Background technology
The research of the anti-error optimization method of power distribution network was risen in the eighties of last century later stage eighties, because it receives many scholars' concern in the important function that reduces network loss and improve the aspect such as system safety.Early stage For Distribution Networks Reconfiguration mainly is the reconstruction of research distribution planning stage.Along with the progressively intensification to power distribution network reconfiguration understanding, scholars find to add in the electrical power distribution automatization system network reconfiguration not only economical and technical feasible, and can greatly optimize the operation of distribution system.
Power distribution network when normal operation radially, determine that by network structure tree is not unique, so that power distribution network determines that by network structure tree-shaped operational mode is not unique, the switch folding condition can form multiple combination, and arbitrary combination all consists of a kind of operational mode.Although interconnection switch disconnects when normal operation, with the tree-shaped running status of the Radiation of keeping system, its existence is so that the webbed structure of system.The essence of distribution reconstruct is exactly to satisfy under certain constraints, by changing on off state in the network, optimize the network configuration of power distribution network, thereby the trend of improving distribution system distributes, ideal situation is to reach optimal load flow to distribute, and makes distribution system loss minimization or other indexs optimum.
Different operational mode correspondences different trends and is distributed, and causes different via net loss, so just exists the problem of economical operation, namely in the combination that consists of tree-shaped operational mode, have a kind of combination, move by the mode of this combination, a certain index of network is optimum.Different operational mode correspondences different trend distributions, via net loss and network operation reliability, thereby produced the problem that network configuration is optimized, namely for the multiple radial operational mode of distribution, exist a kind of operational mode can be so that a certain index of network is optimum.
Number of switches is huge in the power distribution network of reality, so the anti-error optimization of distribution network is extensive a, MIXED INTEGER, multiple target, nonlinear combinatorial optimization problem.One of method of processing multi-objective optimization question is exactly the dimensionality reduction optimization method, namely selects a main target function, and other target is processed as constraint.
Existing algorithm is target function mainly with loss minimization greatly, is satisfying under the various service conditionss, and the anti-error optimization of the power distribution network take loss minimization as target function is still a non-linear hybrid optimization problem.Calculate because the nonlinear characteristic of the anti-error optimization of power distribution network, the iteration that is optimized each time all need to carry out the primary distribution trend, continuous distribution power system load flow calculation needs a large amount of computing times.In order to improve computational speed, assurance draws the Distributing network structure of global optimum or suboptimum, main reconstruction and optimization algorithm has as follows at present: branch exchange method (Branch Exchange Method, BEM), optimal flow pattern (Optimal Flow Pattern, OFP), expert system approach (Expert System, ES), artificial neural network method (Artificial Neural Networks, ANN), simulated annealing (Simulated Annealing, SA), genetic algorithm (Genetic Algorithm, GA), tabu search algorithm (Tabu Search, TS) etc.
The power distribution network topological structure radially, radial structure is not unique so that power distribution network determines that by network structure tree-shaped operational mode is not unique, the switch folding condition can form multiple combination, arbitrary combination all consists of a kind of operational mode.In the distribution system of reality, the permutation and combination number of switching manipulation is very huge, so the anti-error optimization problem of power distribution network is a huge non-linear integer combinations optimization problem in theory.Because as the switch combination enormous amount of optimized variable, exhaustive search will face " multiple shot array " problem.And solution space is too huge, so that amount of calculation is very large when direct Mathematical, thereby will take a large amount of machines the time, and can't guarantee the reliability that restrains.
Embodiment
The anti-error optimization method of a kind of power distribution network comprises the steps:
One, the power distribution network topological system is divided into several independently unit according to current running status, guarantees that each unit has and only have a switch to disconnect, the quantity of unit equals the number of switches of the state of opening the light for disconnecting like this;
Two, the on off state with each unit carries out real coding by the switch exchange algorithm, all unit is connected into a chromosome again;
Three, by target setting function and constraints, with Chaos Genetic Algorithm network is calculated, thereby try to achieve optimal solution.
During enforcement
The Mathematical Modeling of 1 anti-error optimization
1.1 the target function of anti-error optimization
The target function that the present invention adopts is to reduce matching net wire loss with minimum on-off times, makes the loss minimization of network.Its network loss target function can be expressed as:
In the formula, b is a way of power distribution network; K is the running status of branch road i, and 0 this branch road of expression disconnects, this branch road operation of 1 expression; R is the resistance of branch road i; I is the electric current by branch road i.
Network loss is calculated by trend and is tried to achieve.Owing to be open loop when distribution normally moves, so push back for the method Load Flow before the present invention adopts.On-off times draws by the state contrast of front-rear switch, and last general objective function representation is:
(2)
In the formula, the income of F for optimizing, m is front network loss, and f is the network loss after optimizing, and k1 is the unit electricity price, and k2 is each switch motion cost, n is wanted the switch motion number of times by optimization.
1.2 the constraints of optimizing
The anti-error optimization of power distribution network also requires formula (1) should satisfy following constraints:
(1) power capacity constraint:
,
In the formula,
,
Be respectively power calculation value and maximum power value thereof that each branch road i flows through,
,
What be respectively transformer confesses power and maximum capacity thereof;
(2) node voltage constraint:
In the formula,
,
Be respectively the permission minimum value of i node voltage and the maximum of permission;
(3) network configuration constraint: if the network open loop after the optimization, looped network can not appear;
(4) power supply constraint: all loads all will have Power supply belt, the isolated island situation can not occur;
(5) trend constraint: the anti-error optimization of distribution will be satisfied power flow equation.
The anti-error optimized algorithm of 2 distributions
2.1 power distribution network network structure is divided ring
So-called loop refers to form on the grid structure circuit of closed loop in the anti-error optimization of distribution, and two kinds of forms are arranged: 1, from a power supply point of power distribution network, each point arrives the ring of another power supply point only through 1 time; 2, from certain point of power distribution network, each node is got back to again the ring of this point only through 1 time.The running status of the present invention before the anti-error optimization of power distribution network divided ring as the basis, guarantee that every loop has and only have a switch that disconnects under current running status.
2.2 improved genetic algorithm
Genetic algorithm at first will form chromosome, then copies, the work such as crossover and mutation, and selecting at last the highest new individuality of fitness is preferred plan.
2.2.1 chromosome forms and optimizes
The anti-error optimization of distribution is reconstructed when optimizing with traditional genetic algorithm, splits to close to be numbered, and the on off state in the network is opened with 0() or 1(close) expression, each on off state accounts for chromosomal one, chromosomal length equals the sum of switch.
If because the network after the anti-error optimization of power distribution network network will satisfy network open loop after the optimization, all loads all will have Power supply belt.So in the network some must closed switching branches can be removed from loop according to these 2, the switching branches that will link to each other with power supply point is removed from loop.Having removed some must closed branch road, and chromosomal length reduces.
The interconnection switch number that disconnects before the network optimization will equate with the interconnection switch number that disconnects after the network optimization.The present invention is by the search of network topology, and the network operation state based on before optimizing through optimizing, is divided into some loops with network, and the interconnection switch number of loop number and disconnection is the same.Pressing each loop is that body is arranged sequentially good with what fix with all loops one by one, then respectively the on off state of every ring is encoded respectively.Because each loop is that to have and only have a switch be to be in off-state, other switches must be in closure state, carry out real coding according to these characteristics.The position of each switch is fixed in the loop, in the loop i switch be disconnect then this loop is encoded to i (i is natural number), the span of i is that 1~N(N is the total number of switch in the loop).All loops all carry out real coding according to the method described above, then according to the order between existing loop whole network loop are integrated, and obtain the chromosome of a real coding.
The below illustrates real coding as an example of ieee standard 3 feeder lines 16 node distribution systems example.
As shown in Figure 1, the on off state of branch road 1, branch road 7, branch road 10, branch road 15 can be removed from chromosome according to the principle of optimizing.Divide the ring operation with remaining branch road, the below obtains a kind of minute ring scheme:
Loop 1: branch road 3-4-6-8; Loop 2: branch road 9-13-14; Loop 3: branch road 2-5-11-12-16
In Fig. 1 loop 1, the switch of branch road 4 disconnects, and is 1011 with traditional binary coding representation, and can be expressed as 2 with the real coding mode, and the transformation range of state encoding is 1~4.In like manner the state of loop 2 is expressed as 2 with real coding, and the transformation range of state encoding is 1~3; The state of loop 3 is expressed as 3 with real coding, and the transformation range of state encoding is 1~5; Arranged sequentially according to loop 1 loop 2 loops 3, network operation state can represent with real coding 223 among Fig. 1.
Upper example can be found out, adopt a minute ring real number coding method that chromosomal length is reduced on a large scale, chromosomal all codings all are feasible codings, the coding of the normal operation of distribution can not occur not meeting, so need not the chromosome that does not meet the distribution operation is revised.This coded system has significant improvement the efficient of the anti-error optimized Genetic Algorithm of distribution.
2.2.2 chromosomal genetic manipulation
The clone method that the present invention adopts is squirrel wheel method.Copy and finish, will carry out interlace operation behind the formation population.What the present invention adopted is the real coding of minute loop, and interlace operation occurs between loop and the loop.Below or take the interlace operation of the present invention as example illustrates of ieee standard 3 feeder lines, 16 node distribution systems.
Suppose that through real coding copy operation after the pairing, the right chromosome of an assembly occurs: A:234 B:322 in twos; Crossover location is chosen in second, then forms two new chromosome A1:222 B1:334.Concrete operations as shown in Figure 2.
Chromosome after gene morphs just mistake may occur, does not meet the operation condition of network of distribution.Occur to revise chromosome after the chromosome mistake.It is that number of switches is out-of-limit that mistake appears in chromosomal variation of the present invention, will be with the loop transcoding, coding transform of mistake in the transformation range of state encoding after out-of-limit.Above routine chromosome A1 is example.The second of A1 morphs, the chromosome A1:252 that the formation after the variation is new.And deputy excursion is 1~3, so will revise second, second is transformed to 1~3 random number, revised chromosome A1:232.The network operation state that obtains meets the network configuration of distribution.
2.2.3 Chaos Genetic Algorithm
Chaos refers to occur in the deterministic system seems to be at random irregular movement.Defective based on traditional genetic algorithm, the present invention proposes a kind of Chaos Genetic Algorithm, chaotic optimization algorithm is joined in the middle of the genetic algorithm as additional factor, the individuality that fitness in the population is relatively poor replaces to chaos emigration, increase randomness and the ergodic of gene individuality in the population, promote effective evolution of population, prevent that algorithm is absorbed in local optimum.Great many of experiments shows that when colony's total scale was 50~100, immigrant's ratio should select 0.2~0.4, and immigrant's ratio that the present invention chooses is 0.2.
2.2.4 locally optimal solution solution
What the present invention adopted is the branch exchange algorithm, because there is limitation in the branch exchange method, although i.e. branch exchange algorithm fast convergence rate is absorbed in locally optimal solution easily.Now provide a kind of solution that prevents from being absorbed in locally optimal solution.Divide when ring carrying out power distribution network network structure, because the branch road of network configuration much causes the result who minute encircles not unique.And when being optimized with the above-mentioned algorithm of the present invention, just obtain an optimal solution on a minute good loop basis, divide ring structure then helpless to other different networks.This solution is considered the diversity that network divides ring, specifically by Fig. 3 explanation.1 ' is power supply point among Fig. 3, and branch road 1-2 is for must then dividing ring not consider by closing section.Divide the ring method can obtain following a kind of scheme by the present invention: A loop 1: branch road 2 '-8 '-9 '-3 '-2 '; B loop 2: branch road 2 '-5 '-6 '-3 '; C loop 3: branch road 9 '-10 '-4 '-3 '; D loop 4: branch road 6 '-7 '-4 '.
From such scheme, can find out, the node that occurrence number is maximum is 3 ' node, occur altogether 3 times, and 3 ' node connects 4 branch roads, the network topology of Fig. 4 can guarantee that all switch situations of cut-offfing can represent with these three kinds of schemes with three kinds of different minute ring schemes.Can infer thus each network the switch situation of cut-offfing can with N-1(N be connect in the network that the maximum node of branch road connects way) plant different minute ring schemes and represent.This N-1 kind scheme is calculated by algorithm of the present invention, just then compare and to draw optimal solution, thereby the limitation problem is resolved.Whole progress as shown in Figure 4.
Referring to Fig. 5, the anti-error optimization method flow chart of power distribution network comprises following processing step:
The first step: the topological structure of determining power distribution network;
Second step: judge whether network is correct network; Whether the network judgement: whether is active network, be Radial network if comprising;
The 3rd step: Adoption Network divides and around-Francely divides ring based on current network state;
The 4th step: divided good ring to carry out real coding to last link; Guarantee that each loop only has a switch to be in open mode, so each real number corresponding state that particular switch is opened only;
The 5th step: the real coding of all ring sections is compiled a chromosome according to sequencing;
The 6th step: the real coding process is complete;
The 7th step: call genetic algorithm; The scheme that is resolved obtains scheme, provides its network loss size, the parameters such as on-off times;
The 8th step: anti-error optimization finishes.