CN104037776B - The electric network reactive-load capacity collocation method of random inertial factor particle swarm optimization algorithm - Google Patents

Abstract

The present invention provides the electric network reactive-load capacity collocation method of a kind of random inertial factor particle swarm optimization algorithm, and the method comprises the following steps: the systematic parameter of I, in real time acquisition WAMS system, sets the boundary condition of particle；II, initialization population, determine the adaptive value of described particle；III, division iteration phase；IV, the speed updating described particle and position；V, judge whether iterations arrives global search stage maximum iteration time；VI, judge whether iterations arrives and primary solve stabilization sub stage maximum iteration time；VII, judge whether iterations arrives the iteration upper limit；VIII, fall generation to maximum times, export online reactive capability allocation plan.Compared with canonical algorithm and TSP question algorithm, the method of the present invention improves the precision of optimization, while ensureing convergence rate, in conjunction with the practical situation of idle work optimization, achieve the raising of early stage ability of searching optimum, afterwards its Local Search precision, finally give globally optimal solution.

Description

The electric network reactive-load capacity collocation method of random inertial factor particle swarm optimization algorithm

Technical field

The present invention relates to a kind of method that electric network intelligent scheduling supports system operation of power networks state estimation and early warning field, tool Body relates to the electric network reactive-load capacity collocation method of a kind of random inertial factor particle swarm optimization algorithm.

Background technology

Along with society, the economic and fast development of power industry, electrical network is developing progressively extensive, remote, extra-high voltage Alternating current-direct current interconnection and new forms of energy access the forms such as ratio increasing, add the uncertainty of operation of power networks.And receiving end electrical network is main Centered by load area of concentration, it is connected by periphery interconnection power supply of making a start with remote broad sense, and then realizes the confession of electric energy Need balance.Due to not mating of the energy and load center area distribution, and consider the factor restrictions such as environment, inside receiving-end system Power supply underbraced, a large amount of electric energy need to carry out long-distance transmissions from a distant place, and receiving-end system scale rapidly increases and its complexity Get over complexity.

Since the eighties in 20th century, the most multiple large-scale power systems occur many starting voltage to continue on the low side, voltage in succession Crash event, causes huge economic loss and social influence, makes voltage stabilization be increasingly becoming Jiao that International Power educational circles pays close attention to Point, has higher requirement to the receiving end Network Voltage Stability on-line monitoring under running environment.It is presently used for power system quiet State voltage stabilization and the numerical algorithm comparative maturity of Dynamic voltage stability emulation, cause simulation result and real system to misfit Reason is mainly the inaccurate of component models and parameter in system.It addition, analysis method based on mathematical modeling and emulation, by electricity The restriction of the factors such as pessimistic concurrency control, parameter and numerical computations, is difficult to adapt to voltage at aspects such as application scale, speed and reliabilities Stablize the requirement of online real-time assessment.

Configuration that AC network is idle is that the important real-time voltage improving systematic function controls management technique.Generally, Power transmission network reactive-load compensation means can be divided into two big classes, i.e. power plant reactive power output regulation and substation capacitors voltage to prop up Support.Combination of the two has appreciable impact, to it to the transmission of reactive power in power transmission network and the numerical value of network node voltage Optimization belong to multiple target combination optimization problem.

In reactive capability optimization allocation based on loss minimization, it is intended to simultaneously a series of given conditions it Under, the value of control variable is most optimally set.These control variable include the input of generator reactive, transformer voltage ratio, parallel connection The idle output etc. of electric capacity/anti-device.Many documents have carried out Modeling Research to it in recent years, and use evolution algorithmic to carry out it Solve, such as genetic algorithm, ant group algorithm and TS algorithm etc..

Particle swarm optimization algorithm is the one in intelligent algorithm.Particle cluster algorithm is excellent owing to modeling is simple, convergence is fast etc. Point, distributes optimization problem rationally at reactive power capacity and solves field and obtained sufficient development.

Owing to particle swarm optimization algorithm has fast convergence rate, easily realizes and need the advantages such as parameter is few, existing many Document is with regard to reactive capability optimization problem, it is proposed that the PSO algorithm of improvement.But, when PSO is applied to higher-dimension challenge, hold Premature Convergence easily occur and causes the problems such as local optimum, result in this algorithm it cannot be guaranteed that converge to global optimum.There is this The main cause of the situation of kind is that premature convergence speed is fast, does not obtain operative constraint to the later stage and makes algorithm depart from minimal point.

Summary of the invention

In order to overcome the defect of above-mentioned prior art, the invention provides a kind of random inertial factor particle swarm optimization algorithm Electric network reactive-load capacity collocation method.

In order to realize foregoing invention purpose, the present invention adopts the following technical scheme that:

A kind of electric network reactive-load capacity collocation method of random inertial factor particle swarm optimization algorithm, it thes improvement is that: Said method comprising the steps of:

The systematic parameter of I, in real time acquisition WAMS system, sets the boundary condition of particle；

II, initialization population, set up the idle work optimization model of electrical network, determine the adaptive value of described particle；

III, division iteration phase；

IV, the speed updating described particle and position；

V, judge whether iterations arrives global search stage maximum iteration time；

VI, judge whether iterations arrives and primary solve stabilization sub stage maximum iteration time；

VII, judge whether iterations arrives the iteration upper limit；

VIII, fall generation to maximum times, export online reactive capability allocation plan.

Further, the described systematic parameter of described step I includes the PMU measured value of described network system and described EMS Data；

Described PMU measured value includes bus current, busbar voltage, active power and reactive power；

Described EMS data include busbar voltage, busbar voltage stream, active power and reactive power；

The border of described particle is estimated result and the current scheduling operation rule of state situation based on described network system Fixed real-time each point voltage upper lower limit value, the solution space scope of the corresponding algorithm of setting；

Further, described step II comprises the following steps:

S201, the nodal information obtaining distribution network system and branch road information, arrange the number of control variable and respectively control to become The span of amount and the population size of initial population；

Described initial population is initialized and initial parameter is set, it is thus achieved that primary group；

Initialize the described initial population of acquisition and refer in particle span random to the particle in described initial population Selecting initial velocity and the initial position of particle, described initial parameter includes maximum iteration time and adapts to threshold value；

S202, to choose power transmission network active power loss be object function, as following formula determines the mathematical model of idle work optimization:

In formula, and k=(i, j), i ∈ N_B, N_BFor all bus nodes set, j ∈ N_i, N_iFor be associated with bus nodes i Node set；For power transmission network active power loss；g_kAdmittance for branch road k；v_i,v_jRespectively bus nodes i and j Voltage magnitude；θ_ijDifferential seat angle for load bus i and j；

S202, determine equality constraint such as following formula:

In formula, P_giThe generator active power of node i injects；P_diLoad active power for node i；g_ij、B_ijIt is respectively Conductance between node i, j and susceptance；Q_giIt it is the generator reactive power injection of node i；Q_diReactive load merit for node i Rate；v_i,v_jBus nodes i and the voltage magnitude of j respectively；θ_ijDifferential seat angle for load bus i and j.

Further, described step III comprises the following steps:

S301, primary iteration number of times are 1, are divided into global search stage, primary solution to stablize rank iteration according to iterations Section and high precision solution stabilization sub stage；As inertial factor w is adjusted by following formula (1):

In formula, w_maxIt is positioned at primary iteration, w_minBeing positioned at the end in iteration period, iter is current iteration number, iter_maxFor Maximum iteration time,

S302, as following formula (2), (3)) (4) determine inertial factor w and accelerated factor c of different iteration phase respectively₁、c₂:

In formula, k_MFor global search stage maximum iteration time, k_NFor primary solution stabilization sub stage maximum iteration time, k_MAXFor High precision solution stabilization sub stage maximum iteration time.

Further, in described step IV, the speed v of particle as described in following formula (5), (6) update respectively_i ^k+1With position x_i ^{k +1}:

In formula, v_i ^k+1For the i-th particle velocity when+1 generation of kth；W is the inertial factor of particle；v_i ^kIt is i-th Individual particle kth for time velocity；c_1,c₂For accelerator coefficient；r₁,r₂In the range of the number randomly generated between [0,1] Word；For optimum position based on population iteration historical i-th particle, g_bestiFor G population particle global optimum position Put；x_i ^k+1Position for i-th particle when+1 generation of kth；x_i ^kFor kth for time the position of i-th particle；χ for penalize because of Son；

Judge whether described particle control variable exceedes the boundary condition of described particle, if exceeding, value again.

Further, in described step V, if judge iterations not less than global search stage maximum iteration time, then Revise inertial factor w and accelerated factor c₁、c₂, iteration kth+1 word iterative value；

In described step VI, if judging, iterations not less than primary solution stabilization sub stage maximum iteration time, is then revised used Sex factor w and accelerated factor c₁、c₂, iteration kth+1 word iterative value；

In described step VII, if judging, iterations not less than the iteration upper limit, then revises inertial factor w and accelerated factor c₁、c₂, iteration kth+1 word iterative value.

Further, described step VIII includes:

If the history individuality extreme value that the current particle state of particle is better than in iterative process, then with this state more new historical Body extreme valueIf neighborhood particle having the neighborhood history extreme value that the current state particle of particle is better than in iterative process, then with this State updates neighborhood history optimum g_best；

Provide voltage level of power grid according to the node voltage that described WAMS measures, determine reactive capability allocation plan.

Compared with prior art, the beneficial effects of the present invention is:

1, the method for the present invention makes full use of the basic electric network model of EMS system, parameter and runs section information, in conjunction with The high accuracy high-density acquisition data of WAMS system, it is achieved that online quiescent voltage enabling capabilities assessment.

2, the method for the present invention is to be some subspaces by search space partition, carries out POS algorithm optimizing in each subspace； On the basis of analyzing the inertial factor mechanism of action, each sub regions devises one according to population diversity and evolution The inertial factor computational methods of algebraically Automatic adjusument, by transformation search step-length, improve the local search ability of algorithm.

3, compared with canonical algorithm and TSP question algorithm, the method for the present invention improves the precision of optimization, is ensureing While convergence rate, in conjunction with the practical situation of idle work optimization, it is achieved that early stage ability of searching optimum, afterwards its Local Search precision Raising, finally give globally optimal solution.

4, the algorithm of present invention systematic parameter based on WAMS system is capable of when system operation conditions changes Dynamic self-adapting.

Accompanying drawing explanation

The electric network reactive-load capacity collocation method flow process of the random inertial factor particle swarm optimization algorithm that Fig. 1 provides for the present invention Figure.

Detailed description of the invention

The invention will be further described below in conjunction with the accompanying drawings.

With the active loss (i.e. network loss) of power system to be optimized as fitness function, to find system minimum For the purpose of network loss；The trend that solves of random inertial factor particle swarm optimization algorithm, each branch road network loss superposition is used to ask for total system Network loss.The globally optimal solution tried to achieve is system minimum network loss, and corresponding optimal particle is generator voltage, load tap changer The control variable parameters such as gear, Shunt Capacitor Unit switching group number.

As it is shown in figure 1, the electric network reactive-load capacity of random inertial factor particle swarm optimization algorithm that Fig. 1 provides for the present invention is joined Put method flow diagram；The method comprises the following steps:

The systematic parameter of step one, in real time acquisition WAMS system, sets the boundary condition of particle；

Step 2, initialization population, determine the adaptive value of described particle；

Step 3, division iteration phase；

Step 4, the speed updating described particle and position；

Step 5, judge whether iterations arrives global search stage maximum iteration time；

Step 6, judge whether iterations arrives and primary solve stabilization sub stage maximum iteration time；

Step 7, judge whether iterations arrives the iteration upper limit；

Step 8, fall generation to maximum times, export online reactive capability allocation plan.

Step one, obtains the systematic parameter of WAMS system in real time, sets the boundary condition of particle.

Described systematic parameter includes the PMU measured value of described network system and described EMS data；

Described PMU measured value includes bus current, busbar voltage, active power and reactive power；

Described EMS data include busbar voltage, bus current, active power and reactive power；

The border of described particle is estimated result and the current scheduling operation rule of state situation based on described network system Fixed real-time each point voltage upper lower limit value, the solution space scope of the corresponding algorithm of setting；

In embodiment one, obtain systematic parameter in real time from WAMS system and carry out capacity configuration, including the 500kV of goal systems System PMU measured value (such as 500kV bus current, busbar voltage, meritorious and reactive power) and the EMS data of 220kV system The state estimation result of (such as 220kV busbar voltage, bus current, meritorious and reactive power) formation current system, and based on The state estimation result of network system formed above and real-time each point voltage upper lower limit value (this electricity of current scheduling operating provisions Press lower limit manually to set, such as, drafted the voltage of the change of countershaft at any time of some node at present by higher level traffic department Bound curve), set the solution space scope in corresponding algorithm, i.e. the boundary condition of particle.

Described state estimation result refers in the case of given SCADA data and PMU data, by algorithm for estimating, calculates The voltage magnitude of a certain moment each node of electrical network that goes out, phase angle, and then draw active power and the reactive power on each road.

Owing to the quantity of particle masses' particle is inversely proportional to the calculating speed asked for, it is directly proportional to computational accuracy, therefore depends on The situation degree of risk current according to operation of power networks sets population quantity.

Population boundary condition is specified by electrical network actual motion, as 1 day 24 hours the most in the same time, a season for Electrical voltage point has the voltage float area to certain point of special magnitude of voltage bandwidth curve to have regulation.Algorithm final search arrives Optimal solution must be the solution within corresponding moment bandwidth curve.

Population quantity is as the change of power grid time and dynamically sets, and such as, 9 o'clock of the morning is the morning peak phase Between, network load fluctuation is relatively big, now pursues speed more than pursuing precision, and therefore population quantity set is less, in order to quickly Seek out capacity configuration solution.And at 12 in evening to 5:00 AM, system loading is the most relatively low, now load variations is little, precision Requiring bigger, therefore set population population quantity will be of a relatively high.It is currently set value can use specific to quantity The method that experience sets.

Step 2, initialization population, determine the adaptive value of described particle.Comprise the following steps:

S201, the nodal information obtaining distribution network system and branch road information, arrange the number of control variable and respectively control to become The span of amount and the population size of initial population；

Described initial population is initialized and initial parameter is set, it is thus achieved that primary group；

Described initialization refers to randomly choose the particle in described initial population at the beginning of particle in particle span Beginning speed and initial position, described initial parameter includes maximum iteration time and adapts to threshold value；

S202, determine the object function of idle work optimization such as following formula:

In formula, and k=(i, j), i ∈ N_B, N_BFor all bus nodes set, j ∈ N_i, N_iFor be associated with bus nodes i Node set；For power transmission network active power loss；g_kAdmittance for branch road k；v_i,v_jRespectively bus nodes i and j Voltage magnitude；θ_ijDifferential seat angle for load bus i and j；

Described power transmission network active power loss obtains systematic parameter in real time according to WAMS system and determines, joins in the present embodiment Number include goal systems 500kV system PMU measured values (such as 500kV busbar voltage, electric current, meritorious and reactive power) and The EMS data (such as 220kV busbar voltage, bus current, meritorious and reactive power) of 220kV system, according to above-mentioned measured value Determine active power loss.

S202, determine equality constraint such as following formula:

Active power balance retrains

I.e. reactive power equilibrium constraint

In formula, P_giThe generator active power of node i injects；P_diLoad active power for node i；g_ij、B_ijIt is respectively Conductance between node i, j and susceptance；Q_giIt it is the generator reactive power injection of node i；Q_diReactive load merit for node i Rate；v_i,v_jBus nodes i and the voltage magnitude of j respectively；θ_ijDifferential seat angle for load bus i and j.

Step 3, division iteration phase.Comprise the following steps:

Primary iteration number of times is 1, is divided into global search stage, primary to solve stabilization sub stage and high iteration according to iterations The precision solution stabilization sub stage；As inertial factor w is adjusted by following formula (1):

In formula, w_maxIt is positioned at primary iteration, w_minBeing positioned at the end in iteration period, iter is current iteration number, iter_maxFor Maximum iteration time,

As following formula (2), (3)) (4) determine inertial factor w and accelerated factor c of different iteration phase respectively₁、c₂:

In formula, k_MFor global search stage maximum iteration time, k_NFor primary solution stabilization sub stage maximum iteration time, k_MAXFor High precision solution stabilization sub stage maximum iteration time.

Step 4, updates speed and the position of described particle.

The speed and the position that update described particle comprise the following steps:

Following formula (5), (6) respectively to as described in the speed v of particle_i ^k+1With position x_i ^k+1Do and update as follows:

In formula, v_i ^k+1For the i-th particle velocity when+1 generation of kth；W is the inertial factor of particle；v_i ^kIt is i-th Individual particle kth for time velocity；c_1,c₂For normal number, scope is [0,2.5]；r₁,r₂Between [0,1] The numeral randomly generated；For optimum position based on population iteration historical i-th particle；g_bestiFor G population grain Sub-global optimum position；x_i ^k+1Position for i-th particle when+1 generation of kth；x_i ^kFor kth for time i-th particle Position；χ is penalty factor, is used for guaranteeing convergence；

Judging whether described particle control variable exceedes the described particle boundary condition of a kind of setting of step, if exceeding, weighing New value.

In step 5, if judging, iterations not less than global search stage maximum iteration time, then revises inertial factor w With accelerated factor c₁、c₂, iteration kth+1 word iterative value；

In step 6, if judge iterations not less than primary solution stabilization sub stage maximum iteration time, then revise inertia because of Sub-w and accelerated factor c₁、c₂, iteration kth+1 word iterative value；

In step 7, if judging, iterations not less than the iteration upper limit, then revises inertial factor w and accelerated factor c₁、c₂, Iteration kth+1 word iterative value.

Step 8, fall generation to maximum times, export online reactive capability allocation plan.

If the history individuality extreme value that the current particle state of particle is better than in iterative process, then with this state more new historical Body extreme valueIf neighborhood particle having the neighborhood history extreme value that the current state particle of particle is better than in iterative process, then with this State updates neighborhood history optimum g_best；

Provide voltage level of power grid according to the node voltage that described WAMS measures, determine reactive capability allocation plan.

Finally should be noted that: above example is only in order to illustrate that technical scheme is not intended to limit, to the greatest extent The present invention has been described in detail by pipe with reference to above-described embodiment, and those of ordinary skill in the field are it is understood that still The detailed description of the invention of the present invention can be modified or equivalent, and any without departing from spirit and scope of the invention Amendment or equivalent, it all should be contained in the middle of scope of the presently claimed invention.

Claims (6)

1. the electric network reactive-load capacity collocation method of a random inertial factor particle swarm optimization algorithm, it is characterised in that: described side Method comprises the following steps:

The systematic parameter of I, in real time acquisition WAMS system, sets the boundary condition of particle；

II, initialization population, set up the idle work optimization model of electrical network, determine the adaptive value of described particle；

III, division iteration phase；

IV, the speed updating described particle and position；

V, judge whether iterations arrives global search stage maximum iteration time；

VI, judge whether iterations arrives and primary solve stabilization sub stage maximum iteration time；

VII, judge whether iterations arrives high precision solution stabilization sub stage maximum iteration time；

VIII, iterate to each iteration phase maximum iteration time, export online reactive capability allocation plan；

In described step V, if judging, iterations not less than global search stage maximum iteration time, then revises inertial factor w With accelerated factor c₁、c₂,+1 iterative value of iteration kth；

In described step VI, if judge iterations not less than primary solution stabilization sub stage maximum iteration time, then revise inertia because of Sub-w and accelerated factor c₁、c₂,+1 iterative value of iteration kth；

In described step VII, if judging, iterations not less than high precision solution stabilization sub stage maximum iteration time, then revises inertia Factor w and accelerated factor c₁、c₂,+1 iterative value of iteration kth.

2. the method for claim 1, it is characterised in that: the described systematic parameter of described step I includes network system PMU measured value and EMS data；

Described PMU measured value includes bus current, busbar voltage, active power and reactive power；

Described EMS data include busbar voltage, bus current, active power and reactive power；

The boundary condition of described particle is state estimation result based on described network system and the reality of current scheduling operating provisions Time each point voltage upper lower limit value, the solution space scope of the corresponding algorithm of setting；

3. the method for claim 1, it is characterised in that: described step II comprises the following steps:

S201, obtain the nodal information of distribution network system and branch road information, the number of control variable and each control variable are set Span and the population size of initial population；

Described initial population is initialized and initial parameter is set, it is thus achieved that primary group；

Initialize the described primary group of acquisition to refer in particle span, the particle in described initial population be selected at random Selecting initial velocity and the initial position of particle, described initial parameter includes maximum iteration time and adaptive value；

S202, to choose power transmission network active power loss be object function, as following formula determines the mathematical model of idle work optimization:

$\min \underset{k\∈N_E}{\Σ}{P}_{kloss}=\underset{k\∈N_E}{\Σ}{g}_k({v_i}^2+{v_j}^2-2v_i{v}_j{\mathrm{cos\θ}}_{ij})$

In formula, and k=(i, j), i ∈ N_B, N_BFor all bus nodes set, j ∈ N_i, N_iFor the node being associated with bus nodes i Set；For power transmission network active power loss；g_kAdmittance for branch road k；v_i,v_jIt is respectively bus nodes i and the electricity of j Pressure amplitude value；θ_ijDifferential seat angle for load bus i and j；

S203, determine equality constraint such as following formula:

With

In formula, P_giThe generator active power of node i injects；P_diLoad active power for node i；g_ij、B_ijIt is respectively node Conductance between i, j and susceptance；Q_giIt it is the generator reactive power injection of node i；Q_diReactive load power for node i；v_i, v_jBus nodes i and the voltage magnitude of j respectively；θ_ijDifferential seat angle for load bus i and j.

4. the method for claim 1, it is characterised in that: described step III comprises the following steps:

S301, primary iteration number of times are 1, according to iterations iteration is divided into the global search stage, primary solve the stabilization sub stage and The high precision solution stabilization sub stage；As inertial factor w is adjusted by following formula (1):

$w=w_{max}-\frac{w_{max}-w_{\min }}{{\mathrm{iter}}_{max}}\×iter---\left(1\right)$

In formula, w_maxIt is positioned at primary iteration, w_minBeing positioned at the end in iteration period, iter is current iteration number, iter_maxFor maximum Iterations,

S302, determine inertial factor w and accelerated factor c of different iteration phase respectively such as following formula (2), (3), (4)₁、c₂:

$c_1=\left\{\begin{array}{c}2.1,1\&le;iter<k_M\\ 1.05,k_M\&le;iter<k_N\\ 0.525,k_N\&le;iter<k_{max}\end{array}\right.---\left(2\right)$

$c_2=\left\{\begin{array}{c}2.0,1\&le;iter<k_M\\ 1,k_M\&le;iter<k_N\\ 0.5,k_N\&le;iter<k_{max}\end{array}\right.---\left(3\right)$

$w=\left\{\begin{array}{c}2.1,1\&le;iter<k_M\\ 1.05,k_M\&le;iter<k_N\\ 0.525,k_N\&le;iter<K_{MAX}\end{array}\right.---\left(4\right)$

In formula, k_MFor global search stage maximum iteration time, k_NFor primary solution stabilization sub stage maximum iteration time, k_MAXFor high-precision Degree solves stabilization sub stage maximum iteration time.

5. the method for claim 1, it is characterised in that: in described step IV, as described in following formula (5), (6) update respectively The speed v of particle_i ^k+1With position x_i ^k+1:

$v_i^{k+1}=w\&times;v_i^k+c_1\&times;r_1\&times;(P_{besti}-x_i^k)+c_2\&times;r_2\&times;(g_{besti}-x_i^k)---\left(5\right)$

$x_i^{k+1}=x_i^k+\mathrm{\&chi;}\&times;v_i^{k+1}---\left(6\right)$

In formula, v_i ^k+1For the i-th particle velocity when+1 generation of kth；W is the inertial factor of particle；v_i ^kFor i-th grain Son kth for time velocity；c₁,c₂For accelerated factor；r₁,r₂In the range of the numeral randomly generated between [0,1]；For optimum position based on population iteration historical i-th particle, g_bestiFor G population particle global optimum position； x_i ^k+1Position for i-th particle when+1 generation of kth；x_i ^kFor kth for time the position of i-th particle；X for punishment because of Son；

Judge whether particle control variable exceedes the boundary condition of described particle, if exceeding, value again.

6. method as claimed in claim 5, it is characterised in that: described step VIII includes:

If the history individuality extreme value that the current particle state of particle is better than in iterative process, then with this state more new historical individuality pole Value p_best；If neighborhood particle having the neighborhood history extreme value that the current particle state of particle is better than in iterative process, then with this shape State updates neighborhood history optimal value g_best；

Provide voltage level of power grid according to the node voltage that described WAMS measures, determine reactive capability allocation plan.