CN104123683A - Electrical power system black-start scheme generation method based on dynamic programming - Google Patents

Abstract

The invention discloses an electrical power system black-start scheme generation method based on dynamic programming. The electrical power system black-start scheme generation method adopts the idea of gradually deducing black-start recovery operations, operation sequences with low recovery efficiency or quality are reduced gradually by means of state reduction technology, so that the process of decision making after the full enumeration of recovery conditions and cohesive relations thereof of all elements of the whole grid, which is required in technology such as Petri network technology, is avoided, and efficiency of algorithm is improved while requirements for computing storage capacity are lowered. Without any hypothesis of black-start recovery phase, each recovery operation can be chosen freely from all types, optional combination of restart of a generator unit, grid recovery and load recovery is allowed to carry out according to needs, problems of over-high voltage of line nodes in early stage of black start and inadaptability of the grid to the load recovery in later stage of black start are easy to solve, and the electrical power system black-start scheme generation method is naturally adaptable to different power grid after-fault scenes.

Description

Power system blackstart scheme generation method based on dynamic programming

Technical field

The invention belongs to power system restoration technical field, more specifically, relate to a kind of power system blackstart scheme generation method based on dynamic programming.

Background technology

Power system blackstart is often referred to whole system because of after the reason such as cascading failure stops transport, do not relying under the help of external network power supply, by thering is the startup of the unit of self-startup ability in system, drive the unit of non self starting, progressively expand the recovery scope of system power supply, finally realize the rejuvenation of whole electric system.So far a lot of accidents of having a power failure on a large scale that occur all over the world the sixties in 20th century show, in modern power systems, exist all the time local fault and deal with the hidden danger that causes accident expanded range improperly, therefore as preventive measure, the formulation of power system blackstart prediction scheme is necessary all the time.This work at present still mainly relies on and manually completes, often can only consider the wherein recovery order of several most important units, its efficiency and scheme integrality are all difficult to adapt to the application requirements in large-scale complex power grid, need the reliable black-start scheme automatic generating calculation of research badly.

But many factors and complexity that the black startup of actual electric network is considered, during program decisions, at least need to consider the problem of the startup of large electric power plant set auxiliary machinery, rack reconstruct and load restoration simultaneously, therefore up to now, most of existing algorithm research is paid close attention to the solution of some of them subproblem, and how the recovery decision-making of the whole black start-up course of comprehensive coordinate is still difficult point.Wherein, distributed decision making aspect, although the principal and subordinate who has provided based on many agencies, consideration graded dispatching is escorted Decision-making structures from one place to another, and is applied to Shandong Power, does not relate to the sub-power system restoration algorithm of refinement; Centralized decision aspect, has introduced Petri net as decision-making technic, but slightly large its Petri web frame of electrical network scale will become extreme complicated difficult to calculate.

In addition, the target first starting by unit steepest partly recovers rack, complete unit restarts, consider again the treatment principle of load restoration, although greatly improved scheme search efficiency, can in polynomial time, obtain feasible power system restoration scheme, but this algorithm is only confined to consider the recovery of short supply path between generator in the early stage, when therefore load is not enough on path, may cause line node overtension, recovered rack and also may be not suitable with the requirement of later stage load restoration its early stage, causes the prolongation of overall release time.

Summary of the invention

Above defect or Improvement requirement for prior art, the invention provides a kind of power system blackstart scheme generation method based on dynamic programming, the refinement restoration schedule scheme of taking all elements into account can be provided, improved efficiency of algorithm, reduced calculating the requirement of storage, be easy to overcome black startup line node overtension in early stage, later stage rack is not suitable with the problem of load restoration, has naturally adapted to scene after different electric network faults simultaneously.

For achieving the above object, the invention provides a kind of power system blackstart scheme generation method, it is characterized in that, comprise the steps:

(1) electric network information after read failure obtains initial electric network state s ₀, with variable g (s), represent that the shortest accumulative total of arrival electric network state s is consuming time, initialization g (s ₀)=0, progression k=0, original state collection S ₀={ s ₀, initialization terminal state set S _efor empty set, preferred plan T release time _bestfor infinity;

(2) to k level state set S _kin each state s _ki, determine corresponding recovery operation, these recovery operations are carried out to transient state verification, to judge the security of electric network state transient process; To each state s _ki, by each recovery operation of verification, can generate a new state, by these new states without repeatedly adding k+1 level state set S to _k+1in; Wherein, i=1,2 ..., n _k, n _kbe k level state set S _kthe sum of middle electric network state;

(3) to k+1 level state set S _k+1in each state carry out stable state verification, to judge the security of Power System Steady-state electric weight index, cannot from set, delete by the state by verification;

(4) to k+1 level state set S _k+1in each state s _{(k+1) j}, calculate the shortest accumulative total g (s consuming time that arrives this state _{(k+1) j})=min{g (s _ki)+r (s _ki, s _{(k+1) j}), and by the shortest accumulative total g (s consuming time _{(k+1) j}) corresponding restoration path is as the state of arrival s _{(k+1) j}optimal recovery path; Wherein, r (s _ki, s _{(k+1) j}) represent from state s _kito s _{(k+1) j}recovery operation consuming time, j=1,2 ..., n _k+1, n _k+1be k+1 level state set S _k+1the sum of middle electric network state;

(5) if k+1 level state set S _k+1in there is qualified dbjective state, all these qualified dbjective states are moved to SOT state of termination collection S _e, make T _minfor arriving the shortest accumulative total minimum value consuming time of these qualified dbjective states, if preferred plan T release time _best>T _min, make T _best=T _min, otherwise T _bestconstant, execution step (6); If k+1 level state set S _k+1in there is not qualified dbjective state, directly execution step (6);

(6) if k+1 level state set S _k+1for sky, or k+1 level state set S _k+1the shortest accumulative total T that has all been greater than consuming time of all states of middle residue _best, by SOT state of termination collection S _ein corresponding to T _bestoptimal recovery path as black-start scheme, stop to calculate; Otherwise execution step (7);

(7) to k+1 level state set S _k+1carry out state and cut down rearmounted k=k+1, return to step (2).

Preferably, in described step (2), to k level state set S _kin each state s _ki, before transient state verification, optional recovery operation at least comprises following five kinds:

(A1) make electric network state s _kiin a certain uncharged circuit a enter " in charging " state, this circuit must directly be connected with the region that restores electricity in electrical network; In the new electric network state of corresponding generation, line status records the charged state variable S of circuit a in B _bacorresponding change, the accumulative total charging period T of circuit a _baset to 0 all the other electric network state information and s _kiidentical;

(A2) make electric network state s _kiin a certain NBS unit b of not restarting enter " subsidiary engine restart in " state, unit b must meet following two conditions: the one, at electric network state s _kiin its machine end bus restore electricity; The 2nd, electric network state s _kiin the whole network can raise generated output P _upbe greater than the starting power of unit b; In the new electric network state of corresponding generation, generator records the starting state variable S of unit b in G _gbcorresponding change, the accumulative total of unit b is restarted period T _gbset to 0 all the other electric network state information and s _kiidentical;

(A3) make electric network state s _kiin a certain subsidiary engine restart complete NBS unit c and enter " steady combustion before emersion power " state; In the new electric network state of corresponding generation, generator records the starting state variable S of unit c in G _gccorresponding change, all the other electric network state information and s _kiidentical;

(A4) select sub-load to enter the state that restores electricity; In the new electric network state of corresponding generation, the load power that each node is restoring electricity increases by self-defining allocation strategy, all the other electric network state information and s _kiidentical;

(A5) as electric network state s _kicircuit in middle existence charging, the unit in restarting or restoring electricity load time, according to issued default T for Status Change instruction _stepduration is proceeded follow-up quantity of state adjustment, generates new electric network state.

Preferably, in described recovery operation (A4), allocation strategy specifically comprises the steps:

(B1) calculate electric network state s _kiin the whole network can raise generated output P _up, calculated load access limit value P _lP=α _lp _up, wherein, α _l≤ 1 is the default scale-up factor of algorithm;

(B2) according to electric network state s _kicarry out trend calculating, the whole network load bus is sorted from big to small by node voltage perunit value;

(B3) the front N after selected sequence _lindividual load bus is as both candidate nodes; N _lvalue principle be: if the whole network still unrecovered load total amount be less than or equal to P _lP, N _l=n _n, n _nsum for node in electrical network; If the whole network still unrecovered load total amount is greater than P _lP, N _lvalue should make front N _l-1 node still unrecovered load total amount is less than or equal to P _lP, front N _lindividual node still unrecovered load total amount is greater than P _lP;

(B4) determine the concrete load of recovering of each both candidate nodes.

Preferably, the burden with power increment of establishing o both candidate nodes and increasing is successively (D _o1, D _o2... D _oMo), amount to M _othe load of individual grade, described step (B4) further comprises following sub-step:

(C1) initialization index o=1, continues cycle labeling F _goon=0;

(C2) if o the current load restoration grade of both candidate nodes u _o=M _o, show that its all loads have all entered and waited to return to form, and skip to step (C4); Otherwise execution step (C3);

(C3) the individual node of I (o) in o the corresponding electrical network of both candidate nodes, R _{nI (x)}be x the load power that both candidate nodes is restoring electricity, be the u of o both candidate nodes _o+ 1 stage load amount is compared with the load increment of previous stage load, if the load power R o both candidate nodes being restored electricity _{nI (o)}increase one-level load, make juxtaposition u _o=u _o+ 1, F _goon=1, then perform step (C4); Otherwise directly perform step (C4);

(C4) if o+1≤N _l, return to step (C2) after making o=o+1, otherwise execution step (C5);

(C5) if F _goonbe 1, the o that resets is 1, F _goonbe to return to step (C2) after 0, otherwise stop circulation.

Preferably, described recovery operation (A5) further comprises following sub-step:

(D1) the new electric network state s* generating is initialized as and electric network state s _kiidentical state;

(D2) in new electric network state s*, the accumulative total duration of charging of all being in " in charging " status line increases a time step T _stepif the accumulative total duration of charging of arbitrary circuit is more than or equal to its default duration of charging, changing its state is " charging complete " state, accumulative total duration of charging zero clearing;

The accumulative total reboot time of all being in new electric network state s* " subsidiary engine restart in " state unit increases a time step T _step, its unit output is unit starting power P _startby step-length P _stepnegative value after discretize; If the accumulative total reboot time of arbitrary unit is more than or equal to its default start-up time, changing its state is " subsidiary engine is restarted complete " state, the zero clearing of accumulative total reboot time;

Exerting oneself by climbing rate before steady combustion and time step T of " emersion power before steady combustion " the state unit of all being in new electric network state s* _stepproduct increase, if unit output is more than or equal to its minimum technology P that exerts oneself after calculating _gminwith starting power P _startpoor, changing its state is " unit is restarted complete " state, unit output is P _gmin-P _startby step-length P _stepvalue after discretize;

In the state recording L of the load of new electric network state s*, the load power P having restored electricity of q the node of resetting _nq=P _nq+ R _nq, and by the load power R restoring electricity of q node _nqzero clearing, q=1 wherein, 2 ..., n _n, n _nsum for node in electrical network.

Preferably, in described step (7), state is cut down and is comprised successively non-mastery sequence reduction, hierarchical classification and three steps of single goal sequence reduction;

Wherein, the execution flow process of hierarchical classification is as follows:

(E1) init state collection S _k+1in the residing classification of all electric network states layer be the 1st layer;

(E2) put classification layer v=1, by default similar discriminant function f ₁() is to state set S _k+1in all electric network states classify for the first time;

(E3) if total number of categories is greater than number of categories upper limit C _max, perform step (E4), otherwise execution step (E5);

(E4) adopt default distance function d _v() calculates the distance summation between v each state class of layer and other state class, retains the C larger apart from summation _maxindividual state class, deletes other state class, and the electric network state that deleted state class is comprised is from state set S _k+1in remove, finish hierarchical classification flow process;

(E5) maximum one of electric network state quantity in selected all state class, make v equal this state class classification layer of living in, if v<3 makes v=v+1, putting the residing classification layer of all electric network states in this state class is v layer, and by default similar discriminant function f _v() reclassifies this batch of electric network state, returns to step (E3); Otherwise finish hierarchical classification flow process.

In general, the above technical scheme of conceiving by the present invention compared with prior art, has following beneficial effect:

1, flow process has covered that the whole network generator is restarted, rack recovers and the recovery decision-making of each load bus, thereby the refinement restoration schedule scheme of taking all elements into account can be provided.

2, adopt the thinking of progressively deducing black startup recovery operation, and constantly simplify and recover efficiency or the not good sequence of operation of quality by the state technology of cutting down, thereby without as the technology such as Petri net prior by the recovery condition of all elements of the whole network and mutually joining relation entirely enumerate and make decision again, improved efficiency of algorithm, reduced calculating the requirement of storage.

3, not based on any black startup recovery stage hypothesis, every step recovery operation all can freely be selected from all kinds, therefore allow as required to genset restart, rack recovers and load restoration carries out combination in any, be easy to overcome black startup line node overtension in early stage, later stage rack is not suitable with the problem of load restoration, has naturally adapted to scene after different electric network faults simultaneously.

4, in recovery operation, the status switching instruction of element is separated with implementation, thereby can adapt to parallel recovery operation.

5, method flow and concrete electric network state represent, recovery operation setting, transient state/steady state constraint verification content be irrelevant, because electric network element and the verification project of considering in the method all can increase and decrease arbitrarily on demand.

6, the enforcement of state reduction technology has avoided computation complexity excessively to increase with the increase of electrical network scale, thereby this method can be effectively applied in actual electric network, and single goal sequence reduction step wherein provides by its sequence index the approach of introducing manual decision's tendency.

Accompanying drawing explanation

Fig. 1 is the power system blackstart scheme generation method schematic flow sheet based on dynamic programming of the embodiment of the present invention;

Fig. 2 is the generation schematic diagram of new state;

Fig. 3 is optimal recovery path schematic diagram;

Fig. 4 is the conceptual schematic view of hierarchical classification;

Fig. 5 is the line chart of IEEE5 machine 14 node systems;

Fig. 6 is the relation of the state class upper limit and Riming time of algorithm;

Fig. 7 is the relation of the state class upper limit and best decision release time;

The line chart of Tu8Shi New England 10 machine 39 node systems;

Fig. 9 is load accumulation recovery curve;

Figure 10 is power system restoration process voltage magnitude curve;

Figure 11 is circuit overload multiple histogram;

Power system restoration situation when Figure 12 is the 42nd minute.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.In addition,, in each embodiment of described the present invention, involved technical characterictic just can not combine mutually as long as do not form each other conflict.

As shown in Figure 1, the power system blackstart scheme generation method based on dynamic programming of the embodiment of the present invention comprises the steps:

(1) electric network information after read failure obtains initial electric network state s ₀, and the target grid state s after recovery is set _t, with variable g (s), represent that the shortest accumulative total of arrival electric network state s is consuming time, initialization g (s ₀)=0, accumulative total recovery operation number of times (being called " progression " below) k=0 and original state collection S ₀={ s ₀, initialization terminal state set S _efor empty set, preferred plan T release time _bestfor infinity.

Wherein, electric network state s should at least comprise that generator state recording G, line status record B and load condition records L tri-parts, that is:

s＝[G B L]

Notice that in electric network state record, all performance number and time values all need according to given power and time step P _stepand T _stepcarry out discretize, but still by continuous quantity, process below no longer repeat specification while participating in computing.P in practical application _stepshould be taken as the minimum value in the whole network NBS unit (Non-Black Start Unit does not have the unit of self-startup ability) starting power and each node access load increment; T _stepsuggestion is each system element minimum value of release time, but should not be greater than each system element minimum value of release time, otherwise may cause, occurs idle period in rejuvenation.

Generator state recording G is by the recovery information I of every unit _gmform I _gmcomprise unit starting state variable S _gm, unit accumulative total restarts period T _gmwith the current meritorious P that exerts oneself of unit _gm:

$G=[I_{G1},I_{G2},...,I_{{\mathrm{Gn}}_G}]$

I _Gm＝[S _Gm,T _Gm,P _Gm],(m∈[1,n _G])

N in above formula _gfor genset sum in electrical network.

Line status records B by the recovery information I of every circuit _bpform, comprise line charging state variable S _bpand circuit adds up to charge period T _bp:

$B=[I_{B1},I_{B2},...,I_{{\mathrm{Bn}}_B}]$

I _Bp＝[S _Bp,T _Bp],(p∈[1,n _B])

N in formula _bfor the circuit sum in electrical network.

The state recording L of load is by the load restoration information vector I of each node _lqform, comprise the load power P having restored electricity _nqand the load power R restoring electricity _nq:

$L=[I_{L1},I_{L2},...,I_{{\mathrm{Ln}}_N}]$

I _Li＝[P _Nq,R _Nq],(q∈[1,n _N])

N in formula _nsum for node in electrical network.

If still there are other equipment that need to carry out emphasis consideration in rejuvenation (such as transformer etc.) in electrical network, can copy above-mentioned data structure to expand arbitrarily.

(2) to k level state set S _kin each state s _ki, determine corresponding recovery operation, these recovery operations are carried out to transient state verification, to judge the security of electric network state transient process; To each state s _ki, by each recovery operation of verification, can generate a new state, referring to Fig. 2, by these new states without repeatedly adding k+1 level state set S to _k+1in; Wherein, i=1,2 ..., n _k, n _kbe k level state set S _kthe sum of middle electric network state.

Wherein, transient state verification mainly comprises switching overvoltage, self-excitation, three projects of meritorious frequency dynamic response, need to differentiate by electromagnetic transient simulation technology.

To k level state set S _kin each state s _ki, before transient state verification, optional recovery operation at least comprises following five kinds:

(A1) make electric network state s _kiin a certain uncharged circuit a enter " in charging " state, this circuit must directly be connected with the region that restores electricity in electrical network.In the new electric network state of corresponding generation, line status records the charged state variable S of circuit a in B _bavalue change to 1 (" in charging " state), and the accumulative total of circuit a charging period T _baset to 0 all the other electric network state information and s _kiidentical.This operation is only assigned alteration command to line status, does not do actual execution, so it operates r consuming time ₁=0.

(A2) make electric network state s _kiin a certain NBS unit b of not restarting enter " subsidiary engine restart in " state.In the new electric network state of corresponding generation, generator records the starting state variable S of unit b in G _gbvalue change to 1 (" subsidiary engine restart in " state), and the accumulative total of unit b is restarted period T _gbset to 0 all the other electric network state information and s _kiidentical.This operation is only assigned alteration command to set state, does not do actual execution, so it operates r consuming time ₂=0.

It should be noted that the selected NBS unit of subsidiary engine reboot operation b must meet two conditions: the first is at electric network state s _kiin its machine end bus restore electricity, it two is electric network state s _kiin the whole network can raise generated output P _upbe greater than the starting power (P of unit b _upfor electric network state s _kimiddle the whole network unit is at time step T _stepthe load R that interior maximum climbing power summation deduction is restoring electricity _nqand the residue of the unit in " subsidiary engine restart in " state starting power E _gmafter afterpower).

(A3) make electric network state s _kiin a certain subsidiary engine restart complete NBS unit c and enter " steady combustion before emersion power " state.In the new electric network state of corresponding generation, generator records the starting state variable S of unit c in G _gcvalue change to 3 (" emersion power before steady combustion " states), all the other electric network state information and s _kiidentical.This operation is only assigned alteration command to set state, does not do actual execution, so it operates r consuming time ₃=0.

(A4) select sub-load to enter the state that restores electricity.In the new electric network state of corresponding generation, the load power that each node is restoring electricity increases by self-defining allocation strategy, all the other electric network state information and s _kiidentical.This operation is only assigned incoming instruction to sub-load, does not do actual execution, so it operates r consuming time ₄=0.

It is discrete that the load of considering actual electric network accesses always classification, and the burden with power increment that establishing o both candidate nodes may increase is successively (D _o1, D _o2... D _oMo), amount to M _othe load of individual grade, a kind of feasible load distribution strategy is the higher load of climbing capacity priority access place node voltage according to the whole network unit, its detailed allocation flow is as follows:

(B1) calculate electric network state s _kiin the whole network can raise generated output P _up, calculated load access limit value P _lPas shown in the formula:

P _LP＝α _LP _up

Wherein, α _l≤ 1 is the default scale-up factor of algorithm, and the less load restoration of its value is slower, but reserve nargin is larger, needs by actual electric network service requirement selected.

(B2) according to electric network state s _kicarry out trend calculating, the whole network load bus is sorted from big to small by node voltage perunit value.

(B3) the front N after selected sequence _lindividual load bus is as both candidate nodes, N _lvalue principle be: if the whole network still unrecovered load total amount be less than or equal to P _lP, N _l=n _n; If the whole network still unrecovered load total amount is still greater than P _lP, N _lvalue should make front N _l-1 node still unrecovered load total amount is less than or equal to P _lP, front N _lindividual node still unrecovered load total amount is greater than P _lP.

(B4) determine the concrete load of recovering of each both candidate nodes.Further comprise following sub-step:

(C1) initialization index o=1, continues cycle labeling F _goon=0;

(C2) if o the current load restoration grade of both candidate nodes u _o=M _o, show that its all loads have all entered and waited to return to form, skip to step (C4) after making o=o+1; Otherwise execution step (C3);

(C3) the individual node of I (o) in o the corresponding electrical network of both candidate nodes, R _{nI (x)}be x the load power that both candidate nodes is restoring electricity, be the u of o both candidate nodes _o+ 1 stage load amount is compared with the load increment of previous stage load, if the load power R o both candidate nodes being restored electricity _{nI (o)}increase one-level load, even juxtaposition u _o=u _o+ 1, F _goon=1, then perform step (C4); Otherwise directly perform step (C4);

(C4) if o+1≤N _l, return to step (C2) after making o=o+1, otherwise execution step (C5);

(C5) if F _goonbe 1, the o that resets is 1, F _goonbe to return to step (C2) after 0, otherwise stop circulation.

It should be noted that electric network state s must will just allow to carry out once again (A4) operation through (A5) operation after the whole zero clearings of all load powers that restoring electricity.

(A5) above four operations are assigning of instruction, change be electric network state s _kiin certain element is residing returns to form, its accumulative total release time and exert oneself and wait the generation in this operation of variation of accumulation after element state switches, thereby the physical meaning of this operation can be considered at Preset Time step-length T _stepinside no longer sending the new instruction that changes grid equipment state, and system state change is before this done to actual execution, is a time step so operate consuming time, i.e. r ₅=T _step.It should be noted that this operation and if only if electric network state s simultaneously _kicircuit in middle existence charging, the unit in restarting or restoring electricity load time just allow to carry out, the flow process of the new electric network state of generation is as follows:

(D1) the new electric network state s* generating is initialized as and electric network state s _kiidentical state;

(D2) in new electric network state s*, the accumulative total duration of charging of all being in " in charging " status line increases a time step T _stepif the accumulative total duration of charging of arbitrary circuit is more than or equal to its default duration of charging, changing its state is " charging complete " state, accumulative total duration of charging zero clearing;

The accumulative total reboot time of all being in new electric network state s* " subsidiary engine restart in " state unit increases a time step T _step, its unit output is taken as unit starting power P _startby step-length P _stepnegative value after discretize.If the accumulative total reboot time of arbitrary unit is more than or equal to its default start-up time, changing its state is " subsidiary engine is restarted complete " state, the zero clearing of accumulative total reboot time;

Exerting oneself by climbing rate before steady combustion and time step T of " emersion power before steady combustion " the state unit of all being in new electric network state s* _stepproduct increase, if unit output is more than or equal to its minimum technology P that exerts oneself after calculating _gminwith starting power P _startpoor, changing its state is " unit is restarted complete " state, unit output is taken as P _gmin-P _startby step-length P _stepvalue after discretize;

In the state recording L of the load of new electric network state s*, the load power P having restored electricity of q the node of resetting _nq=P _nq+ R _nq, and by the load power R restoring electricity of q node _nqzero clearing, q=1 wherein, 2 ..., n _n, n _nsum for node in electrical network.

Because (A5) operates the change that is usually directed to each node injecting power of electrical network, therefore exerting oneself of each unit must be carried out corresponding adjusting, thereby the distribution principle of each unit output regulated quantity while needing self-defined network load power to change, the feasible program adopting in this instructions is newly-increased load power to be assigned to each unit of " unit is restarted complete " state by the climbing capacity of unit.

It should be added that, due to (A1) (A2) (A3) (A4) four kinds of recovery operations be instruction issuing, only (A5) operation embodies the practical implementation of electric network element Status Change, therefore this method allows multiple instruction to assign rear execution simultaneously, can naturally contain the strategy of multicomponent parallel recovery.

If considered other electrical equipments except generator, circuit, load during electric network state represents, power system restoration operation can also be expanded arbitrarily beyond above-mentioned five kinds.

(3) to k+1 level state set S _k+1in each state carry out stable state verification, to judge the security of Power System Steady-state electric weight index, cannot from set, delete by the state by verification.

Wherein, stable state verification comprises marginal time verification and two projects of trend verification (being whether power frequency steady state voltage amplitude and Line Flow be out-of-limit) of each unit starting, and wherein trend verification index can directly obtain by the stabilization result of electromagnetic transient simulation.

It should be noted that in this method flow process that transient state verification and stable state verification are all relatively independent links, be therefore easy to add as required or delete the verification project of any amount, there is good expansibility.

(4) to k+1 level state set S _k+1in each state s _{(k+1) j}, by formula (1), calculate arrival state s _{(k+1) j}the shortest accumulative total g (s consuming time _{(k+1) j}) (r (s wherein _ki, s _{(k+1) j}) represent from state s _kito s _{(k+1) j}recovery operation consuming time), and by the shortest accumulative total g (s consuming time _{(k+1) j}) corresponding restoration path is as the state of arrival s _{(k+1) j}optimal recovery path, referring to Fig. 3.Wherein, j=1,2 ..., n _k+1, n _k+1be k+1 level state set S _k+1the sum of middle electric network state.

g(s _(k+1)j)＝min{g(s _ki)+r(s _ki,s _(k+1)j)} (1)

(5) if k+1 level state set S _k+1in there is qualified dbjective state, all these qualified dbjective states are moved to SOT state of termination collection S _e, make T _minfor arriving the shortest accumulative total minimum value consuming time of these qualified dbjective states, if preferred plan T release time _best>T _min, make T _best=T _min, otherwise T _bestconstant, execution step (6); If k+1 level state set S _k+1in there is not qualified dbjective state, directly execution step (6);

(6) if k+1 level state set S _k+1for sky, or k+1 level state set S _k+1the shortest accumulative total T that has all been greater than consuming time of all states of middle residue _best, by SOT state of termination collection S _ein corresponding to T _bestoptimal recovery path as black-start scheme, stop to calculate; Otherwise execution step (7).

(7) to k+1 level state set S _k+1carry out state and cut down rearmounted k=k+1, return to step (2).

State is cut down and is comprised successively non-mastery sequence reduction, hierarchical classification and three steps of single goal sequence reduction, it is implemented object and is to provide approach for introducing decision-making tendentiousness, with the newly-generated status number of seasonal this method, is in all the time in controlled scope to guarantee algorithm execution efficiency.

First in non-mastery sequence is cut down, according to a plurality of electric network state indexs of user's decision-making tendency predefine, to state set S _k+1in all electric network states calculate these desired values, only retain the state be wherein in Pareto forward position (Pareto Frontier), all the other electric network states are from state set S _k+1middle deletion.(because the implementation method of Pareto forward position and non-dominated Sorting is common in existing documents and materials, not describing in detail) recommends five electric network state index C herein herein ₁～C ₅as follows:

$\left\{\begin{array}{c}{C}_1\left(s\right)=\frac{1}{g\left(s\right)}\underset{0<p<n_B}{\mathrm{\&Sigma;}}{1}_{\{S_{\mathrm{Bp}>0\}}}\\ C_2\left(s\right)=\frac{1}{g\left(s\right)}\underset{0<m<n_G}{\mathrm{\&Sigma;}}{1}_{\left\{S_{\mathrm{Gm}>0}\right\}}\\ C_3\left(s\right)=\frac{1}{g\left(s\right)}\underset{0<q<n_N}{\mathrm{\&Sigma;}}{P}_{\mathrm{Nq}}\\ C_4\left(s\right)=\frac{1}{{(U_{\max }-1)}^2+{(U_{\min }-1)}^2+{\mathrm{\&epsiv;}}_V}\\ C_5\left(s\right)=\frac{1}{P_{\mathrm{LL}\max }+{\mathrm{\&epsiv;}}_{\mathrm{LL}}}\end{array}\right.$

In above formula: 1 _{{ }}for indicative function, g (s) is the accumulation release time of state s, U _maxand U _minmaximal value and minimum value for each node voltage perunit value in state s; P _lLmaxfor having recovered the maximum line loss value of circuit in state s; ε _vwith ε _lLfor constant, only, for preventing that denominator from being 0 to add, its value should be fully little; Index C ₁～C ₅physical meaning be respectively the average line way that starts in the unit interval to recover, in the unit interval, start the average unit number of restarting, the load of recovering in the unit interval, degree and the line losses indices of voltage deviation reference value, index more electric network state s is more excellent in this index meaning.C ₁～C ₃characterized comparatively all sidedly recovery efficiency, C ₄～C ₅weighed the quality of the current running status of electrical network.

Next is split as electric network state different classes of by its similarity by hierarchical classification, its batch total is no more than default classification upper limit C _maxin concrete enforcement, need predefine a plurality of " classification layers ", its concept is referring to Fig. 4, in different classification layers, the classification of state adopts the sorting criterion of different finenesses, layer numbering is larger, corresponding sorting criterion is meticulousr, and every layer of default similar discriminant function is for judging whether two electric network states belong to same class in the criteria for classification of this layer, and the similar discriminant function of remembering x layer is here f _x(), every layer also needs to define a same layer between class distance function to quantize the similarity degree with different conditions class in layer simultaneously, and the distance function of remembering x layer is here d _x(), provides a kind of realization example of 3 classification layers as follows here:

$\left\{\begin{array}{c}{f}_1(s^{\&prime;},s)=\underset{0<p<n_B}{\mathrm{\&Pi;}}(1_{\{S_{\mathrm{Bp}}^{\&prime;}>0\}}{1}_{\{S_{\mathrm{Bp}}>0\}}+1_{\{S_{\mathrm{Bp}}^{\&prime;}=0\}}{1}_{\{S_{\mathrm{Bp}}=0\}})\\ d_1(s^{\&prime;},s)=\frac{1}{n_B}\underset{0<p<n_B}{\mathrm{\&Sigma;}}{1}_{\left\{S_{\mathrm{Bp}=0}^{\&prime;}\right\}}{1}_{\left\{S_{\mathrm{Bp}>0}\right\}}\end{array}\right.$

$\left\{\begin{array}{c}{f}_2(s^{\&prime;},s)=f_1(s^{\&prime;},s)\underset{0<m<n_G}{\mathrm{\&Pi;}}(1_{\{S_{\mathrm{Gm}}^{\&prime;}>0\}}{1}_{\{S_{\mathrm{Gm}}>0\}}+1_{\{S_{\mathrm{Gm}}^{\&prime;}=0\}}{1}_{\{S_{\mathrm{Gm}}=0\}})\\ d_2(s^{\&prime;},s)=d_1(s^{\&prime;},s)+\frac{1}{n_G}\underset{0<m<n_G}{\mathrm{\&Sigma;}}{1}_{\left\{S_{\mathrm{Gm}=0}^{\&prime;}\right\}}{1}_{\left\{S_{\mathrm{Gm}>0}\right\}}\end{array}\right.$

$\left\{\begin{array}{c}{f}_3(s^{\&prime;},s)=f_2(s^{\&prime;},s)\underset{0<q<n_N}{\mathrm{\&Pi;}}{1}_{\{P_{\mathrm{Nq}}^{\&prime;}=P_{\mathrm{Nq}}\}}\\ d_3(s^{\&prime;},s)=d_2(s^{\&prime;},s)+\frac{2}{\underset{0<q<n_N}{\mathrm{\&Sigma;}}(P_{\mathrm{Nq}}^{\&prime;}+P_{\mathrm{Nq}})}\underset{0<q<n_N}{\mathrm{\&Sigma;}}|P_{\mathrm{Nq}}^{\&prime;}-P_{\mathrm{Nq}}|\end{array}\right.$

In above formula, s ' and s represent any two different electric network states, notice the charged state that the discriminant function of the 1st layer and distance function are only taken into account power network line, 2nd, 3 layers are added the comparison to electrical network set state, load restoration state on its basis successively, thereby have realized the progressively increase of classification resolution.

According to above definition, the execution flow process of hierarchical classification is as follows:

(E1) init state collection S _k+1in the residing classification of all electric network states layer be the 1st layer;

(E2) put classification layer v=1, by default similar discriminant function f ₁() is to state set S _k+1in all electric network states classify for the first time;

(E3) if total number of categories is greater than number of categories upper limit C _max, perform step (E4), otherwise execution step (E5);

(E4) adopt default distance function d _v() calculates the distance summation between v each state class of layer and other state class, retains the C larger apart from summation _maxindividual state class, deletes other state class, and the electric network state that deleted state class is comprised is from state set S _k+1in remove, finish hierarchical classification flow process.

(E5) maximum one of electric network state quantity in selected all state class, make v equal this state class classification layer of living in, if v<3 makes v=v+1, putting the residing classification layer of all electric network states in this state class is v layer, and by default similar discriminant function f _v() reclassifies this batch of electric network state, returns to step (E3); Otherwise finish hierarchical classification flow process.

Hierarchical classification is carried out rear state set S _k+1in state will be included into quantity and be no more than C _maxstate class in, guaranteed the status flag diversity between these state class simultaneously.

What state was cut down finally passes through a final index C by single goal sequence _felectric network state in each class is sorted, only retain the front R of desired value the best in such _maxindividual state, a kind of feasible quality of voltage index (desired value is the bigger the better) that is chosen as, (U in formula is shown below _qbe the voltage perunit value of q node, n _nrecfor the whole network recovery nodes sum):

$C_F\left(s\right)=\frac{n_{\mathrm{Nrec}}}{\mathrm{\&Sigma;}{(U_q-1)}^2+{\mathrm{\&epsiv;}}_V},(q\&Element;[1,n_N\left]\right)$

In sum, the state of this method technology of cutting down is cut down by non-mastery sequence and has been introduced decision-making tendentiousness, by hierarchical classification and single goal sequence, cut down and guaranteed Guarantee Status collection S of diversified while of electric network state _k+1in state sum be no more than C _maxr _max, thereby guaranteed the efficiency of decision-making technique.

The state that it should be noted that cut down in the index that adopts of three steps can adjust according to actual needs or supplement, the scheme of the present embodiment is only a kind of implementation wherein.

For making those skilled in the art understand better the present invention, below in conjunction with specific embodiment, power system blackstart scheme generation method of the present invention is elaborated.

One, algorithm efficiency of the practice is analyzed

The efficiency of the practice of the inventive method is carried out to validation verification below, hardware environment is Core Duo T65002.10Ghz, DDR3RAM 2.00GB, algorithm implementation is C++ and Matlab hybrid programming, explanation in each index that electric network state, recovery operation, unit output allocation strategy, load restoration strategy, state reduction adopt and above flow process is consistent, and does not make other complement and amendmentses.

In Analysis of operation efficiency, the example of employing is IEEE4 machine 14 node systems, and as shown in Figure 5, concrete systematic parameter adopts associated data files in matpower 4.0 kits to its line chart.In example when test, all supposes that system initial state is for black (all devices powers failure) entirely, and in system, each node load is pressed the discrete value of 2MW, and permission is 3 circuit parallel recoveries at the most, and algorithm parameter is got and determined P _step=2MW, T _step=2min, α L=1.0 arranges constant, not considering transient verification, the required trend of stable state verification is calculated and is completed by calling matpower 4.0 kits.

Algorithm employing state reduction technology of the present invention guarantees efficiency of algorithm, wherein has two parameters: state class upper limit C _maxand the status number upper limit R of each state class _max.On IEEE 14 node systems by adjustment algorithm Verification the two affect result as shown in Fig. 6, Fig. 7 and table 1.

The dynamic programming algorithm that executing state is cut down, because long operational time (working time was over 10 hours), consumption internal memory are large, does not therefore almost have practical value.As seen from Figure 6, the enforcement that state is cut down has been played conclusive effect to the raising of algorithm speed, and the state class upper limit is less in general, and Riming time of algorithm is shorter.

As seen from Figure 7, state class is counted quantitative limitation and may be caused the result of decision to occur that (the known optimal recovery time is 76min for to a certain degree deteriorated, enforcement state cut down after time of obtaining of decision-making roughly within the scope of 80～84min, much smaller than each element release time of cumulative upper limit 361min in turn).And owing to may there being responsive branch in example decision process, the too low status number upper limit will cause unstable (two histogram empty partly represent the failure of algorithm search scheme) of the result of decision.

Table 1R _maxeffect test

In upper table, C _max=35.

From table 1, if counting the upper limit, state classification selects rationally, the state upper limit retaining in each classification does not affect the result of decision substantially, but be conducive to ground, further improves speed of decision.

Two, decision scheme benefit analysis

Below the advantage of the inventive method decision scheme is analyzed, adopted New England's 10 machine 39 node systems as shown in Figure 8 to discuss, getting algorithm parameter is P _step=2MW, T _step=2min, α _l=1.0, Cmax=30, Rmax=1, New England's 10 machine 39 node systems are carried out to black startup decision-making, through 1049s, can be calculated optimal recovery scheme consuming time for 456min, much smaller than each element release time of cumulative upper limit 1411min in turn, existing with regard to its result, to carry out labor as follows:

1) equipment restoration schedule

In system, each unit restoration schedule is as shown in table 2 below.Each unit reboot time all meets the constraint of default marginal time, wherein G9～10 because Ji Duan supply line susceptance is excessive, G4～7 are owing to crossing far and causing reboot time more late to black startup unit electrical distance.

Table 2 genset reboot time

2) load restoration process

As shown in Figure 9, visible load restoration has run through the rejuvenation of whole black startup to the load accumulation recovery curve (not counting each set auxiliary machinery starting power) of black start-up course in Optimal Decision-making.Curve has the mild section in number place, actually corresponding to starting, carries out output distribution before unit and steady combustion take the process of Optimal Power Flow distribution in climbing between unit: wherein 40～64min, 80～96min place adjust as G1 and G2, G8 unit; Within 142～242 minutes, locate for starting the adjustment process between unit and G3～7 unit.

After 270 minutes, because part of generating units has reached maximum output, and the further load restoration of part of nodes need to wait for that the voltage after G9 and G10 are restarted maintains, and load access speed totally slows down.

3) voltage magnitude

In recovery operation overall process, the maximal value of the whole network voltage magnitude and minimum value curve are as shown in figure 10, visible each node can remain in the normal range of 0.9～1.1p.u. all the time, wherein 0～70 minute stage is because the light-loaded circuit of new charging is more, voltage max is higher, and the high value of voltage occurring for 138 and 196 minutes is respectively due to due to 43, No. 45 large susceptance line chargings.

4) recover rear operating mode

Electrical network operating mode after black startup recovers and the contrast of former operating condition are as shown in table 3 below.

Electrical network operating mode after table 3 recovers

From table 3 and Figure 11, each unit output of electrical network that adopts the inventive method to carry out after recovering is reasonable, close with former operating mode on via net loss, therefore can think that recovery operation is comparatively successful.

5) decision scheme advantage

One, the inventive method is by the load superpotential of inhibition initial stage supply path of proper restoration, if otherwise do not consider load access, in setting up the process of supply path, serious superpotential will be there is, its maximal value changes as shown in figure 10, in the time of 20 minutes, 4,5,6,16, No. 17 node starts to occur slight voltage out-of-limit, and 3, No. 18 node also reaches the voltage magnitude upper bound; In the time of 22 minutes, all the other node voltages all seriously out-of-limit (maximum 1.226p.u.) except No. 30 nodes, black startup cannot be proceeded.

They are two years old, the inventive method can be taken into account the common load outside the necessary path of recovering to restart NBS unit in black initial start stage, (recovered part overstriking represents) as shown in figure 12, in the time of 42 minutes, except G2 and G8, all the other generators are due in supply path charging or because overvoltage problem can not obtain instant recovery, and existing unit available has obvious redundancy, therefore in algorithm decision-making to 7, 8, the important load centers such as 18 nodes have recovered power supply, both be conducive to make full use of the time slot during unit supply path is set up, also for climbing before the steady combustion of unit provides how available balanced load.

Above-mentioned analytic explanation the advantage that hockets as required of the recovery of generator, circuit and load, and the method that the present invention proposes can be coordinated this effectively.

Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any modifications of doing within the spirit and principles in the present invention, be equal to and replace and improvement etc., within all should being included in protection scope of the present invention.

Claims (6)

1. a power system blackstart scheme generation method, is characterized in that, comprises the steps:

(1) electric network information after read failure obtains initial electric network state s ₀, with variable g (s), represent that the shortest accumulative total of arrival electric network state s is consuming time, initialization g (s ₀)=0, progression k=0, original state collection S ₀={ s ₀, initialization terminal state set S _efor empty set, preferred plan T release time _bestfor infinity;

(2) to k level state set S _kin each state s _ki, determine corresponding recovery operation, these recovery operations are carried out to transient state verification, to judge the security of electric network state transient process; To each state s _ki, by each recovery operation of verification, can generate a new state, by these new states without repeatedly adding k+1 level state set S to _k+1in; Wherein, i=1,2 ..., n _k, n _kbe k level state set S _kthe sum of middle electric network state;

(3) to k+1 level state set S _k+1in each state carry out stable state verification, to judge the security of Power System Steady-state electric weight index, cannot from set, delete by the state by verification;

(4) to k+1 level state set S _k+1in each state s _{(k+1) j}, calculate the shortest accumulative total g (s consuming time that arrives this state _{(k+1) j})=min{g (s _ki)+r (s _ki, s _{(k+1) j}), and by the shortest accumulative total g (s consuming time _{(k+1) j}) corresponding restoration path is as the state of arrival s _{(k+1) j}optimal recovery path; Wherein, r (s _ki, s _{(k+1) j}) represent from state s _kito s _{(k+1) j}recovery operation consuming time, j=1,2 ..., n _k+1, n _k+1be k+1 level state set S _k+1the sum of middle electric network state;

(5) if k+1 level state set S _k+1in there is qualified dbjective state, all these qualified dbjective states are moved to SOT state of termination collection S _e, make T _minfor arriving the shortest accumulative total minimum value consuming time of these qualified dbjective states, if preferred plan T release time _best>T _min, make T _best=T _min, otherwise T _bestconstant, execution step (6); If k+1 level state set S _k+1in there is not qualified dbjective state, directly execution step (6);

(6) if k+1 level state set S _k+1for sky, or k+1 level state set S _k+1the shortest accumulative total T that has all been greater than consuming time of all states of middle residue _best, by SOT state of termination collection S _ein corresponding to T _bestoptimal recovery path as black-start scheme, stop to calculate; Otherwise execution step (7);

(7) to k+1 level state set S _k+1carry out state and cut down rearmounted k=k+1, return to step (2).

2. power system blackstart scheme generation method as claimed in claim 1, is characterized in that, in described step (2), to k level state set S _kin each state s _ki, before transient state verification, optional recovery operation at least comprises following five kinds:

(A1) make electric network state s _kiin a certain uncharged circuit a enter " in charging " state, this circuit must directly be connected with the region that restores electricity in electrical network; In the new electric network state of corresponding generation, line status records the charged state variable S of circuit a in B _bacorresponding change, the accumulative total charging period T of circuit a _baset to 0 all the other electric network state information and s _kiidentical;

(A2) make electric network state s _kiin a certain NBS unit b of not restarting enter " subsidiary engine restart in " state, unit b must meet following two conditions: the one, at electric network state s _kiin its machine end bus restore electricity; The 2nd, electric network state s _kiin the whole network can raise generated output P _upbe greater than the starting power of unit b; In the new electric network state of corresponding generation, generator records the starting state variable S of unit b in G _gbcorresponding change, the accumulative total of unit b is restarted period T _gbset to 0 all the other electric network state information and s _kiidentical;

(A3) make electric network state s _kiin a certain subsidiary engine restart complete NBS unit c and enter " steady combustion before emersion power " state; In the new electric network state of corresponding generation, generator records the starting state variable S of unit c in G _gccorresponding change, all the other electric network state information and s _kiidentical;

(A4) select sub-load to enter the state that restores electricity; In the new electric network state of corresponding generation, the load power that each node is restoring electricity increases by self-defining allocation strategy, all the other electric network state information and s _kiidentical;

(A5) as electric network state s _kicircuit in middle existence charging, the unit in restarting or restoring electricity load time, according to issued default T for Status Change instruction _stepduration is proceeded follow-up quantity of state adjustment, generates new electric network state.

3. power system blackstart scheme generation method as claimed in claim 2, is characterized in that, in described recovery operation (A4), allocation strategy specifically comprises the steps:

(B1) calculate electric network state s _kiin the whole network can raise generated output P _up, calculated load access limit value P _lP=α _lp _up, wherein, α _l≤ 1 is the default scale-up factor of algorithm;

(B2) according to electric network state s _kicarry out trend calculating, the whole network load bus is sorted from big to small by node voltage perunit value;

(B3) the front N after selected sequence _lindividual load bus is as both candidate nodes; N _lvalue principle be: if the whole network still unrecovered load total amount be less than or equal to P _lP, N _l=n _n, n _nsum for node in electrical network; If the whole network still unrecovered load total amount is greater than P _lP, N _lvalue should make front N _l-1 node still unrecovered load total amount is less than or equal to P _lP, front N _lindividual node still unrecovered load total amount is greater than P _lP;

(B4) determine the concrete load of recovering of each both candidate nodes.

4. power system blackstart scheme generation method as claimed in claim 3, is characterized in that, the burden with power increment that establishing o both candidate nodes increases is successively (D _o1, D _o2... D _oMo), amount to M _othe load of individual grade, described step (B4) further comprises following sub-step:

(C1) initialization index o=1, continues cycle labeling F _goon=0;

(C2) if o the current load restoration grade of both candidate nodes u _o=M _o, show that its all loads have all entered and waited to return to form, and skip to step (C4); Otherwise execution step (C3);

(C3) the individual node of I (o) in o the corresponding electrical network of both candidate nodes, R _{nI (x)}be x the load power that both candidate nodes is restoring electricity, be the u of o both candidate nodes _o+ 1 stage load amount is compared with the load increment of previous stage load, if the load power R o both candidate nodes being restored electricity _{nI (o)}increase one-level load, make juxtaposition u _o=u _o+ 1, F _goon=1, then perform step (C4); Otherwise directly perform step (C4);

(C4) if o+1≤N _l, return to step (C2) after making o=o+1, otherwise execution step (C5);

(C5) if F _goonbe 1, the o that resets is 1, F _goonbe to return to step (C2) after 0, otherwise stop circulation.

5. the power system blackstart scheme generation method as described in any one in claim 2 to 4, is characterized in that, described recovery operation (A5) further comprises following sub-step:

(D1) the new electric network state s* generating is initialized as and electric network state s _kiidentical state;

(D2) in new electric network state s*, the accumulative total duration of charging of all being in " in charging " status line increases a time step T _stepif the accumulative total duration of charging of arbitrary circuit is more than or equal to its default duration of charging, changing its state is " charging complete " state, accumulative total duration of charging zero clearing;

The accumulative total reboot time of all being in new electric network state s* " subsidiary engine restart in " state unit increases a time step T _step, its unit output is unit starting power P _startby step-length P _stepnegative value after discretize; If the accumulative total reboot time of arbitrary unit is more than or equal to its default start-up time, changing its state is " subsidiary engine is restarted complete " state, the zero clearing of accumulative total reboot time;

Exerting oneself by climbing rate before steady combustion and time step T of " emersion power before steady combustion " the state unit of all being in new electric network state s* _stepproduct increase, if unit output is more than or equal to its minimum technology P that exerts oneself after calculating _gminwith starting power P _startpoor, changing its state is " unit is restarted complete " state, unit output is P _gmin-P _startby step-length P _stepvalue after discretize;

In the state recording L of the load of new electric network state s*, the load power P having restored electricity of q the node of resetting _nq=P _nq+ R _nq, and by the load power R restoring electricity of q node _nqzero clearing, q=1 wherein, 2 ..., n _n, n _nsum for node in electrical network.

6. power system blackstart scheme generation method as claimed in claim 1, is characterized in that, in described step (7), state is cut down and comprised successively non-mastery sequence reduction, hierarchical classification and three steps of single goal sequence reduction;

Wherein, the execution flow process of hierarchical classification is as follows:

(E1) init state collection S _k+1in the residing classification of all electric network states layer be the 1st layer;

(E2) put classification layer v=1, by default similar discriminant function f ₁() is to state set S _k+1in all electric network states classify for the first time;

(E3) if total number of categories is greater than number of categories upper limit C _max, perform step (E4), otherwise execution step (E5);

(E4) adopt default distance function d _v() calculates the distance summation between v each state class of layer and other state class, retains the C larger apart from summation _maxindividual state class, deletes other state class, and the electric network state that deleted state class is comprised is from state set S _k+1in remove, finish hierarchical classification flow process;

(E5) maximum one of electric network state quantity in selected all state class, make v equal this state class classification layer of living in, if v<3 makes v=v+1, putting the residing classification layer of all electric network states in this state class is v layer, and by default similar discriminant function f _v() reclassifies this batch of electric network state, returns to step (E3); Otherwise finish hierarchical classification flow process.