is controlling elements, and T is iterations, TmaxFor the maximum iteration time pre-set, wmax For maximum inertial factor, wminFor minimum inertial factor.

Further, in step 3, when carrying out mutation operation, to vector Xr1G d ' the position of () carries out mutation operation, i.e. 1 becomes It is 0,0 to become 1.

Further, in step 3, when intersection operates, newly-generated vectorial Ui(g+1) fooled had N number of position When producing 0, then the binary coding on the most all positions is both configured to 1, stops intersecting operating;When intersecting operation After completing, newly-generated Ui(g+1) when the figure place of upper 0 is less than N, then random polishing becomes N number of 0.

Further, in step 5, shown in described multiple objective function such as formula (2):

Minf=λ1f12f23f3 (2)

In formula (2), f1For distribution network loss, f2For variation index, f3For switching manipulation cost;λ123 It is respectively indices proportion in general objective function.

Further, in step 6, choose each and every one body of preferably M simultaneously and form new mixed population as searching for next time Parent;Described preferably M individuality is the individuality chosen according to the descending arrangement of adaptive value.

Compared with prior art, its remarkable advantage is the present invention, and existing two kinds of intelligent optimization algorithms are carried out by the present invention Merge, and made further to improve to intersection operation and the selection operation of differential evolution algorithm in hybrid algorithm, improve Convergence rate and the ability jumping out locally optimal solution;The present invention carries out weight by reality contains the power distribution network of distributed power source Structure, it was demonstrated that this algorithm also is able to use well in real network;The present invention carries out multiple target reconstruct to power distribution network, Not only can ensure that when power distribution network runs, active power loss is minimum, and it is also ensured that balancing the load index is minimum, voltage is inclined Move volume index minimum, thus ensure that electrical network is in optimal operational condition.

Accompanying drawing explanation

Fig. 1 is the inventive method outline flowchart.

Fig. 2 is the inventive method detail flowchart.

Fig. 3 is power distribution network promise simplification figure during the present invention tests.

Fig. 4 be the inventive method with PSO convergence of algorithm to such as scheming.

Fig. 5 the inventive method and DE convergence of algorithm comparison diagram.

Detailed description of the invention

In general, when power distribution network is reconstructed, need the structure standard of power distribution network is first changed to coding.The present invention will join Electric network composition represents with binary coding, as PSO algorithm (particle cluster algorithm) and DE algorithm (differential evolution Algorithm) in particle and individuality, on power distribution network every branch road switch folding condition represent by binary coding intuitively, Wherein 1 represents branch road Guan Bi, and 0 represents that branch road disconnects, and the structure of whole network then represents by a binary coding, Figure place and the power distribution network number of branches of coding are equal.A multiple objective function optimum is obtained by calculating of the inventive method Binary coding, then draws optimal network according to coding.Specifically comprise the following steps that

Step one, initializing and arrange basic parameter, basic parameter includes: the ginseng in PSO algorithm (particle cluster algorithm) Number population size M, maximum iteration time Tmax, maximum inertial factor wmax, minimum inertial factor wmin, control because of Sub-λ, accelerated factor c1And c2, parameter maximum zoom factor F in DE algorithm (differential evolution algorithm)max, Little zoom factor Fmin, mutation probability upper limit CRmaxAnd mutation probability lower limit CRmin;Wherein,

Population size M, i.e. generates the individuality representated by M initial solution;

Maximum iteration time Tmax, i.e. iterate to exit iterative computation during this number of times;

Maximum inertial factor wmax, minimum inertial factor wmin, the size of inertial factor w affects the present invention side of calculating The Global Attractor of method, controls the impact on rear an iteration speed of the front iteration speed, bigger inertial factor w The ability of searching optimum of PSO can be strengthened, and less inertial factor w can strengthen the local search ability of PSO.

Controlling elements λ, for controlling the parameter of the smoothness of w and T change curve.

Accelerated factor c1And c2, show respectively two and optimize the weight solved.

Maximum zoom factor Fmax, minimum zoom factor Fmin, bigger F can make population more multiformity, now Ability of searching optimum stronger;When having arrived the later stage evolved, F quickly reduces, and the speed of convergence can be accelerated and this Time local search ability stronger.

Mutation probability upper limit CRmax, mutation probability lower limit CRmin;Less mutation probability CR can make population more various Property, ability of searching optimum now is stronger;When having arrived the later stage evolved, mutation probability CR becomes larger, convergence Speed can accelerate and now local search ability stronger.

Step 2, initial population generation module, be divided into two populations, one of them population by the colony being initially generated As the population PSO-DE (PSO) of particle cluster algorithm, another population is as the population PSO-DE of differential evolution algorithm (DE), the initialized location of these two populations lays respectively at different regions;

Step 3, the number counter T=0 of iteration is set;

Step 4, to the particle in PSO-DE (PSO) population, i.e. individuality in PSO-DE (PSO) population is carried out Speed and the renewal of position, make a variation to the individuality in PSO-DE (DE) population, hybridize and select operation;

Wherein, the state of each particle j in PSO-DE (PSO) population is represented by when its position and search speed, position Putting with information vector representation is Xj=(xj1,xj2,...xjM) ', wherein, XjRepresent the position vector that particle j represents, xjn N-th 's of expression particle j is position encoded, and velocity information vector representation is Vj=(vj1,vj2,...vjM) ', VjFor grain The velocity vector that sub-j represents, vjnThe velocity encoded cine of n-th of expression particle j.In search iteration, particle can be by two Individual optimal location constantly adjusts the state of flight of oneself, and one is the optimum that up to the present particle j self is searched Position, represents self human-subject test of particle, is designated as Pj=(pj1,pj2,...pjM) ', PjThe speed represented for particle j Vector, pjnN-th 's of expression particle j is position encoded, and the particle fitness of this position is designated as pbestj;Another is The searched optimal location of the most whole colony, represents social cognition's level of particle, is designated as Gj=(gj1,gj2,...gjM) ', represents that particle j is colony's optimal particle, GjRepresent the position vector that this particle represents, gjn N-th 's of expression jth particle is position encoded, and colony's fitness of this position is designated as gbest.Particle by follow the tracks of and Study the two optimal location constantly changes the search speed of self, determines next step the direction of search, and then updates certainly Position and the speed of body look for optimal solution.

Particle rapidity updates as shown in formula (1):

v jm k + 1 = wv jm k + c 1 rand ( 0,1 ) 1 k ( p best , jm k - x jm k ) + c 2 rand ( 0,1 ) 2 k ( g best , jm k - x jm k ) - - - ( 1 )

Particle position updates as shown in formula (2):

x jm k + 1 = x jm k + v jm k + 1 - - - ( 2 )

In formula (1) and (2):

Represent+1 iteration speed of particle kth;

Represent the position of the search of particle+1 iteration of kth;

W is inertia weight, determine particle present speed inherit number, equalize particle exploring ability and exploitation energy Power, inertia weight strategy is improved by the present invention further, add controlling elements λ control inertia weight w along with Iterations T and the smoothness of curve that changes, the inertia weight after improvement is variable parameter, as shown in formula (3):

w = w min + ( w max - w min ) · exp [ - λ ( T T max ) 2 ] - - - ( 3 )

Wherein λ is controlling elements, and generally this value takes 3.

c1、c2For the accelerated factor more than 0, show respectively two and optimize the weight solved;

WithIt is respectively k+1 the rear individuality of particle j iteration position corresponding to optimal value and colony is optimum Position corresponding to value.

When in PSO-DE (PSO) population, individuality carries out mutation operation, intersection operation and selects operation, first to kind In Qun, each individuality carries out mutation operation, then the individuality after variation carries out intersecting operation, finally to carried out variation, The individuality after operating that intersects carries out selecting operation.Concrete grammar is as follows:

(1) mutation operation

If Xi(g)=(xi1(g),xi2(g),...,xin(g)) represent that the i-th in g generation is individual.Randomly select out from population 3 random individual X in g generationr1(g)、Xr2(g)、Xr3(g) and i ≠ r1 ≠ r2 ≠ r3, its binary coding such as table 1 institute Show:

Any three vector codings of table 1

In mutation operation, represent vector X with Hamming distancesr2(g) and vector Xr3Difference between (g).If Xr2(g) and Xr3G the Hamming distances between () is D (Xr2(g),Xr3(g))=d, F be zoom factor, the distance after scaling takes D '=Fd, because this value is real number, so needing d ' upwardly or downwardly round numbers, i.e. int (d ') or int (d ')+1, Wherein int () is floor operation.The probability distribution that d ' rounds up and down is:

P (d ' → (int) d '+1)=d '-(int) d ' (4)

P (d ' → (int) d ')=1-(d '-(int) d ') (5)

If Xr1(g) make a variation with it after vector Mi(g+1) Hamming distances between is D (Mi(g+1),Xr1(g)), represent Xr1G () vector needs the figure place of variation, calculate as follows:

D ( M i ( g + 1 ) , X r 1 ( g ) ) = ( int ) d &prime; + 1 if rand < d &prime; - ( int ) d &prime; ( int ) d &prime; otherwise - - - ( 6 )

In formula, r is equally distributed random number between [0,1].

Zoom factor is improved by the present invention according to formula (7):

F = F max - ( F max - F min ) ( T T max ) 2 - - - ( 7 )

Wherein, FmaxIf 0.9, FminIt is set to 0.2, TmaxBeing the maximum iteration time set, T is the number of times of current iteration.

Mutation operation has been carried out further improving by the present invention, to vector Xr1G d ' the position of () carries out mutation operation, i.e. 1 becomes 0, and 0 becomes 1.In order to ensure that opening number of switches is N, can set the switch number phase of the switch opened and Guan Bi With, it is all K, wherein shown in K such as formula (8),

Now at random K switch is become 0 from 1 state, K is respectively switched become 1 from 0 at random, so ensure that net Network is the essential condition of radiation network, reduces infeasible solution, increases the efficiency of intersection operation.

(2) intersection operation

The intersection operation of DE algorithm is by the vector M that makes a variationiAnd object vector X (g+1)iG () is intersected, generate a survey Examination vector Ui(g+1).Intersect operation as shown in formula (9):

u ij ( g + 1 ) = m ij ( g + 1 ) rand ij < CR or j = rand ( d ) x ij ( g ) otherwise - - - ( 9 )

In formula, randijFor the equally distributed random number between [0,1];CR is crossover probability, comes according to formula (10) Set:

CR = CR min + ( CR max - CR min ) ( T T max ) 2 - - - ( 10 )

Wherein, CRmaxIf 0.9, CRminIt is set to 0.2, TmaxBeing the maximum iteration time set, T is current iteration Number of times.

Rand (i) is the random integers between [1, n].This intersection operation ensure that Ui(g+1) at least one-component in By Mi(g+1) provide.

Same, when intersection operates, newly-generated vectorial Ui(g+1) when having had N number of position to produce 0 on, then Binary coding on the most all positions is both configured to 1, stops intersecting operating;After intersection operation completes, newborn The U becomei(g+1) figure place of upper 0 is less than N, then random polishing becomes N number of 0.

By the mutation operation improved and the operation that intersects, in the new population that DE produces, infeasible solution greatly reduces.

(3) operation is selected

DE algorithm carries out selecting operation by criterion of " selecting the superior and eliminating the inferior " in natural law, it is possible to constantly to optimum Solve and evolve.If object function is f (X), carry out selecting to operate by formula (11):

X i ( g + 1 ) = U i ( g + 1 ) f ( U i ( g + 1 ) ) < f ( X i ( g ) ) X i ( g ) otherwise - - - ( 11 )

Work as Ui(g+1) fitness is than object vector XiThe when of g the fitness of () is little, Ui(g+1) will replace XiG (), enters the next generation and proceeds to evolve, otherwise then by XiG () is retained in population, continue as in the next generation Individual vector.

Step 5, judge whether the distribution network corresponding to individuality in two populations is radial networks respectively, if spoke Penetrate shape net and then proceed step 6, if not radial net then returns step 4.

Step 6, radial networks is carried out Load flow calculation, according to the result calculating target function of Load flow calculation, according to mesh Scalar functions obtains adaptive value, and adaptive value is the inverse of object function;

The present invention is with multiple objective function for optimizing purpose, and multiple objective function uses weigthed sums approach, and weigthed sums approach is many mesh Mark optimize in most common method, the method, according to each target significance level in optimization problem, determines each mesh Target coefficient, then the single target phase Calais of Weighted Coefficients coefficient is combined into new evaluation function, i.e. a multiple objective function. Adaptive value is the inverse of this multiple objective function.

The present invention considers the multiple target letter after distribution network loss, variation index, three indexs of switching manipulation cost Number is as shown in formula (12):

Minf=λ1f12f23f3 (12)

Wherein: f1For distribution network loss, f2For variation index, f3For switching manipulation cost;λ123Respectively For indices proportion in general objective function, parameter can like acquisition, λ according to policymakeriValue the biggest, represent fiWeight in generic function is the biggest, also implies that this target is the most important.

Adaptive value function:

Step 7, in two populations, choose the maximum individuality of adaptive value as optimized individual in mixed population, select simultaneously Take each and every one body of preferably M and form new mixed population as the parent searched for next time;Described preferably M individuality is The individuality chosen according to the descending arrangement of adaptive value;

Step 8, discriminating whether to reach maximum iteration time restrictive condition, if reaching, stopping search, export optimum individual And the distribution network parameter that individuality is corresponding, otherwise return step 3~eight and continue search for.

The present invention can be further illustrated by following experiment:

This certain section is 10kV power distribution network, has 8 buses, 7 union switch, 23 regular taps, altogether 30 Bar branch road, total load is 7806.7+j3030kVA, owing to the distribution situation of every circuit of actual power distribution network is extremely complex, So network is simplified.The process simplified: close load is merged, and will merge at the circuit of same switch Same circuit, deletes less branch, only retains main line and important branch road.Network after simplification not only comprises all of Switch, and network structure is simple, can effectively judge network structure.Each for this network node is numbered, should 8 feeder lines of network stand alone as load and power, and in order to effectively differentiate the radioactivity of network, this power distribution network do an equivalence, All feeder line root nodes are merged into a node, and whole power distribution network can be considered that one tree is powered.After equivalence, this network is 24 nodes, branch road 30, is wherein a Miniature wind electric field at 9 nodes, and wind energy turbine set is directly accessed as DG joins In electrical network.Network is as it is shown on figure 3, the circuit of power distribution network and node parameter, and parameter is as shown in table 2, table 3.

Table 2 power distribution network branch parameters

Branch number Branch road start node Branch road end-node Branch resistance Branch road reactance 2-1 1 2 0.425 0.85 2-2 2 3 0.0255 0.051 2-3 3 4 0.0085 0.017 2-4 4 5 0.0595 0.119 2-5 5 6 0.0476 0.0952 2-6 6 7 0.0255 0.051 2-7 7 8 0.1773 0.3546 2-8 1 8 0.1955 0.391 2-9 6 9 0.1907 0.118 2-10 1 10 0.2861 0.1677 2-11 10 11 0.5722 0.3354 2-12 1 12 0.2988 0.1751 2-13 12 13 0.5404 0.3168 2-14 13 14 0.17 0.34 2-15 14 15 0.1275 0.255 2-16 15 16 0.1105 0.221 2-17 1 17 0.034 0.068 2-18 13 17 0.102 0.204 2-19 1 18 0.425 0.85 2-20 18 19 0.255 0.51 2-21 19 20 0.017 0.034 2-22 20 21 0.3179 0.1863 2-23 21 22 0.17 0.34 2-24 1 22 0.3349 0.6698 2-25 1 23 0.4352 0.8704 2-26 23 24 0.0908 0.1816 2-27 21 24 0.0187 0.0374

2-28 2 15 0.3179 0.1863 2-29 8 10 0.3179 0.1863 2-30 6 21 0.3179 0.1863

Table 3 power distribution network node parameter

Node serial number Node active power Node reactive power 1 0 0 2 189.7 100 3 154.1 50 4 47.4 20 5 179.6 60 6 534.6 110 7 385.3 20 8 819.6 400 9 331.3 100 10 193.2 20 11 254.3 100 12 47.4 10 13 1015.4 500 14 301.7 100 15 84.8 20 16 246 100 17 84.8 70 18 683.4 220 19 647.3 320 20 364.5 150 21 325.4 20 22 481.3 300 23 71.1 50 24 364.5 200

When Miniature wind electric field input power is P=300kW, during Q=225kW, with the present invention, this network is carried out multiple target Power distribution network reconfiguration, initialize and arrange basic parameter: initial population scale M is 30, maximum iteration time TmaxFor 50, Studying factors C1=C2=2, wmax=0.9, wmin=0.4, λ=3, Fmax=0.9, Fmin=0.2, CRmax=0.9, CRmin=0.2.

Owing to this experiment having 8 transformer station's outlet lines, when 8 transformators are transmitted electricity for this power distribution network simultaneously, circuit 1, these 8 switches at 8,10,12,17,19,24,25 places must be Guan Bi, needs to set in a program this The vector value perseverance at several switch places is 1.The reconstruction result obtained takes optimal solution, as shown in table 4:

Table 4 power distribution network reconfiguration result

From the results, it was seen that the on off state before and after Chong Gou only has 2 to change, but distribution network loss is the most significantly Reduce, be reduced to 35.0154kW from original 64.7913kW, reduce 42.0%.Lowest section point voltage is from 0.9790 Being raised to 0.9861, voltage deviation is reduced to 0.0010 by 0.0026, and the reliability that power distribution network is powered there has also been the biggest proposing Rise.Visible power distribution network reconfiguration also can be used in real network well, not only makes network loss reduce, also improves power distribution network The reliability of power supply, effect is obvious.

The present invention uses PSO algorithm, DE algorithm and the inventive method to be reconstructed power distribution network, and receive the most respectively Hold back the contrast of speed.Conventional method advantage contrast is all separate computations, calculates repeatedly, comes with the average of iterations Contrast.But due to the impact of initial population, and the randomness during calculating, the speed of iteration and convergence The each difference of effect is very big, inconspicuous with average contrast effect.In order to avoid twice calculates initial population to algorithm receipts The impact held back, it is ensured that the concordance of initial population.The present invention carries out two kinds of algorithms simultaneously, i.e. divides according to same initial population Do not carry out follow-up calculating and iteration.Record obtains every generation optimal solution during calculating, and makees according to the optimal solution of every generation Go out to restrain comparison diagram.

The inventive method contrasts as shown in Figure 4 with PSO convergence of algorithm, and the inventive method contrasts with DE convergence of algorithm As shown in Figure 5.From aforementioned convergence comparison diagram, for the PSO-DE algorithm of same initial population and PSO algorithm, DE algorithm carries out calculating and compares, and PSO-DE hybrid algorithm convergence rate is significantly faster than PSO algorithm and DE algorithm, and And it is more readily available globally optimal solution.Illustrate that the inventive method is effective, and with the obvious advantage.

Claims (6)

1. a reconstruction method of power distribution network based on the PSO-DE hybrid algorithm improved, it is characterised in that comprise the following steps:
Step one, initializing and arrange basic parameter, basic parameter includes: the parameter population size M in particle cluster algorithm, maximum iteration time Tmax, maximum inertial factor wmax, minimum inertial factor wmin, controlling elements λ, accelerated factor c1And c2, parameter maximum zoom factor F in differential evolution algorithmmax, minimum zoom factor Fmin, mutation probability upper limit CRmaxAnd mutation probability lower limit CRmin
Step 2, generation initial mixing population, be divided into two populations by the initial population of generation, and one of them population is as the population of particle cluster algorithm, and another population is as the population of differential evolution algorithm;
Step 3, the individuality in particle cluster algorithm population is carried out the renewal of speed and position, the individuality in differential evolution algorithm population is made a variation, hybridizes and selected operation;
Step 4, judge whether the distribution network corresponding to individuality in two populations is radial networks, if radial net then proceeds step 6, if not radial net then returns step 3 respectively;
Step 5, radial networks is carried out Load flow calculation, calculate multiple objective function according to the result of Load flow calculation, obtain adaptive value according to multiple objective function;
Step 6, in two populations, choose the maximum individuality of adaptive value as optimized individual in mixed population;
Step 7, judging whether to reach maximum iteration time restrictive condition, if reaching, stopping search, output optimum individual and individual corresponding distribution network parameter, otherwise return step 3.
2. reconstruction method of power distribution network based on the PSO-DE hybrid algorithm improved as claimed in claim 1, it is characterised in that in step 3, shown in the inertia weight w such as formula (1) used when the individuality in particle cluster algorithm population being carried out speed and updating:
In formula (1), λ is controlling elements, and T is iterations, TmaxFor the maximum iteration time pre-set, wmaxFor maximum inertial factor, wminFor minimum inertial factor.
3. reconstruction method of power distribution network based on the PSO-DE hybrid algorithm improved as claimed in claim 1, it is characterised in that in step 3, when carrying out mutation operation, to vector Xr1G d ' the position of () carries out mutation operation, i.e. 1 becomes 0, and 0 becomes 1.
4. reconstruction method of power distribution network based on the PSO-DE hybrid algorithm improved as claimed in claim 1, it is characterised in that in step 3, when intersection operation, newly-generated vectorial Ui(g+1) it is taken in when having had N number of position to produce 0, then the binary coding on the most all positions is both configured to 1, stop intersecting operation;After intersection operation completes, newly-generated Ui(g+1) when the figure place of upper 0 is less than N, then random polishing becomes N number of 0.
5. reconstruction method of power distribution network based on the PSO-DE hybrid algorithm improved as claimed in claim 1, it is characterised in that in step 5, shown in described multiple objective function such as formula (2):
Minf=λ1f12f23f3 (2)
In formula (2), f1For distribution network loss, f2For variation index, f3For switching manipulation cost;λ123It is respectively indices proportion in general objective function.
6. reconstruction method of power distribution network based on the PSO-DE hybrid algorithm improved as claimed in claim 1, it is characterised in that in step 6, choose each and every one body of preferably M simultaneously and form new mixed population as the parent searched for next time;Described preferably M individuality is the individuality chosen according to the descending arrangement of adaptive value.
CN201510119921.4A 2015-03-18 2015-03-18 Power distribution network reconstruction method based on improved PSO-DE hybrid algorithm CN106033887A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510119921.4A CN106033887A (en) 2015-03-18 2015-03-18 Power distribution network reconstruction method based on improved PSO-DE hybrid algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510119921.4A CN106033887A (en) 2015-03-18 2015-03-18 Power distribution network reconstruction method based on improved PSO-DE hybrid algorithm

Publications (1)

Publication Number Publication Date
CN106033887A true CN106033887A (en) 2016-10-19

Family

ID=57148704

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510119921.4A CN106033887A (en) 2015-03-18 2015-03-18 Power distribution network reconstruction method based on improved PSO-DE hybrid algorithm

Country Status (1)

Country Link
CN (1) CN106033887A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106408135A (en) * 2016-10-26 2017-02-15 重庆邮电大学 Power system optimal power flow method based on feedback learning cuckoo algorithm
CN106654319A (en) * 2016-12-27 2017-05-10 东南大学 Temperature modeling method for proton exchange membrane fuel cell (PEMFC) system based on variation particle swarm and differential evolution hybrid algorithm
CN106680669A (en) * 2016-12-24 2017-05-17 国网吉林省电力有限公司培训中心 Power distribution network fault section positioning method based on binary mixing algorithm
CN107017640A (en) * 2017-06-12 2017-08-04 广东工业大学 A kind of optimal load flow computational methods of power system, apparatus and system
CN108390407A (en) * 2018-02-28 2018-08-10 国家电网公司 Distributed photovoltaic access amount computational methods, device and computer equipment
CN109739086A (en) * 2018-12-28 2019-05-10 浙江大学 Propylene polymerization production process optimal online forecasting system based on Recognition with Recurrent Neural Network

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007288877A (en) * 2006-04-14 2007-11-01 Hitachi Ltd Power quality maintenance supporting method and system for power distribution system in linkage with a plurality of decentralized power sources
CN104332995A (en) * 2014-11-14 2015-02-04 南京工程学院 Improved particle swarm optimization based power distribution reconstruction optimization method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007288877A (en) * 2006-04-14 2007-11-01 Hitachi Ltd Power quality maintenance supporting method and system for power distribution system in linkage with a plurality of decentralized power sources
CN104332995A (en) * 2014-11-14 2015-02-04 南京工程学院 Improved particle swarm optimization based power distribution reconstruction optimization method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
刘丹丹: "基于改进粒子群算法的电力系统无功优化", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
周湶等: "基于化整为零策略和改进二进制差分进化算法的配电网重构", 《电网技术》 *
宋刚: "基于粒子群和差分进化算法的新型混合算法在配电网重构中的应用", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
栾丽君等: "一种基于粒子群优化算法和差分进化算法的新型混合全局优化算法", 《信息与控制》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106408135A (en) * 2016-10-26 2017-02-15 重庆邮电大学 Power system optimal power flow method based on feedback learning cuckoo algorithm
CN106680669A (en) * 2016-12-24 2017-05-17 国网吉林省电力有限公司培训中心 Power distribution network fault section positioning method based on binary mixing algorithm
CN106654319A (en) * 2016-12-27 2017-05-10 东南大学 Temperature modeling method for proton exchange membrane fuel cell (PEMFC) system based on variation particle swarm and differential evolution hybrid algorithm
CN106654319B (en) * 2016-12-27 2019-03-19 东南大学 A kind of PEMFC system temperature modeling method based on mutation particle swarm and differential evolution hybrid algorithm
CN107017640A (en) * 2017-06-12 2017-08-04 广东工业大学 A kind of optimal load flow computational methods of power system, apparatus and system
CN107017640B (en) * 2017-06-12 2019-11-22 广东工业大学 A kind of optimal load flow calculation method of electric system, apparatus and system
CN108390407A (en) * 2018-02-28 2018-08-10 国家电网公司 Distributed photovoltaic access amount computational methods, device and computer equipment
CN109739086A (en) * 2018-12-28 2019-05-10 浙江大学 Propylene polymerization production process optimal online forecasting system based on Recognition with Recurrent Neural Network

Similar Documents

Publication Publication Date Title
Barthélemy et al. Modeling urban street patterns
Zhu Optimal reconfiguration of electrical distribution network using the refined genetic algorithm
Lancichinetti et al. Benchmarks for testing community detection algorithms on directed and weighted graphs with overlapping communities
Nguyen et al. Distribution network reconfiguration for power loss minimization and voltage profile improvement using cuckoo search algorithm
Sheng et al. Optimal placement and sizing of distributed generation via an improved nondominated sorting genetic algorithm II
Wang et al. Optimization of network configuration in large distribution systems using plant growth simulation algorithm
Sum-Im et al. Differential evolution algorithm for static and multistage transmission expansion planning
Marinakis et al. Honey bees mating optimization algorithm for the Euclidean traveling salesman problem
Chandrasekaran et al. Network and reliability constrained unit commitment problem using binary real coded firefly algorithm
Duan et al. Reconfiguration of distribution network for loss reduction and reliability improvement based on an enhanced genetic algorithm
Niknam et al. Distribution feeder reconfiguration considering fuel cell/wind/photovoltaic power plants
Yuan et al. Application of cultural algorithm to generation scheduling of hydrothermal systems
Narimani et al. Enhanced gravitational search algorithm for multi-objective distribution feeder reconfiguration considering reliability, loss and operational cost
Carpaneto et al. Distribution system minimum loss reconfiguration in the hyper-cube ant colony optimization framework
Marinakis et al. A new bilevel formulation for the vehicle routing problem and a solution method using a genetic algorithm
Jabr Optimization of AC transmission system planning
Song et al. Advanced engineered-conditioning genetic approach to power economic dispatch
Swarnkar et al. Adapted ant colony optimization for efficient reconfiguration of balanced and unbalanced distribution systems for loss minimization
CN102611119A (en) Multi-target reactive power optimization method for electric system
CN103345657B (en) Method for scheduling task based on heredity and ant group under cloud computing environment
Miranda et al. Differential evolutionary particle swarm optimization (DEEPSO): a successful hybrid
Abdelaziz Genetic algorithm-based power transmission expansion planning
Afshar et al. Non-dominated archiving multi-colony ant algorithm for multi-objective optimization: Application to multi-purpose reservoir operation
CN103679544A (en) Comprehensive assessment method for running of intelligent power distribution network
Yammani et al. A Multi-objective Shuffled Bat algorithm for optimal placement and sizing of multi distributed generations with different load models

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20161019

RJ01 Rejection of invention patent application after publication