CN105205549A - Light-preserved system tracking day-ahead plan scheduling method based on chance constrained programming - Google Patents

Abstract

The invention provides a light-preserved system tracking day-ahead plan scheduling method based on chance constrained programming. The method comprises the following steps that (1) the relevant data of a photovoltaic power station and an energy storage system are read; (2) day-ahead actual photovoltaic power generation power is simulated via a monte carlo technique and range of photovoltaic planed output upper and lower limit is set; (3) a chance constrained programming mathematical model including a control coefficient is established based on a short-term photovoltaic prediction power value and a random prediction deviation value; and (4) charge and discharge power of the energy storage system is determined by adopting an improved self-adaptive particle swarm algorithm. Polling of each forecast point is performed one day in advance, and an effect that light-preserved tracking planed output is in the range of the planed upper and lower limit is realized by adopting the improved self-adaptive particle swarm algorithm; besides, tracking control is enabled to be more flexible by adjustment of the control coefficient in a target function, and charge and discharge power and the state of charge of the energy storage system are basically maintained within an appropriate range so that charge and discharge capability is enhanced and the requirement for energy storage is reduced.

Description

A kind of light-preserved system based on chance constrained programming follows the tracks of planned dispatching method a few days ago

Technical field

The present invention relates to one planned dispatching method a few days ago, be specifically related to a kind of light-preserved system based on chance constrained programming and follow the tracks of planned dispatching method a few days ago.

Background technology

Sun power has been acknowledged as one of following most competitive power energy, has the feature such as aboundresources, environmental protection.According to International Energy Agency (IEA) prediction, to the year two thousand fifty solar energy power generating by accounting for 20% ~ 25% of global generated energy, become one of basic energy resource.But photovoltaic generation is intermittent energy source; affect by intensity of solar radiation, environment temperature etc.; its output power has uncertainty; usually harmful effect can be caused to the quality of power supply, power supply reliability and stability, grid benefit etc. when grid-connected; predict by exerting oneself to photovoltaic plant; contribute to the cooperation of electric power system dispatching department overall arrangement conventional energy resources and photovoltaic generation; timely adjustment operation plan; reasonable arrangement power system operating mode; the impact of effective reduction photovoltaic access on electric system, thus improve stability and the security of operation of power networks.But at present, because photovoltaic prediction is subject to effect of natural conditions, still have that prediction deviation is excessive, the immature problem of forecasting techniques.Not enough for making up this, nowadays from utilize energy storage technology to be formed angle that the plan of light storage association system tracking in time exerts oneself to reduce photovoltaic predicated error size, indirectly improve precision of prediction and become new study hotspot.

At present, launched multinomial research both at home and abroad for photovoltaic forecasting techniques, be mostly single consideration wind storage situation, the analysis based on light storage use in conjunction is then little, and the research of especially exerting oneself for accumulator system control realization light storage tracking plan is then less.Prior art has proposition to exert oneself with wind-light storage to be the control strategy of target to the maximum with standing plans power degree of closeness a few days ago, but the method have ignored scene prediction bound scope, each calculating can only fix formulation objective plan, not only makes energy storage control to lose dirigibility but also the requirement that adds accumulator system and cost.Also technology is had to propose to follow the tracks of based on ultra-short term wind predicted power the control strategy of plan, although the method can realize real-time follow-up by the change control coefrficient that rolls, but only considered ultra-short term situation and wind-powered electricity generation scope, not exerting oneself to short-term conditions and photovoltaic tracking plan performs an analysis.

Summary of the invention

In order to overcome above-mentioned the deficiencies in the prior art, the invention provides a kind of light-preserved system based on chance constrained programming and following the tracks of planned dispatching method a few days ago.This method makes light store up to exert oneself and be limited in for target in intended scope, by controlling accumulator system flexibly, making its charge-discharge electric power and state-of-charge all remain on optimum range and at utmost reducing energy system requirements effectively to control accumulator system.

In order to realize foregoing invention object, the present invention takes following technical scheme:

Light-preserved system based on chance constrained programming follows the tracks of a planned dispatching method a few days ago, and described method comprises the steps:

(1) related data of photovoltaic plant and accumulator system is read;

(2) formulate photovoltaic meter mark power bound scope by photovoltaic actual power power before Monte Carlo technique simulating sun;

(3) set up based on short-term photovoltaic predicted power value, stochastic prediction departure the chance constrained programming mathematical model containing control coefrficient;

(4) improvement APSO algorithm determination accumulator system charge-discharge electric power is adopted.

Preferably, in described step (1), described related data comprises: short-term photovoltaic predicted power value, accumulator system charge-discharge electric power upper lower limit value, state-of-charge upper lower limit value.

Preferably, described step (2) comprises the steps:

Step 2-1, Monte Carlo technique simulation process is adopted to be stochastic variable ξ (t) photovoltaic prediction deviation;

Step 2-2, using described short-term photovoltaic predicted power value as determining variable;

Step 2-3, described formulation photovoltaic meter mark power bound scope.

Preferably, in described step 2-1, described photovoltaic prediction deviation be photovoltaic generation a few days ago predicted power and the same day photovoltaic actual power power difference, photovoltaic actual power power P _actt () is by formula P _act(t)=P _pret ()+ξ (t) obtains, P _pret () is the predicted power of t photovoltaic generation a few days ago, the probability distribution function of described stochastic variable ξ (t) adopts that to meet average be 0, and variance is σ ²t the Normal probability distribution of (), σ (t) is by formula σ (t)=0.2P _pre(t)+0.02C _aptry to achieve, C _apfor photovoltaic installed capacity.

Preferably, in described step 2-3, described photovoltaic meter marks power bound scope with described short-term photovoltaic predicted power value for working out according to according to ± 25% fluctuation, and described photovoltaic meter marks power bound and calculated by following formula:

P _limit=ξ _allowc _ap

P _{plan_up}(t)＝P _pre(t)+P _limit

P _{plan_dn}(t)＝P _pre(t)-P _limit

In formula, ξ _allowfor predicated error allows percentage value, by being defined as 25% a few days ago; P _limitfor ξ _allowthe photovoltaic determined goes out fluctuation limit value; C _apfor photovoltaic installed capacity; P _{plan_up}t () marks power higher limit, P for photovoltaic meter _{plan_dn}t () marks power lower limit for photovoltaic meter.

Preferably, in described step (3), comprise the steps:

Step 3-1, to exert oneself within the scope of bound is exerted oneself in plan as target at utmost to make light store up, set up objective function as shown in the formula:

p _{plan_adj}(t)＝u{P _{plan_up}(t)+c[P _{plan_up}(t)-P _{plan_dn}(t)]}

$\min f=\sqrt{\underset{t=1}{\overset{m}{\mathrm{\Σ}}}[P_{\mathrm{bess}}\left(t\right)+P_{\mathrm{pre}}\left(t\right)+\mathrm{\ξ}\left(t\right)-P_{\mathrm{plan}\_\mathrm{adj}}\left(t\right){]}^2}$

P in formula _{plan_adj}t () is target control power, P _{plan_up}t () marks power higher limit, P for photovoltaic meter _{plan_dn}t () marks power lower limit for photovoltaic meter, f is objective function, and u controls the switching coefficient whether energy storage work, and it is that accumulator system is in running order that u gets 1, is in idle condition when u gets 0, and c is the target power control coefrficient between 0 to 1; P _besst () is decision variable, i.e. t energy storage charge-discharge electric power, P _besst () is greater than zero, represent energy storage device electric discharge, is less than null representation charging;

Step 3-2, set up chance constrained programming condition;

Step 3-3, set up accumulator system constraint condition.

Preferably, in described step 3-2, by photovoltaic active power output smoothing rate η _eerabsolute value be less than or equal to allowed band δ as chance constrained programming condition, make the probability P that it is set up _rbe not less than confidence level a, as shown in the formula:

η _eer＝P _pre(t)+P _bess(t)+ξ(t)-P _{plan_adj}(t)/P _{plan_adj}(t)

P _r{|η _eer|≤δ}≥α

In formula, P _rfor photovoltaic active power output smoothing rate η _eerthe probability set up when being less than or equal to allowed band δ.

Preferably, in described step 3-3, described accumulator system constraint condition comprises:

Power constraints, when namely charging, P _ch.max≤ P _bess(t)≤0, during electric discharge, 0≤P _bess(t)≤P _dis.max, in formula, P _ch.manfor the maximum charge power that negative value is accumulator system; P _dis.maxfor on the occasion of the maximum discharge power being accumulator system;

SOC constraint condition, SOC _min≤ SOC (t)≤SOC _max, in formula, SOC _minfor the minimum value of accumulator system state-of-charge; SOC _maxfor the maximal value of accumulator system state-of-charge; SOC (t) is the state-of-charge of t; ζ is corresponding discharge and recharge coefficient, and ζ > 1 during electric discharge, illustrates in discharge process to there is certain loss, and ζ < 1 during charging, illustrates in charging process also there is certain loss; Δ t is the sampling time interval of power; C is the rated capacity of accumulator system.

Preferably, in described power constraints, P is worked as _act(t) > P _{plan_up}t, time (), accumulator system is in charged state, i.e. P _ch.max≤ P _bess(t)≤0; Work as P _act(t) < P _{plan_dn}t, time (), accumulator system is in discharge condition, i.e. 0≤P _bess(t)≤P _dis.max.

Preferably, in described step (4), comprise the steps:

The parameter of step 4-1, setting particle swarm optimization algorithm, comprising: population sum N, iterations K, maximum iteration time k _max, inertia constant ω and Studying factors c ₁and c ₂;

The position of step 4-2, initialization population and speed;

Step 4-3, determine the fitness of each particle in population according to objective function f;

The fitness of step 4-4, more each particle, determines the optimum P of the individuality of each particle _best, from whole individual optimum P _bestin determine global optimum G _best; Speed and the position of each particle is upgraded according to individual optimum and global optimum;

Step 4-5, again calculate each particle fitness now according to objective function f, judge whether to upgrade individual optimum P _bestwith global optimum G _best;

Step 4-6, judge whether Search Results reaches iterations, if do not reach, continue the speed and the position that upgrade each particle; Otherwise stopping iteration, exports optimum solution.

Preferably, in described step 4-4, the speed of each particle of described renewal and position, formula is as follows:

$v_{\mathrm{id}}^{k+1}=K[\mathrm{\&omega;}{v}_{\mathrm{id}}^k+c_1{r}_1(P_{\mathrm{id}}^k-x_{\mathrm{id}}^k)+c_2{r}_2(G_{\mathrm{id}}^k-x_{\mathrm{id}}^k)$

$x_{\mathrm{id}}^{k+1}=x_{\mathrm{id}}^k+v_{\mathrm{id}}^{k+1}$

In formula, be respectively iteration to the kth generation speed of i-th particle and position; be respectively iteration to kth+1 generation i-th particle speed and position; for iteration is to the individual extreme value of kth generation i-th particle; for the global extremum of population in front k generation; c ₁, c ₂for Studying factors, it can be avoided being absorbed in local optimum by convergence speedup; r ₁, r ₂it is the random number between [0,1]; K is constraint factor, in formula wherein k is current particle iterations; K _maxfor particle cluster algorithm starts the maximum iteration time of setting; ω _min, ω _maxminimum and maximum inertia weight respectively.

Preferably, in described step 4-5, described in judge whether to upgrade individual optimum P _bestwith global optimum G _bestmode be: if $f\left(x_{\mathrm{id}}^{k+1}\right)<f\left(x_{\mathrm{id}}^k\right)$ Set up, then $p_{\mathrm{id}}^k=x_i^{k+1};$ Otherwise, $P_{\mathrm{id}}^k=x_i^k,$ Then $G_{\mathrm{id}}^k=\min \left(P_{\mathrm{id}}^k\right).$

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

The invention provides a kind of light-preserved system based on chance constrained programming and follow the tracks of planned dispatching method a few days ago, the method, by shifting to an earlier date each forecast point of poll on the one, adopts improvement APSO algorithm to achieve light storage tracking plan and exerts oneself in plan bound range effect; In addition by the adjustment to control coefrficient in objective function, make tracing control more flexible, charge-discharge electric power and the state-of-charge of accumulator system remain on optimum range all substantially, improve charging and discharging capabilities, reduce energy storage requirement.

Accompanying drawing explanation

Fig. 1 is the process flow diagram that a kind of light-preserved system based on chance constrained programming provided by the invention follows the tracks of planned dispatching method a few days ago.

Fig. 2 be in the embodiment of the present invention one day photovoltaic predicted power, plan bound and day front simulation real power curve map;

Fig. 3 be in the embodiment of the present invention in fixed coefficient situation control coefrficient c, u at intraday change curve;

Fig. 4 be in the embodiment of the present invention in fixed coefficient situation the c=0.5 time store up tracking plan and to exert oneself design sketch;

Fig. 5 be in the embodiment of the present invention in fixed coefficient situation the c=0.2 time store up tracking plan and to exert oneself design sketch;

Fig. 6 be in the embodiment of the present invention in variation factor situation control coefrficient c, u at intraday change curve;

Fig. 7 is that in the embodiment of the present invention, in variation factor situation, light storage tracking plan is exerted oneself design sketch;

Fig. 8 is SOC change curve in each situation in the embodiment of the present invention;

Fig. 9 is accumulator system charge-discharge electric power figure in variation factor situation in the embodiment of the present invention;

Figure 10 is algorithm convergence curve map in the embodiment of the present invention;

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in further detail.

In order to solve the problem in prior art, the discharge and recharge of accumulator system being lacked to consideration of overall importance, the embodiment of the present invention proposes a kind of light-preserved system based on chance constrained programming and follows the tracks of planned dispatching method a few days ago, the method is first worked out photovoltaic meter by short-term forecasting power and is marked power bound scope, consider prediction deviation randomness, real power before employing Monte Carlo technique simulating sun, then the chance constrained programming mathematical model containing control coefrficient is set up, finally utilize and improve APSO algorithm acquisition accumulator system charge-discharge electric power, and then improve light storage associating tracking plan capacity and the requirement reduced accumulator system.Accumulator system mentioned by the method can be Power Flow, mechanical energy storage or electrochemical energy storage, is described in the present embodiment for battery energy storage system, as shown in Figure 1, comprises the steps:

The related data of step 1, reading photovoltaic plant and accumulator system, comprising: short-term photovoltaic predicted power value, accumulator system charge-discharge electric power upper lower limit value, state-of-charge upper lower limit value;

The predicted time yardstick of short-term photovoltaic power predicted value is 24h, and amount to 96 periods, predicted time resolution is 15min, that is: be the grid-connected power prediction to following 24 hours, every 15 minutes forecast points, every day rail vehicle roller test-rig once;

Step 2, by photovoltaic before Monte Carlo technique simulating sun actual exert oneself and formulate photovoltaic meter mark power bound scope;

Prior art is many does not specifically consider prediction deviation randomness, have ignored prediction bound scope.For this reason, by adopting Monte Carlo simulation to be treated to stochastic variable ξ (t) photovoltaic prediction deviation in the present embodiment, the prediction of short-term photovoltaic being exerted oneself to be treated to and determines variable, the actual P that exerts oneself of photovoltaic _actt () is by formula P _act(t)=P _pret ()+ξ (t) obtains.Photovoltaic meter marks power bound model with short-term photovoltaic power generation power prediction value for working out according to ± 25% fluctuation according to according to " specification " requirement, and through type (1)-(3) calculate:

P _limit=ξ _allowc _ap(1)

P _{plan_up}(t)＝P _pre(t)+P _limit(2)

P _{plan_dn}(t)＝P _pre(t)-P _limit(3)

In formula, ξ _allowfor predicated error allows percentage value, by being defined as 25% a few days ago; P _limitfor ξ _allowthe photovoltaic determined goes out fluctuation limit value; C _apfor photovoltaic installed capacity; P _pret () is t photovoltaic predicted power a few days ago.

Monte Carlo simulation is adopted to be treated to stochastic variable photovoltaic prediction deviation.

It is 0 that the probability distribution function employing of stochastic variable ξ (t) meets average, and variance is σ ²t the Normal probability distribution of (), σ (t) is by formula σ (t)=0.2P _pre(t)+0.02C _aptry to achieve.

Step 3, to set up containing the chance constrained programming mathematical model of control coefrficient based on short-term photovoltaic predicted power value, stochastic prediction departure.

Storing up at utmost to make light to exert oneself within the scope of bound is exerted oneself in plan as target, set up objective function as shown in the formula:

p _{plan_adj}(t)＝u{P _{plan_up}(t)+c[P _{plan_up}(t)-P _{plan_dn}(t)]}(4)

$\min f=\sqrt{\underset{t=1}{\overset{m}{\mathrm{\&Sigma;}}}[P_{\mathrm{bess}}\left(t\right)+P_{\mathrm{pre}}\left(t\right)+\mathrm{\&xi;}\left(t\right)-P_{\mathrm{plan}\_\mathrm{adj}}\left(t\right){]}^2}---\left(5\right)$

P in formula _{plan_adj}t () is target control power, f is objective function, and u is the switching coefficient whether control energy storage works, and getting 1 is that accumulator system is in running order, is in idle condition when getting 0, and c is the target power control coefrficient between 0 to 1; P _besst () is decision variable, i.e. t energy storage charge-discharge electric power, P _besst () is greater than zero, represent energy storage device electric discharge, be less than zero.

Step 3.1, set up chance constrained programming condition;

Using the absolute value of photovoltaic active power output smoothing rate not higher than allowed band δ as chance constrained programming condition, the probability making it set up is not less than confidence level α, such as formula (6)-(7):

η _eer＝[p _pre(t)+p _bess(t)+ξ(t)-p _{plan_adj}(t)]/p _{plan_adj}(t)(6)

p _r{|η _eer|≤δ}≥α(7)

Step 3.2, set up accumulator system constraint condition

Power constraints, when namely charging, P _ch.max≤ P _bess(t)≤0, during electric discharge, 0≤P _bess(t)≤P _dis.max, in formula, P _ch.maxfor the maximum charge power that negative value is accumulator system; P _dis.maxfor on the occasion of the maximum discharge power being accumulator system;

SOC constraint condition, SOC _min≤ SOC _min(t)≤SOC _min, in formula, SOC _minfor the minimum value of accumulator system state-of-charge; SOC _manfor the maximal value of accumulator system state-of-charge; SOC _tit is the state-of-charge of t.

Step 3.3, power constraints to be improved

Work as P _act(t) > P _{plan_up}t, time (), accumulator system is in charged state, i.e. P _ch.max≤ P _bess(t)≤0;

Work as P _act(t) < P _{plan_dn}t, time (), accumulator system is in discharge condition, i.e. 0≤P _bess(t)≤P _dis.max.

Step 4, employing improve APSO algorithm determination accumulator system charge-discharge electric power

Concrete calculation process is as follows:

The parameter of step 4.1, setting particle swarm optimization algorithm, comprising: population sum N, iterations k, maximum iteration time k _max, inertia constant ω and Studying factors c ₁and c ₂;

The position of step 4.2, beginningization population and speed.Initialization particle rapidity, by corresponding energy storage charge-discharge electric power of each particle of accumulator system constraint condition random selecting each period, and carry out chance constrained programming condition Verification, if meet, repeat this step and carry out all particle positions of initialization, otherwise proceed random selecting until checking meets.

The objective function of step 4.3, through type (5) determines the fitness of each particle.

Step 4.4, record extreme value.The fitness of more each particle, determines the optimum P of the individuality of each particle _bestfrom whole individual extreme value P _bestin determine global optimum G _best.

Step 4.5, upgrade speed and the position of each particle according to individual extreme value and global extremum.

$v_{\mathrm{id}}^{k+1}=K[\mathrm{\&omega;}{v}_{\mathrm{id}}^k+c_1{r}_1(P_{\mathrm{id}}^k-x_{\mathrm{id}}^k)+c_2{r}_2(G_{\mathrm{id}}^k-x_{\mathrm{id}}^k)---\left(8\right)$

$x_{\mathrm{id}}^{k+1}=x_{\mathrm{id}}^k+v_{\mathrm{id}}^{k+1}---\left(9\right)$

In formula (8) (9), be respectively iteration to the kth generation speed of i-th particle and position; be respectively iteration to kth+1 generation i-th particle speed and position; for iteration is to the individual extreme value of kth generation i-th particle; for the global extremum of population in front k generation; c ₁, c ₂for Studying factors, it can be avoided being absorbed in local optimum by convergence speedup; r ₁, r ₂it is the random number between [0,1]; K is constraint factor, in formula

Step 4.6, recalculate each particle fitness now according to objective function f, judge whether to upgrade individual extreme value P _bestwith global extremum G _best.If set up, then otherwise, then

Step 4.7, judge whether Search Results reaches iterations, if do not reach, jump to step 45; Otherwise stopping iteration, exports optimum solution.

The embodiment of the present invention also proposes the storage of a kind of light a few days ago based on chance constrained programming and follows the tracks of planning system, comprising:

Data capture unit, for reading the related data of photovoltaic plant and accumulator system;

Data pre-processing unit, for by photovoltaic before Monte Carlo technique simulating sun actual exert oneself and formulate photovoltaic meter mark power bound scope;

Control module, for setting up based on short-term photovoltaic predicted power value, stochastic prediction departure the chance constrained programming mathematical model containing control coefrficient;

Calculating output module, improving APSO algorithm determination accumulator system charge-discharge electric power for adopting.

Described data pre-processing unit comprises further:

First pretreatment unit, for generation of prediction deviation random value, before simulating sun, photovoltaic is actual exerts oneself;

Second pretreatment unit, for according to short-term photovoltaic predicted power, calculates fluctuation limit value, determines plan bound scope;

Described calculating output module comprises further:

Module being set, for setting the parameter of particle swarm optimization algorithm, comprising: population sum N, iterations k, inertia constant ω and Studying factors c ₁and c ₂;

Initialization module, for position and the speed of initialization population;

Fitness computing module, for determining the fitness of each particle in population;

Extreme value computing module, for the fitness of more each particle, determines the individual extreme value P of each particle _bestwith global extremum G _best;

Update module, for upgrading speed and the position of each particle according to individual extreme value and global extremum, and recalculates each particle fitness now, judges whether to upgrade individual extreme value P _bestwith global extremum G _best;

Perform output module, judge whether Search Results reaches iterations, if do not reach, continue the speed and the position that upgrade each particle; Otherwise stopping iteration, exports optimum solution.

Sample calculation analysis

With certain wind-light storage demonstration project for background, choose one day in July short-term photovoltaic predicted data as sample calculation analysis object, in this demonstration project, photovoltaic generation total installation of generating capacity is 40MW, and energy storage total installation of generating capacity is 20MW/70MWh, the charged original state of setting accumulator system is 0.5, SOC _min=0.3, SOC _max=0.9.Optimum configurations in PSO: population scale is 40, particle dimension is 96, c ₁=c ₂=1.4962, ω=0.7298, particle rapidity scope is [-3,3], and maximum iteration time gets 500.MonteCarlo number realization is set to 1000.

Fig. 2 is for this demonstration project short-term one day in July photovoltaic predicted data, according to actual physics situation, according to prediction corresponding in " specification " exert oneself ± 25% fluctuation worked out this day photovoltaic operation plan bound and to have exerted oneself scope, and to be exerted oneself by the actual photovoltaic that Monte Carlo simulation technique simulates proxima luce (prox. luc) 0 ~ 24h.

For checking herein propose validity and the dirigibility of control strategy, example carries out simulation calculation contrast respectively in fixed coefficient situation and variation factor situation.Energy storage switching coefficient u=1 is set in fixed coefficient situation, keep accumulator system in running order all the time, fixed target power control ratio c 0.5 and 0.2 carries out emulation respectively as Fig. 3, control strategy effectively can realize light and store up the target of combining force tracking plan and exerting oneself herein, actual exert oneself accumulator system supplement under be substantially all limited within the scope of plan bound, tracking effect is as shown in Fig. 4 and 5.In addition, although tracking effect when c being fixed as 0.5 is obviously excellent in the situation 0.2 time, but can find as Fig. 8 in SOC change curve, it is always in running order that c gets 0.5 accumulator system major part period within 96 periods, and it is darker when depth of discharge is compared with 0.2, last beyond SOC lower limit in example, be undesirable.And though the SOC variation tendency of accumulator system is rational substantially when c gets 0.2, still very high to the requirement of accumulator system.For reducing energy storage burden, according to charge-discharge electric power constraint improvement condition on fixed coefficient c=0.2 basis, accumulator system is made only to exert oneself lower than the period electric discharge of plan lower limit at actual photovoltaic, exceed the period charging of the plan upper limit, all the other periods keep idle condition to regulate as shown in Figure 6 switching coefficient d.As shown in Figure 7, simulation result still reaches Expected Results.

Can find out that the SOC change curve to carrying out emulating gained accumulator system after fixed coefficient c=0.2 situation variable coefficient optimizing regulation is obviously optimum in fig. 8, after variation factor, energy storage just carries out work in the small part period, all the other major part periods are all in idle condition, and this is very favorable to prolongation accumulator system serviceable life.Each discharge and recharge is all carried out within the scope of example SOC bound in addition, all can carry out a certain amount of charged/discharged, also further increased the charging and discharging capabilities of accumulator system before charge/discharge.

For simulated effect under verifying variation factor situation is further more excellent, to not adding energy storage, after adding energy storage, in fixed coefficient (c=0.2) and variation factor situation, by predicated error, the probability be limited within the scope of this specifically calculates, the each scheme of comparative analysis reduces predicated error, improves the degree of precision of prediction.Result shows, not only requires to reduce to energy storage after variation factor, more can 100% error to be reduced and to be limited in acceptability limit, as shown in table 1.

Table 1. photovoltaic power error meet the demands probability contrast

Drop into energy storage situation	Do not drop into energy storage	Fixed coefficient situation	Variation factor situation
				Probability within ± 25% error band scope	77.08％	98.96％	100％

Therefore analyze tracking plan to exert oneself effect by above comprehensive, improve the working condition of precision of prediction degree and accumulator system, variable coefficient control strategy can be adopted to control as with reference to the energy storage of scheme to the same day, and specifically each period charge-discharge electric power value as shown in Figure 9.

Evolution of Population process as shown in Figure 10, can find out, fitness value reduces gradually along with the increase of evolutionary generation, show light storage associating power curve and target control powertrace more and more close, when iterations reaches near 200 times, fitness value reaches optimum and substantially no longer changes, and describes algorithm and has good convergence.

Table 2 makes simulation comparison by getting different value to confidence level, find that confidence level is about 0.65 time, error qualification rate is also few compared to improving when not adding energy storage, explanation tracking effect is not ideal enough, get confidence level more than 0.75, along with the raising of confidence level, meet error requirements acceptable level and all substantially reach more than 90%, but the total discharge and recharge needed increases gradually, also more strict to the requirement of accumulator system, the confidence level be suitable for can be selected to calculate according to accumulator system concrete condition in actual tracing control.

The different confidence level comparison of computational results of table 2.

Force control method is planned out in the storage of the light a few days ago based on the chance constrained programming tracking that the present invention proposes, work out photovoltaic meter according to short-term forecasting power and mark power bound scope, consider prediction deviation randomness, real power before employing Monte Carlo technique simulating sun, set up chance constrained programming mathematical model, utilize improvement APSO algorithm to solve accumulator system to exert oneself, result shows that light storage associating tracking plan is exerted oneself and reaches good effect.Considering the feasibility of energy storage practical application simultaneously, adjusting tracking target power at any time by arranging control coefrficient, compared with fixed coefficient control strategy, make energy storage a few days ago exert oneself control program more flexibly, the requirement of accumulator system is reduced further.

Finally should be noted that: above embodiment is only in order to illustrate that technical scheme of the present invention is not intended to limit, although with reference to above-described embodiment to invention has been detailed description, those of ordinary skill in the field are to be understood that: still can modify to the specific embodiment of the present invention or equivalent replacement, and not departing from any amendment of spirit and scope of the invention or equivalent replacement, it all should be encompassed in the middle of right of the present invention.

Claims (12)

1. the light-preserved system based on chance constrained programming follows the tracks of a planned dispatching method a few days ago, it is characterized in that, described method comprises the steps:

(1) related data of photovoltaic plant and accumulator system is read;

(2) formulate photovoltaic meter mark power bound scope by photovoltaic actual power power before Monte Carlo technique simulating sun;

(3) set up based on short-term photovoltaic predicted power value, stochastic prediction departure the chance constrained programming mathematical model containing control coefrficient;

(4) improvement APSO algorithm determination accumulator system charge-discharge electric power is adopted.

2. dispatching method according to claim 1, it is characterized in that, in described step (1), described related data comprises: short-term photovoltaic predicted power value, accumulator system charge-discharge electric power upper lower limit value, state-of-charge upper lower limit value.

3. dispatching method according to claim 1, it is characterized in that, described step (2) comprises the steps:

Step 2-1, Monte Carlo technique simulation process is adopted to be stochastic variable ξ (t) photovoltaic prediction deviation;

Step 2-2, using described short-term photovoltaic predicted power value as determining variable;

Step 2-3, described formulation photovoltaic meter mark power bound scope.

4. dispatching method according to claim 3, is characterized in that, in described step 2-1, described photovoltaic prediction deviation be photovoltaic generation a few days ago predicted power and the same day photovoltaic actual power power difference, photovoltaic actual power power P _actt () is by formula P _act(t)=P _pret ()+ξ (t) obtains, P _pret () is the predicted power of t photovoltaic generation a few days ago, the probability distribution function of described stochastic variable ξ (t) adopts that to meet average be 0, and variance is σ ²t the Normal probability distribution of (), σ (t) is by formula σ (t)=0.2P _pre(t)+0.02C _aptry to achieve, C _apfor photovoltaic installed capacity.

5. dispatching method according to claim 3, it is characterized in that, in described step 2-3, described photovoltaic meter marks power bound scope with described short-term photovoltaic predicted power value for working out according to according to ± 25% fluctuation, and described photovoltaic meter marks power bound and calculated by following formula:

P _limit=ξ _allowc _ap

P _{plan_up}(t)＝P _pre(t)+P _limit

P _{plan_dn}(t)＝P _pre(t)-P _limit

In formula, ξ _allowfor predicated error allows percentage value, by being defined as 25% a few days ago; P _limitfor ξ _allowthe photovoltaic determined goes out fluctuation limit value; C _apfor photovoltaic installed capacity; P _{plan_up}t () marks power higher limit, P for photovoltaic meter _{plan_dn}t () marks power lower limit for photovoltaic meter.

6. dispatching method according to claim 1, is characterized in that, in described step (3), comprise the steps:

Step 3-1, to exert oneself within the scope of bound is exerted oneself in plan as target at utmost to make light store up, set up objective function as shown in the formula:

p _{plan_adj}(t)＝u{P _{plan_up}(t)+c[P _{plan_up}(t)-P _{plan_dn}(t)]}

$\min f=\sqrt{\underset{t=1}{\overset{m}{\mathrm{\&Sigma;}}}{[P_{\mathrm{bess}}\left(t\right)+P_{\mathrm{pre}}\left(t\right)+\mathrm{\&xi;}\left(t\right)-P_{\mathrm{plan}\_\mathrm{adj}}\left(t\right)]}^2}$

P in formula _{plan_adj}t () is target control power, P _{plan_up}t () marks power higher limit, P for photovoltaic meter _{plan_dn}t () marks power lower limit for photovoltaic meter, f is objective function, and u controls the switching coefficient whether energy storage work, and it is that accumulator system is in running order that u gets 1, is in idle condition when u gets 0, and c is the target power control coefrficient between 0 to 1; P _besst () is decision variable, i.e. t energy storage charge-discharge electric power, P _besst () is greater than zero, represent energy storage device electric discharge, is less than null representation charging;

Step 3-2, set up chance constrained programming condition;

Step 3-3, set up accumulator system constraint condition.

7. dispatching method according to claim 6, is characterized in that, in described step 3-2, by photovoltaic active power output smoothing rate η _eerabsolute value be less than or equal to allowed band δ as chance constrained programming condition, make the probability P that it is set up _rbe not less than confidence level α, as shown in the formula:

η _eer＝P _pre(t)+P _bess(t)+ξ(t)-P _{plan_adj}(t)/P _{plain_adj}(t)

P _r{|η _eer|≤δ}≥α

In formula, P _rfor photovoltaic active power output smoothing rate η _eerthe probability set up when being less than or equal to allowed band δ.

8. dispatching method according to claim 6, it is characterized in that, in described step 3-3, described accumulator system constraint condition comprises:

Power constraints, when namely charging, P _ch.max≤ P _bess(t)≤0, during electric discharge, 0≤P _bess(t)≤P _dis.max, in formula, P _ch.maxfor the maximum charge power that negative value is accumulator system; P _dis.maxfor on the occasion of the maximum discharge power being accumulator system;

SOC constraint condition, SOC _min≤ SOC (t)≤SOC _max, in formula, SOC _minfor the minimum value of accumulator system state-of-charge; SOC _maxfor the maximal value of accumulator system state-of-charge; SOC (t) is the state-of-charge of t; ζ is corresponding discharge and recharge coefficient, and ζ > 1 during electric discharge, illustrates in discharge process to there is certain loss, and ζ < 1 during charging, illustrates in charging process also there is certain loss; Δ t is the sampling time interval of power; C is the rated capacity of accumulator system.

9. dispatching method according to claim 8, is characterized in that, in described power constraints, work as P _act(t) > P _{plan_up}t, time (), accumulator system is in charged state, i.e. P _ch.max≤ P _bess(t)≤0; Work as P _act(t) < P _{plan_up}t, time (), accumulator system is in discharge condition, i.e. 0≤P _bess(t)≤P _dis.max.

10. dispatching method according to claim 1, is characterized in that, in described step (4), comprise the steps:

The parameter of step 4-1, setting particle swarm optimization algorithm, comprising: population sum is from iterations k, maximum iteration time k _max, inertia constant ω and Studying factors c ₁and c ₂;

The position of step 4-2, initialization population and speed;

Step 4-3, determine the fitness of each particle in population according to objective function f;

The fitness of step 4-4, more each particle, determines the optimum P of the individuality of each particle _bess, from whole individual optimum P _bessin determine global optimum G _best; Speed and the position of each particle is upgraded according to individual optimum and global optimum;

Step 4-5, again calculate each particle fitness now according to objective function f, judge whether to upgrade individual optimum P _besswith global optimum G _best;

Step 4-6, judge whether Search Results reaches iterations, if do not reach, continue the speed and the position that upgrade each particle; Otherwise stopping iteration, exports optimum solution.

11. dispatching methods according to claim 10, is characterized in that, in described step 4-4, and the speed of each particle of described renewal and position, formula is as follows:

$v_{\mathrm{id}}^{k+1}=K[{\mathrm{\&omega;v}}_{\mathrm{id}}^k+c_1{r}_1(P_{\mathrm{id}}^k-x_{\mathrm{id}}^k)+c_2{r}_2(G_{\mathrm{id}}^k-x_{\mathrm{id}}^k)$

$x_{\mathrm{id}}^{k+1}=x_{\mathrm{id}}^k+v_{\mathrm{id}}^{k+1}$

In formula, be respectively iteration to the kth generation speed of i-th particle and position; be respectively iteration to kth+1 generation i-th particle speed and position; for iteration is to the individual extreme value of kth generation i-th particle; for the global extremum of population in front k generation; c ₁, c ₂for Studying factors, it can be avoided being absorbed in local optimum by convergence speedup; r ₁, r ₂it is the random number between [0,1]; K is constraint factor, in formula wherein k is current particle iterations; k _maxfor particle cluster algorithm starts the maximum iteration time of setting; ω _min, ω _maxminimum and maximum inertia weight respectively.

12. dispatching methods according to claim 10, is characterized in that, in described step 4-5, described in judge whether to upgrade individual optimum P _bestwith global optimum G _bestmode be: if set up, then otherwise, $p_{\mathrm{id}}^k=x_{\mathrm{id}}^k,$ Then $G_{\mathrm{id}}^k=\min \left(P_{\mathrm{id}}^k\right).$