CN105913158A - DP-MPSO algorithm based weekly repair plan compiling and optimizing method for power dispatching - Google Patents

Abstract

The invention provides a DP-MPSO algorithm based weekly repair plan compiling and optimizing method for power dispatching, which is characterized by and comprises the following steps: simplifying a power grid equipment repair model into a power charging and loading two-end model; performing a constraint processing to the power grid equipment repair model; breaking down and transforming the problem of repair plan compiling into the problem of an optimized combining of single day superposed load and the installed capacity balance across the power grid; formulating a peak value optimizing and sequencing strategy; converting into a 0/1 knapsack problem and through the DP-MPSO algorithm, obtaining the most optimized solution to the peak value; dating back all optimized results of the peak value; inspecting a cross-work period repair ticket optimized set and determining if the acceptance or rejection results of one single cross-work period repair ticket are consistent or not in different work periods. With the method provided by the invention, the goals of repair compiling are concentrated and refined. Being intelligent and high efficient, the plan compiling is added by a DP-MPSO algorithm in the optimization process with the consequence of an improved calculating efficiency.

Description

A kind of dispatching of power netwoks weekly repair planning optimization side based on DP-MPSO algorithm Method

Technical field

The present invention relates to dispatching automation of electric power systems technical field, particularly relate to a kind of based on The dispatching of power netwoks weekly repair planning optimization method of DP-MPSO algorithm.

Background technology

Along with economic fast development, society's power consumption the most quickly increases, electrical network scale Rapid Expansion, Electric network composition is the most complicated；The power equipment quantity such as bus, transformator, disconnecting link quickly increase, then Bring is grid equipment safety, the problem that economical operation ensures.

After strong intelligent grid plan comes into effect, the maintenance of electrical network is carried by the development of many nascent technology Go out new requirement, including electricity market, Distributed Power Generation and repair based on condition of component etc., wherein The development of electricity market promotes electric power enterprise to pay attention to the economy of maintenance.Along with electricity market reform not Breaking deeply, domestic electricity power enterprise and power supply enterprise have been separated from, and this just requires that power supply enterprise is to user While firm energy is provided, it is necessary to make great efforts to improve the economy of operational management, to improve the competing of enterprise Strive power.

Electric power apparatus examination, as the important process of power supply enterprise's day-to-day operation, is raising equipment healthy water Flat, the properly functioning important component part being to ensure that power network safety operation of guarantee power equipment, Among these, Maintenance Schedule Optimization is one of key component of maintenance decision technology, and its task is exactly The time is implemented in the maintenance of Optimum equipment, is ensureing maintenance task and is meeting the various constraint of system On the basis of, reduce the manpower and materials cost that inspection and repair shop is paid as best one can, reaching the operation of raising system can By targets such as property and the states improving equipment self.Therefore, can reasonable efficient pin in conjunction with maintenance constraint The repair schedule submitting all departments collects the key that arrangement is repair schedule establishment.

Traditional maintenance model has the disadvantage that

(1) system multiobject can be gone to weigh operation of power networks index and grid equipment is overhauled the shadow brought Ring, the factor such as minimum short of electricity amount of system when system reserve capacity after stopping transport such as unit maintenance, maintenance, Make maintenance model the most loaded down with trivial details, not easy care.

(2) subproblem during the out-of-limit DECOMPOSED OPTIMIZATION of superposition load is not separate, can be limited by Overstate the impact of duration maintenance loss electricity.

(3) PSO algorithm is a kind of random, parallel optimized algorithm, but it is for there being multiple office The function of portion's extreme point, is easily trapped in Local Extremum, can not get correct result.

Therefore, it is badly in need of the optimization method of a kind of dispatching of power netwoks weekly repair planning, gets rid of operation of power networks Electrical network primary equipment is overhauled the impact brought by safety other factors outer so that the decision-making of establishment more collects Middle lean, ensures grid equipment safety, provides for economical operation and supports.

Summary of the invention

It is an object of the invention to provide a kind of dispatching of power netwoks weekly repair plan based on DP-MPSO algorithm Establishment optimization method, by the whole network installed capacity data of scheduling repair schedule system acquisition, establishment week Load prediction data in phase and the maintenance capacity information of application repair ticket, at electric power netting safe running In principle, by build equivalence maintenance operation of power networks model, make maintenance establishment target tightening lean, Intelligent, automatization, improves operation efficiency simultaneously.

For achieving the above object, present invention provide the technical scheme that

A kind of dispatching of power netwoks weekly repair planning optimization method based on DP-MPSO algorithm, including Following steps:

Grid equipment is overhauled the two end models that model simplification is generating and load by step one, with equivalence Unit substitutes all primary equipments in electrical network in addition to unit；

Step 2 according to mutual exclusion maintenance constraint, simultaneously overhaul constraint, force forbid maintenance retrain and Clearly overhaul constraint, grid equipment maintenance model is carried out constraint process；

Step 3, by establishment cycle per diem discretization, sets up final optimized maintenance model, maintenance is counted Draw establishment PROBLEM DECOMPOSITION to be converted into odd-numbered day superposition load and ask with the Combinatorial Optimization that the whole network installed capacity balances Topic；

Step 4 is formulated peak value and is optimized order policies, optimizes according to peak body linkage situation order of packets；

Out-of-limit peak value is optimized by step 5, is put down with the whole network installed capacity by odd-numbered day superposition load The combinatorial optimization problem of weighing apparatus is converted into 0-1 knapsack problem, obtains peak value by DP-MPSO algorithm excellent The optimal solution changed；

Step 6 recalls all peak value optimum results, checks across duration ticket optimization collection, it is determined that same The result being rejected or accepted in the different durations across duration repair ticket is the most inconsistent, it is determined that result is one Cause then to complete optimization method；Result of determination is inconsistent, returns step 5, updates tendency probability factor.

Further, in step one, model is carried out pretreatment, read in establishment cycle, load pre- Survey and the whole network installed capacity, the establishment cycle is per diem started discretization from 0, refine the load of every day Prediction, obtains all repair ticket plan start and end times in the establishment cycle, if plan starts Between less than 0, delete this repair ticket, cumulative residue maintenance capacity in the load prediction of residue maintenance day, Guarantee that the planned start time remaining repair ticket is more than or equal to 0.

Further, in step 2, mutual exclusion is retrained, set up according to the repair ticket of mutual exclusion and split Set, eliminates mutex relation；For overhauling constraint simultaneously, per diem maintenance capacity carries out repair ticket merging, Formed planned start time, the end time by new repair ticket the earliest and the latest in merging ticket, and delete Except original repair ticket；For forcing to forbid maintenance constraint, delete the repair ticket in this period；For Clearly retrain, directly accept this repair ticket.

Further, in step 4, in finding the establishment cycle, occur that superposition load prediction exceedes the whole network The point of installed capacity, whether existing across duration repair ticket of checking that this day relate to relates to other out-of-limit peak days Phenomenon, if it has not, then individually this day is carried out the optimization in step 5；If any, will be related to Peak be one group of optimization carried out in step 5, optimization order is according to the number of the repair ticket related to peak value day Amount size arranges from small to large and carries out the peak value that odd-numbered day optimization is out-of-limit one by one.

Further, in step 5, the fitness of peak value optimized algorithm is

${\mathrm{fitness}}_j=\left\{\begin{array}{cc}f\left(x_j\right),& (\underset{j=1}{\overset{n}{\Σ}}{C}_j{x}_j\≤N)\\ f\left(x_j\right)-H\×K,& (\underset{j=1}{\overset{n}{\Σ}}{C}_j{x}_j>N)\end{array}\right.$

Wherein, fitness_jBeing the fitness of jth particle, K is that this individuality is unsatisfactory for the individual of constraint Number, H is that a very big positive number (is typically set to H=10⁹), C_jFor the capacity of repair ticket, x_jFor 0-1 Decision variable, 0 represents that refusal repair ticket, 1 expression accept repair ticket, and N is knapsack capacity.

Further, it is determined that formula is

${\mathrm{\ω}}_i=\left\{\begin{array}{cc}1+\mathrm{\μ}\tanh (\frac{N_a}{S}-0.5),& x_j=1\\ 1+\mathrm{\μ}\tanh (\frac{N_r}{S}-0.5),& x_j=0\end{array}\right.$

Wherein, μ is convergence control coefrficient, is typically set to 0.4-0.6, N_aFor this linkage groups occurs knot The most inconsistent repair ticket obtains the number of days accepting result, N_rFor being rejected number of days, s is that this repair ticket is overstated The number of days of duration.

Use technique scheme, there is advantages that

The first, maintenance establishment target tightening lean, by overhauling the Equivalent Simplification of model, gets rid of electricity Electrical network primary equipment is overhauled the impact brought by network operation safety other factors outer so that the decision-making of establishment More concentrate lean, be prevented effectively from " cross and repair " and " owing to repair ", may make up the chain of central planning establishment Module link.

The second, intelligent automaticization improve, work out plan with traditional conference type compared with, decrease people The expense of power resource, changes into the combinatorial optimization problem of 0/1 knapsack by being intended to establishment problem, real The intelligent optimum combination of existing odd-numbered day repair ticket, and when the plan worked out occurs across the linkage of duration repair ticket out-of-limit During peak value, according to Backtracking Strategy, problem condition can be judged, automatic adjusting and optimizing algorithm, it is achieved Work out the intelligence global optimization automatically of plan.

3rd, operation efficiency is improved, owing to introducing DP-MPSO during working out plan in optimization Algorithm so that scheduling algorithm realization is easier to, global optimization ability is higher, the place to locally optimal solution Reason is more preferable, algorithmic statement process is more rapid, and overcomes " the dimension calamity " that conventional algorithm easily occurs Problem.

Accompanying drawing explanation

Fig. 1 is the flow chart of DP-MPSO algorithm；

Fig. 2 is the flow chart of optimization method of the present invention.

Detailed description of the invention

In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with the accompanying drawings And embodiment, the present invention is further elaborated.Should be appreciated that structure described herein Figure and specific embodiment, only in order to explain the present invention, are not intended to limit the present invention.

Fig. 2 is the flow chart of optimization method of the present invention, as in figure 2 it is shown, one is based on DP-MPSO The dispatching of power netwoks weekly repair planning optimization method of algorithm, comprises the following steps:

Grid equipment is overhauled the two end models that model simplification is generating and load by step one, with equivalence Unit substitutes all primary equipments in electrical network in addition to unit；

Step 2 according to mutual exclusion maintenance constraint, simultaneously overhaul constraint, force forbid maintenance retrain and Clearly overhaul constraint, grid equipment maintenance model is carried out constraint process；

Step 3, by establishment cycle per diem discretization, sets up final optimized maintenance model, maintenance is counted Draw establishment PROBLEM DECOMPOSITION to be converted into odd-numbered day superposition load and ask with the Combinatorial Optimization that the whole network installed capacity balances Topic；

Step 4 is formulated peak value and is optimized order policies, optimizes according to peak body linkage situation order of packets；

Out-of-limit peak value is optimized by step 5, is put down with the whole network installed capacity by odd-numbered day superposition load The combinatorial optimization problem of weighing apparatus is converted into 0-1 knapsack problem, obtains peak value by DP-MPSO algorithm excellent The optimal solution changed；

Step 6 recalls all peak value optimum results, checks across duration ticket optimization collection, it is determined that same The result being rejected or accepted in the different durations across duration repair ticket is the most inconsistent, it is determined that result is one Cause then to complete optimization method；Result of determination is inconsistent, returns step 5, updates tendency probability factor.

Embodiment 1

In the step one of the inventive method, equivalent unit builds, and obtains whole network equipment from system normal System during operation can load with lotus L_f, and when disconnecting in repair ticket the equipment needing maintenance, system can Load with lotus L_d, the installed capacity C of the equivalent unit of definition_e, specific as follows:

C_e=L_f-L_d

Model preprocessing, reads in establishment cycle T, the load prediction L in T_pAnd the whole network installed capacity W, per diem starts discretization from 0 by T, refines the load prediction L of every day_pi, all in obtaining T Repair ticket M_jPlan start and end time t_js, t_jeIf, t_js< 0, delete this repair ticket, cumulative surplus Remaining maintenance capacity is in the load prediction L of residue maintenance day_pOn, it is ensured that the t of residue repair ticket_js≥0。

In step 2, maintenance constraint processes, and retrains for mutual exclusion, need to build according to the repair ticket of mutual exclusion Vertical fractionation is gathered, and eliminates mutex relation, such as M1 Yu M2 mutual exclusion, then sets up (M1, M2)；(M1, M2), (M1, M2) three kinds of establishment set；For overhauling constraint simultaneously, per diem maintenance capacity is needed to enter Row repair ticket merges, and forms t_js, t_jeBy new repair ticket the earliest and the latest in merging ticket, and delete Repair ticket originally；For forcing to forbid maintenance constraint, only the repair ticket in this period need to be deleted； For clearly retraining, this repair ticket need to be directly accepted, this repair ticket load can be carried out by its duration pre- Survey superposition, and delete this repair ticket.

In step 3, build maintenance capacity peak curve according to maintenance constraint, obtain in repair ticket Maintenance capacity C_j, per diem it is added to as daily load prediction L within the duration that it is subordinate to_piOn, form T Interior superposition prediction load curve.

In step 4,5) formulate peak value optimization order, occur in finding the establishment cycle that superposition load is pre- Surveying and exceed the point of the whole network installed capacity, whether existing across duration repair ticket of checking that this day relate to relates to it The phenomenon of his out-of-limit peak day, if it has not, then individually carry out " peak clipping " to optimize (the most out-of-limit peak value to this day Optimize)；If any, by related to peak and be one group and be optimized, optimization order related to according to peak value day And the population size of repair ticket arrange from small to large and carry out the odd-numbered day " peak clipping " one by one and optimize.

In steps of 5, Optimized model converts, and " peak clipping " is optimized and is converted into 0/1 knapsack problem, fixed Justice knapsack capacity N is the whole network installed capacity W and as daily load prediction L_piDifference, specific as follows:

N=W-L_pi

The weight of article j is capacity C j of repair ticket Mj, and Item Value Vj is this repair ticket capacity With the product of its duration length, specific as follows:

V_j=C_j×(t_js-t_je)

Under out-of-limit peak, the mathematical model of " peak clipping " optimization problem of n repair ticket is converted into:

$\max f(x_1,x_2\mathrm{...}{x}_n)=\underset{j=1}{\overset{n}{\&Sigma;}}{V}_j{x}_j$

$\begin{array}{cc}s.t.& \underset{j=1}{\overset{n}{\&Sigma;}}{C}_j{x}_j\&le;N\end{array}$

Wherein, x_jFor 0-1 decision variable, 0 represents that refusal repair ticket, 1 expression accept repair ticket.

Transformation standard PSO algorithmic function is interval, and the speed equation of motion of DP-MPSO algorithm is:

$\begin{array}{c}{v}_{jd}^{k+1}={\mathrm{\&omega;}}_j\&times;w\&times;v_{jd}^k+c_1\&times;r_1\&times;(p_{jd}-\mathrm{int}(x_{jd}^k\left)\right)+c_2\&times;r_2\\ \&times;(p_{gd}-\mathrm{int}(x_{jd}^k\left)\right)\end{array}$

Wherein, ω_jFor tendency probability factor, being initially 1, w is inertia weight, c₁、c₂For often accelerating Number, r₁、r₂For the random number in [0,1] interval change at random.The position motion side of DP-MPSO algorithm Cheng Wei:

$x_{jd}^{k+1}=\left\{\begin{array}{cc}1,& Sig\left(v_{jd}^{k+1}\right)>{\mathrm{\&rho;}}_{jd}^{k+1}\\ 0,& Sig\left(v_{jd}^{k+1}\right)\&le;{\mathrm{\&rho;}}_{jd}^{k+1}\end{array}\right.$

$Sig\left(v_{jd}^{k+1}\right)=\frac{1}{1+\exp (-v_{jd}^{k+1})}$

Sig function is fuzzy Sigmoid function,It is the random vector of Normal Distribution, now, The position of particle is only converted into (0,1) two kinds, speed withThresholding relevant, its value is the biggest, Particle position be 1 probability the biggest, otherwise the least.

Utilizing DP-MPSO algorithm to carry out " peak clipping " Combinatorial Optimization, the fitness of algorithm is:

${\mathrm{fitness}}_j=\left\{\begin{array}{cc}f\left(x_j\right),& (\underset{j=1}{\overset{n}{\&Sigma;}}{C}_j{x}_j\&le;N)\\ f\left(x_j\right)-H\&times;K,& (\underset{j=1}{\overset{n}{\&Sigma;}}{C}_j{x}_j>N)\end{array}\right.$

Wherein, fitness_jBeing the fitness of jth particle, K is that this individuality is unsatisfactory for the individual of constraint Number, H is that a very big positive number (is typically set to H=10⁹), the solution making this problem with this can be as early as possible Jump out local optimum.

In step 6, trace back and solve globally optimal solution, after having solved each often group linkage peak value, can Can occur that same refusal occurs and to accept result inconsistent across duration repair ticket within the different durations Situation, it is therefore desirable to by particle rapidity in tendency probability factor regulation DP-MPSO algorithm, make connection Dynamic peak value group carries out re-optimization, it is to avoid occur at sixes and sevens across duration repair ticket decision-making.Tendency is general Rate factor ω_jFor:

${\mathrm{\&omega;}}_j=\left\{\begin{array}{cc}1+\mathrm{\&mu;}\tanh (\frac{N_a}{S}-0.5),& x_j=1\\ 1+\mathrm{\&mu;}\tanh (\frac{N_r}{S}-0.5),& x_j=0\end{array}\right.$

Wherein, μ is convergence control coefrficient, is typically set to 0.4-0.6, N_aFor this linkage groups occurs knot The most inconsistent repair ticket obtains the number of days accepting result, N_rFor being rejected number of days, s is that this repair ticket is overstated The number of days of duration.Meet N_a+N_r=S, the equation completes the acceleration in trace-back process solved tendency Disturbance, it is achieved that global solution is to the evolution rationalized.

Embodiment described above only have expressed embodiments of the present invention, and it describes more concrete and detailed Carefully, but therefore can not be interpreted as the restriction to the scope of the claims of the present invention.It should be pointed out that, it is right For those of ordinary skill in the art, without departing from the inventive concept of the premise, it is also possible to do Going out some deformation and improvement, these broadly fall into protection scope of the present invention.Therefore, patent of the present invention Protection domain should be as the criterion with claims.

Claims (6)

1. a dispatching of power netwoks weekly repair planning optimization method based on DP-MPSO algorithm, its It is characterised by, comprises the following steps:

Grid equipment is overhauled the two end models that model simplification is generating and load by step one, with equivalence Unit substitutes all primary equipments in electrical network in addition to unit；

Step 2 according to mutual exclusion maintenance constraint, simultaneously overhaul constraint, force forbid maintenance retrain and Clearly overhaul constraint, grid equipment maintenance model is carried out constraint process；

Step 3, by establishment cycle per diem discretization, sets up final optimized maintenance model, maintenance is counted Draw establishment PROBLEM DECOMPOSITION to be converted into odd-numbered day superposition load and ask with the Combinatorial Optimization that the whole network installed capacity balances Topic；

Step 4 is formulated peak value and is optimized order policies, optimizes according to peak body linkage situation order of packets；

Out-of-limit peak value is optimized by step 5, is put down with the whole network installed capacity by odd-numbered day superposition load The combinatorial optimization problem of weighing apparatus is converted into 0-1 knapsack problem, obtains peak value by DP-MPSO algorithm excellent The optimal solution changed；

Step 6 recalls all peak value optimum results, checks across duration ticket optimization collection, it is determined that same The result being rejected or accepted in the different durations across duration repair ticket is the most inconsistent, it is determined that result is one Cause then to complete optimization method；Result of determination is inconsistent, returns step 5, updates tendency probability factor.

Dispatching of power netwoks weekly repair plan based on DP-MPSO algorithm the most according to claim 1 Establishment optimization method, it is characterised in that in step one, carries out pretreatment to model, reads in establishment Cycle, load prediction and the whole network installed capacity, per diem start discretization from 0 by the establishment cycle, carefully Change the load prediction of every day, obtain all repair ticket plan start and end times in the establishment cycle, as Really planned start time is less than 0, deletes this repair ticket, and cumulative residue maintenance capacity is in residue maintenance day Load prediction on, it is ensured that the planned start time of residue repair ticket is more than or equal to 0.

Dispatching of power netwoks weekly repair plan based on DP-MPSO algorithm the most according to claim 1 Establishment optimization method, it is characterised in that in step 2, retrains for mutual exclusion, according to the inspection of mutual exclusion Repair ticket and set up fractionation set, eliminate mutex relation；For overhauling constraint simultaneously, per diem maintenance capacity enters Row repair ticket merges, formed planned start time, the end time by merging ticket the earliest and the latest New repair ticket, and delete original repair ticket；For forcing to forbid maintenance constraint, delete in this period Repair ticket；For clearly retraining, directly accept this repair ticket.

Dispatching of power netwoks weekly repair plan based on DP-MPSO algorithm the most according to claim 1 Establishment optimization method, it is characterised in that in step 4, superposition load occurs in finding the establishment cycle Prediction exceedes the point of the whole network installed capacity, and whether existing across duration repair ticket of checking that this day relate to relates to The phenomenon of other out-of-limit peak place days, if it has not, then individually carry out the optimization in step 5 to this day； If any, by related to peak and be one group of optimization carried out in step 5, optimization order is according to peak value day The population size of the repair ticket related to arranges from small to large and carries out the peak value that odd-numbered day optimization is out-of-limit one by one.

Dispatching of power netwoks weekly repair plan based on DP-MPSO algorithm the most according to claim 1 Establishment optimization method, it is characterised in that in step 5, the fitness of peak value optimized algorithm is

${\mathrm{fitness}}_j=\left\{\begin{array}{cc}f\left(x_j\right),& \left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{C}_j{x}_j\&le;N\right)\\ f\left(x_j\right)-H\&times;K,& \left(\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{C}_j{x}_j>N\right)\end{array}\right.$

Wherein, fitness_jBeing the fitness of jth particle, K is that this individuality is unsatisfactory for the individual of constraint Number, H is that a very big positive number (is typically set to H=10⁹), C_jFor the capacity of repair ticket, x_jFor 0-1 Decision variable, 0 represents that refusal repair ticket, 1 expression accept repair ticket, and N is knapsack capacity.

Dispatching of power netwoks weekly repair plan based on DP-MPSO algorithm the most according to claim 1 Establishment optimization method, it is characterised in that in step 6, it is determined that formula is

${\mathrm{\&omega;}}_j=\left\{\begin{array}{cc}1+\mathrm{\&mu;}\tanh \left(\frac{N_a}{S}-0.5\right),& x_j=1\\ 1+\mathrm{\&mu;}\tanh \left(\frac{N_r}{S}-0.5\right),& x_j=0\end{array}\right.$

Wherein, μ is convergence control coefrficient, is typically set to 0.4-0.6, N_aFor this linkage groups occurs knot The most inconsistent repair ticket obtains the number of days accepting result, N_rFor being rejected number of days, S is that this repair ticket is overstated The number of days of duration.