CN103295077A - Wind power plant cluster scheduling method with consideration of prediction error distribution characteristics - Google Patents

Abstract

The invention discloses a wind power plant cluster scheduling method with consideration of prediction error distribution characteristics. The wind power plant cluster scheduling method is specifically used for scheduling a wind power plant cluster. The wind power plant cluster scheduling method has the advantages that historical error probability distribution characteristics are analyzed by a statistical process on the basis of historical predicted wind power data, and an optimization objective includes minimizing the sum of values of mathematical expectation of shortage of wind power outputted by various wind power plants after active power instructions are issued by the wind power plant cluster, so that active power instructions which are issued by a system to the wind power plant cluster can be completed to the greatest extent; by the method, power difference caused by the fact that active power outputted by existing wind power plants cannot meet cluster scheduling requirements due to wind power active prediction errors can be reduced, system scheduling participation ability of the wind power plants is improved, and accordingly the system scheduling pressure is relieved.

Description

A kind of wind energy turbine set colony dispatching method of considering the predicated error distribution character

Technical field

The present invention relates to a kind of wind energy turbine set colony dispatching method, specifically is a kind of wind energy turbine set colony dispatching method of considering the predicated error distribution character, belongs to technical field of wind power generation.

Background technology

In recent years, along with the increase of wind-electricity integration capacity with insert the raising of electric pressure, large-scale wind power concentrate be incorporated into the power networks more and more big to the influence degree of power network dispatching system.For the influence that the randomness of tackling wind-powered electricity generation and undulatory property cause system's active power balance, including the wind power information of forecasting operation of in electric system scheduler routine becomes inevitable.Problems such as and existing dispatching system is too simple at the dispatching method of wind-powered electricity generation active power, causes annual wind-powered electricity generation on average to generate electricity and utilizes hourage low, and the target generated output performance is relatively poor.

At the meritorious scheduling problem of wind-powered electricity generation, Chinese scholars has been carried out useful research from different perspectives.But, no matter existing research is gain merit in the wind energy turbine set scheduling distribution or the meritorious scheduling distribution of wind energy turbine set cluster, the wind power information of forecasting that adopts all is point prediction information, because the restriction of wind energy characteristic and existing prediction level, these information of forecastings often have bigger error, bring great difficulty to dispatching of power netwoks operation, this meritorious instruction that also to be existing dispatching system send to wind energy turbine set is well below the reason of wind power predicted data.Compare to future predicting the outcome sometime provide the point prediction of determinacy numerical value, can provide the probability density prediction of the complete probability distribution of future value to become research emphasis gradually.So far also do not study wind power predicated error probability distribution information is applied to during electric system in a few days dispatches.

Summary of the invention

Technical matters to be solved by this invention is to provide a kind of wind energy turbine set colony dispatching method of considering the predicated error distribution character at wind energy turbine set colony dispatching problem.

In order to solve the problems of the technologies described above, technical scheme of the present invention is: a kind of wind energy turbine set colony dispatching method of considering the predicated error distribution character, and it may further comprise the steps:

(1) the actual measurement active power of output data set y in wind farm group x days _i, prediction active power of output data set y _i', obtain this wind farm group active power prediction error data sample set e with this _i=y _i'-y _i

(2) add up the quantity of error information in each window, obtain the discrete probability density function of error:

Window width when setting h and being the error information statistics is divided into the n equal portions that size is h with the scope of statistics of error information sample set in the step (1); If W=int is (e _i/ h), int (x) is bracket function, then error e _iThe window at place be [hW, h (W+1)); Corresponding probability is PRO (W)=m _W/ M, wherein m _WFor window [hW, h (W+1)) interior sample number, M is total sample number.

The target of this method is exactly to reduce the risk that occurs meritorious power shortage when each wind energy turbine set of mind-set issues dispatch command in the wind-powered electricity generation colony dispatching.

(3) establishing the some time, to issue the wind farm group dispatch command be P in the etching system dispatching center _{Dispatch_all}, prediction wind energy turbine set WF _nIt is P that maximum can be sent active power _{Predict_n}, be handed down to wind energy turbine set WF _nThe instruction of exerting oneself for P _{Dispatch_n}, and wind energy turbine set WF _nActual sending can be sent the maximum meritorious P that is _{Real_n}, with each wind energy turbine set wind power vacancy expectation value E in the wind energy turbine set cluster _nThe sum minimum is optimization aim, and the operation plan instruction that wind farm group is issued to wind energy turbine set is optimized, and optimization aim is:

min∑E _n

Wherein the min function is for returning minimum value function in the given parameter list;

Constraint condition is:

∑P _{dispatch_n}=P _{dispatch_all}

P _{dispatch_n}≤P _{max_n}

P wherein _{Max_n}Be wind energy turbine set WF _nMaximum output restriction;

If P _{Real_n}＜P _{Dispatch_n}, then produce (P _{Predict_n}-P _{Real_n}) power shortage.Because P _{Real_n}Unknown in advance, can obtain producing (P according to wind power predicated error probability Distribution Model _{Predict_n}-P _{Real_n}) probability of power shortage is PRO (int ((P _{Predict_n}-P _{Real_n})/h)).

Optimization problem method for solving (find the solution, use based on method for solving such as derivatives as adopting variational principle) by maturation can be obtained wind energy turbine set WF _nThe instruction of exerting oneself for P _{Dispatch_n}, wherein, wind energy turbine set wind power vacancy expectation value E _nDeriving method be:

1) asks for predicted value place window W _{0_n}:

W _{0_n}=int(P _{dispatch_n}/h)

2) ask at W _{0_n}Interior wind energy turbine set wind power vacancy expectation value E _{0_n}:

${\mathrm{<figure-callout\; id="0"\; label="E"\; filenames="HDA00003203660100021.png,HDA00003203660100031.png"\; state="\{\{state\}\}">E</figure-callout>}}_{0\_n}={\&Integral;}_{h(W_{0\_n}-1)}^{P_{\mathrm{dispatch}\_n}}\frac{\mathrm{PRO}(\mathrm{int}(P_{\mathrm{dispatch}\_n}-P)/h)}{h}*(P_{\mathrm{dispatch}\_n}-P)d_P$

3) ask for WF _nWind energy turbine set wind power vacancy expectation value E _n:

$E_n={\mathrm{<figure-callout\; id="0"\; label="E"\; filenames="HDA00003203660100021.png,HDA00003203660100031.png"\; state="\{\{state\}\}">E</figure-callout>}}_{0\_n}+\underset{a=0}{\overset{{\mathrm{<figure-callout\; id="0"\; label="W"\; filenames="HDA00003203660100021.png,HDA00003203660100031.png"\; state="\{\{state\}\}">W</figure-callout>}}_{0\_n}-1}{\mathrm{\&Sigma;}}}{\&Integral;}_{\mathrm{ha}}^{h(a+1)}\frac{\mathrm{PRO}\left(a\right)}{h}*(P_{\mathrm{dispatch}\_n}-P)d_P.$

In general, wind power forecasting research is adopted a certain distribution character of given wind speed through regular meeting, ask for the statistical method of its distribution parameter according to historical data; Yet for the wind power predicated error, existing research can't provide its distribution character accurately.Therefore, for objectively responding the regularity of wind power predicated error historical data, the present invention adopts the method for direct statistics, obtains the discrete distributed model on himself statistical law meaning.

The present invention is directed to wind energy turbine set colony dispatching problem, propose a kind of active power dispatching method of considering wind power predicated error distribution character.This method is based on wind power prediction history data, by statistical method analysis of history probability of error distribution character, the mathematical expectation sum minimum that issues the wind power vacancy of each wind energy turbine set output of meritorious instruction back with the wind energy turbine set cluster is optimization aim, and the active power instruction that the system that makes is handed down to the wind energy turbine set cluster can be finished as far as possible.Can reduce because the meritorious predicated error of wind-powered electricity generation causes the wind energy turbine set active power of output can not satisfy the power difference that the colony dispatching requirement produces by this method, increase the ability of wind-powered electricity generation participation system scheduling, thus mitigation system scheduling pressure.

Description of drawings

The present invention is further detailed explanation below in conjunction with the drawings and specific embodiments.

Fig. 1 considers the wind energy turbine set colony dispatching method method flow diagram of predicated error distribution character for the present invention.

Fig. 2 is wind energy turbine set wind power predicted data probability of error density in the wind energy turbine set cluster in the specific embodiment.

The instruction P that exerts oneself that Fig. 3 receives for wind energy turbine set WF1 among the embodiment _{Dispatch_1}Curve map;

The instruction P that exerts oneself that Fig. 4 receives for wind energy turbine set WF2 among the embodiment _{Dispatch_2}Curve map;

The instruction P that exerts oneself that Fig. 5 receives for wind energy turbine set WF3 among the embodiment _{Dispatch_3}Curve map;

The instruction P that exerts oneself that Fig. 6 receives for wind energy turbine set WF4 among the embodiment _{Dispatch_4}Curve map;

Embodiment

The inventive method example adopts data from the actual wind energy turbine set cluster of Gansu province.Contain four wind energy turbine set in this wind energy turbine set cluster.Wherein WF1 and WF3 capacity are 200MW, and WF2 and WF4 capacity are 150MW.

Flow process of the present invention as shown in Figure 1, the actual measurement active power of output data set y in the wind farm group 300 days _i, prediction active power of output data set y _i', obtain this wind farm group active power prediction error data sample set e with this _i=y _i'-y _i

Window width when setting h and be historical error statistics when namely carrying out error statistics is divided into scope of statistics the n equal portions that size is h.Window width is more little, and the precision that the probability of error distributes is more high, and probability Distribution Model is also just more complicated, and is consuming time also just more many with this follow-up dispatching algorithm of carrying out, and therefore should select suitable window width on the basis of taking all factors into consideration accuracy requirement and algorithm computing time.In conjunction with this example, getting h is 1MW.If W=int is (e _i/ h), int (x) is bracket function, then error e _iThe window at place be [hW, h (W+1)).Corresponding probability is PRO (W)=m _W/ M, wherein m _WFor window [hW, h (W+1)) interior sample number, M is total sample number.

Carry out sample analysis by four wind energy turbine set historical data sample of said method, obtain the probability Distribution Model of wind power predicated error, as shown in Figure 2.It is as shown in table 1 to obtain four wind energy turbine set predicated error statisticss simultaneously.

Table 1 wind energy turbine set wind power predicated error historical data statistics

Consider the wind power forecast confidence, set under the fiducial interval and be limited to the operation plan that system is issued to the wind energy turbine set cluster, concrete grammar is: at certain wind energy turbine set cluster, the prediction of interior m Fixed Time Interval of collection a period of time can send out the active power data and reality can be sent out the active power data and form data set P respectively _{Predict_all}And P _{Real_all}So, (P _{Predict_all}-P _{Real_all}) be wind energy turbine set cluster active power error collection.If it is m that error is concentrated the number of positive error (comprising 0) _Positive, the number of negative error is m _NegativeThe error that error is concentrated sorts.When getting degree of confidence when being C, round-off error ascending

Individual data value is m _u, get negative error descending

Individual data value is m _l, wherein

For rounding symbol downwards.So, corresponding to predicted value P _PredictThe fiducial interval lower limit be (P _Predict+ m _l), m wherein _lFor negative.

Getting m in this example is that 86400, C is 90% o'clock, and establishing the some time, to issue the wind farm group dispatch command be P in the etching system dispatching center _{Dispatch_all}, prediction wind energy turbine set WF _nIt is P that maximum can be sent active power _{Predict_n}, be handed down to wind energy turbine set WF _nThe instruction of exerting oneself for P _{Dispatch_n}, and wind energy turbine set WF _nActual sending can be sent the maximum meritorious P that is _{Real_n}, with each wind energy turbine set wind power vacancy expectation value E in the wind energy turbine set cluster _nThe sum minimum is optimization aim, and the operation plan instruction that wind farm group is issued to wind energy turbine set is optimized, and optimization aim is:

min∑E _n

Wherein the min function is for returning minimum value function in the given parameter list;

Constraint condition is:

∑P _{dispatch_n}=P _{dispatch_all}

P _{dispatch_n}≤P _{max_n}

P wherein _{Max_n}Be wind energy turbine set WF _nMaximum output restriction;

Optimization problem method for solving by maturation can be obtained wind energy turbine set WF _nThe instruction of exerting oneself for P _{Dispatch_n}, wherein, wind energy turbine set wind power vacancy expectation value E _nDeriving method be:

1) asks for predicted value place window W _{0_n}:

W _{0_n}=int(P _{dispatch_n}/h)

2) ask at W _{0_n}Internal power vacancy expectation value E _{0_n}:

${\mathrm{<figure-callout\; id="0"\; label="E"\; filenames="HDA00003203660100021.png,HDA00003203660100031.png"\; state="\{\{state\}\}">E</figure-callout>}}_{0\_n}={\&Integral;}_{h(W_{0\_n}-1)}^{P_{\mathrm{dispatch}\_n}}\frac{\mathrm{PRO}(\mathrm{int}(P_{\mathrm{dispatch}\_n}-P)/h)}{h}*(P_{\mathrm{dispatch}\_n}-P)d_P$

3) ask for WF _nGeneral power vacancy expectation value E _n:

$E_n={\mathrm{<figure-callout\; id="0"\; label="E"\; filenames="HDA00003203660100021.png,HDA00003203660100031.png"\; state="\{\{state\}\}">E</figure-callout>}}_{0\_n}+\underset{a=0}{\overset{{\mathrm{<figure-callout\; id="0"\; label="W"\; filenames="HDA00003203660100021.png,HDA00003203660100031.png"\; state="\{\{state\}\}">W</figure-callout>}}_{0\_n}-1}{\mathrm{\&Sigma;}}}{\&Integral;}_{\mathrm{ha}}^{h(a+1)}\frac{\mathrm{PRO}\left(a\right)}{h}*(P_{\mathrm{dispatch}\_n}-P)d_P.$

Comprehensively above-mentioned, the instruction of exerting oneself of each wind energy turbine set that obtains is as Fig. 3-shown in Figure 6.

The influence that this method produces scheduling decision when adopting different wind power forecast confidence for the research operation plan, compare the scheduling precision that adopts this algorithm and control with changed scale apportion design under 4 kinds of cluster wind-powered electricity generation predicted power degree of confidence, be respectively 99%, 90%, 60% and 15%.

For analyzing this algorithm to the raising of scheduling precision, the power shortage that the wind energy turbine set cluster produces when calculate adopting the dispatch command that every kind of method sends, computing method are:

Wherein, P _{Dispatch_all}For the wind energy turbine set cluster always issues dispatch command, P _ΣBe the total actual active power of sending of wind energy turbine set cluster, η is the vacancy ratio.Calculate wind-powered electricity generation utilization factor under different degree of confidence simultaneously, computing method are:

Wherein, P _{Real_ Σ}For wind energy turbine set cluster reality can be sent out peak power.

When cluster wind-powered electricity generation predicted power degree of confidence got 99%, 90%, 60% and 20% respectively, the dispatch command vacancy that control with changed scale allocation algorithm and this paper method obtain was as shown in table 2.

The different operation plan this method of table 2 and control with changed scale distribution method produce the vacancy situation

As can be seen,

(1) contrast control with changed scale distribution method when cluster wind-powered electricity generation predicted power degree of confidence gets 90%, 60% and 15% respectively, adopts this method to distribute the instruction of wind energy turbine set colony dispatching can improve the wind-powered electricity generation utilization factor to a certain extent.

(2) contrast control with changed scale distribution method, on the basis that the wind-powered electricity generation utilization factor improves, when employing this paper method was distributed the instruction of wind energy turbine set colony dispatching, the wind energy turbine set cluster was because the power shortage that can not satisfy the scheduling requirement and produce of exerting oneself significantly reduces.

(3) contrast control with changed scale distribution method along with the increase of wind energy turbine set operation plan, reduces vacancy and accounts for the wind energy turbine set cluster and send the active power ratio and be respectively 0%, 0.2%, 0.47% and 0.74%.Therefore adopting the wind-powered electricity generation degree of confidence in dispatching of power netwoks, more low can to hold the wind-powered electricity generation ratio more high, and the gain effect that this method is brought is more obvious.

Be that 60% o'clock different wind power predicated errors of scheduling situation analysis distribute to the influence of scheduling precision to adopt cluster wind-powered electricity generation predicted power degree of confidence.The wind power predicted data error of four wind energy turbine set distributes as shown in Figure 2, and concrete error parameter sees Table 1.As shown in table 3 by the dispatch command vacancy that control with changed scale allocation algorithm and the inventive method obtain.

The wind energy turbine set that the different predicated errors of table 3 distribute produces the vacancy situation

As can be seen,

(1) for each wind energy turbine set of cluster inside, when adopting this paper method to distribute the wind energy turbine set colony dispatching to instruct, each wind energy turbine set is dispatched all significantly minimizings of power shortage that require and produce because exert oneself to satisfy.

(2) wind energy turbine set of the equal capacity of contrast, WF1 and WF3, this paper method is 0.51% for the ratio that the bigger wind energy turbine set WF1 of error sample variance reduces vacancy, and is 0.25% for the ratio that the less wind energy turbine set WF3 of error sample variance reduces vacancy; Equally, WF2 and WF4, this paper method is 0.99% for the ratio that the bigger wind energy turbine set WF4 of error sample variance reduces vacancy, and is 0.07% for the ratio that the less wind energy turbine set WF2 of error sample variance reduces vacancy.So it is obvious that this paper method improves effect to the big wind energy turbine set of error sample variance.

Contrast by above and control with changed scale allocation algorithm as can be seen, the wind energy turbine set colony dispatching method of consideration predicated error distribution character can be optimized the scheduling of wind energy turbine set cluster active power and distribute the remarkable power shortage that produces less than dispatching requirement because of the actual active power of wind energy turbine set that reduces; Simultaneously, the gain effect that brings of this method increasing and increase with wind power in the dispatching of power netwoks; For the wind energy turbine set of equal capacity, this method is more obvious to the raising effect of the big wind energy turbine set of wind power predicated error sample variance.

Above-described embodiment does not limit the present invention in any way, and every employing is equal to replaces or technical scheme that the mode of equivalent transformation obtains all drops in protection scope of the present invention.

Claims (1)

1. wind energy turbine set colony dispatching method of considering the predicated error distribution character is characterized in that may further comprise the steps:

(1) the actual measurement active power of output data set y in wind farm group x days _i, prediction active power of output data set y _i', obtain this wind farm group active power prediction error data sample set e with this _i=y _i'-y _i

(2) add up the quantity of error information in each window, obtain the discrete probability density function of error:

Window width when setting h and being the error information statistics is divided into the n equal portions that size is h with the scope of statistics of error information sample set in the step (1); If W=int is (e _i/ h), int (x) is bracket function, then error e _iThe window at place be [hW, h (W+1)); Corresponding probability is PRO (W)=m _W/ M, wherein m _WFor window [hW, h (W+1)) interior sample number, M is total sample number;

(3) establishing the some time, to issue the wind farm group dispatch command be P in the etching system dispatching center _{Dispatch_all}, prediction wind energy turbine set WF _nIt is P that maximum can be sent active power _{Predict_n}, be handed down to wind energy turbine set WF _nThe instruction of exerting oneself for P _{Dispatch_n}, and wind energy turbine set WF _nActual sending can be sent the maximum meritorious P that is _{Real_n}, with each wind energy turbine set wind power vacancy expectation value E in the wind energy turbine set cluster _nThe sum minimum is optimization aim, and the operation plan instruction that wind farm group is issued to wind energy turbine set is optimized, and optimization aim is:

min∑E _n

Wherein the min function is for returning minimum value function in the given parameter list;

Constraint condition is:

∑P _{dispatch_n}=P _{dispatch_all}

P _{dispatch_n}≤P _{max_n}

P wherein _{Max_n}Be wind energy turbine set WF _nMaximum output restriction;

Optimization problem method for solving by maturation can be obtained wind energy turbine set WF _nThe instruction of exerting oneself for P _{Dispatch_n}, wherein, wind energy turbine set wind power vacancy expectation value E _nDeriving method be:

1) asks for predicted value place window W _{0_n}:

W _{0_n}=int(P _{dispatch_n}/h)

2) ask at W _{0_n}Interior wind energy turbine set wind power vacancy expectation value E _{0_n}:

$E_{0\_n}={\&Integral;}_{h(W_{0\_n}-1)}^{P_{\mathrm{dispatch}\_n}}\frac{\mathrm{PRO}(\mathrm{int}(P_{\mathrm{dispatch}\_n}-P)/h)}{h}*(P_{\mathrm{dispatch}\_n}-P)d_P$

3) ask for WF _nTotal wind energy turbine set wind power vacancy expectation value E _n:

$E_n=E_{0\_n}+\underset{a=0}{\overset{W_{0\_n}-1}{\mathrm{\&Sigma;}}}{\&Integral;}_{\mathrm{ha}}^{h(a+1)}\frac{\mathrm{PRO}\left(a\right)}{h}*(P_{\mathrm{dispatch}\_n}-P)d_P.$

Images