CN105844343A - Water-optical daily joint scheduling method based on analysis of photovoltaic properties - Google Patents

Abstract

The invention discloses a water-optical daily joint scheduling method based on analysis of photovoltaic properties. The method comprises a first step of according to the form of historical daily output sequences of a photovoltaic power station, classifying the historical daily output sequences into M types of typical sequences, and obtaining a daily output envelope of each typical sequence; a second step of calculating a daily output clearness index R, an output residual standard deviation D2, and an output residual Gini coefficient NG of each historical daily output sequence; a third step of clustering the historical daily output sequence; a fourth step of constructing an output residual model; a fifth step of combining classified output scenes of the photovoltaic power station and inflow situations of a hydropower station, generating a predicted daily joint scheduling system scene; and a sixth step of taking the water-optical joint scheduling scene as the background, solving the output residual model, and obtaining an optimal operation scheme for the predicted daily water-optical joint scheduling. According to the invention, the constructed model is clear in physical mechanism, and the method is simple and reliable, easy to use and operate, and capable of use for guiding the actual scheduling of a power grid.

Description

A kind of water light day combined scheduling method analyzed based on photovoltaic property

Technical field

The present invention relates to photovoltaic plant power producing characteristics and water light day combined dispatching technical field, a kind of based on light The water light day combined scheduling method of volt specificity analysis.

Background technology

Photovoltaic plant is exerted oneself and is had the feature of uniqueness: (1) does not exerts oneself night；(2) affected by solar radiation, photovoltaic plant The process of exerting oneself is unimodal arc, and high noon, theory was exerted oneself maximum；(3) affected by cloud layer, temperature, exert oneself and there is random fluctuation Property.

When large-scale photovoltaic accesses electrical network, farthest receive photovoltaic electricity and capacity for electrical network and ensure that electrical network is pacified Complete stable, need other power supplys that photovoltaic plant is compensated, make photovoltaic plant and other cooperation station outputs With network load process compatible after superposition.When photovoltaic plant exerts oneself reduction, need conventional power plant increase to exert oneself, work as light Overhead utility is exerted oneself when increasing, and needs conventional power plant to reduce and exerts oneself.And the turbine-generator units start and stop in power station are rapid, regulation and control Nimble, unit output variable amplitude is relatively big, and therefore water power is to compensate the good energy that photovoltaic is exerted oneself.Regimen is carried out in power station Condition and the impact of storage capacity, water power short-term electricity generation is somewhat limited, power station exert oneself simultaneously by installed capacity, Storage capacity size and the restriction carrying out water state.Therefore, water, the target of light cooperation are exactly at balance photovoltaic plant load ripple System optimal is reached in the case of Dong.

At present, forecast model of exerting oneself photovoltaic plant research is more, but photovoltaic plant and routine electricity under research different mode Cooperation of standing dispatches the most less of rule.

Summary of the invention

The deficiency existed for prior art, the invention provides a kind of water light day combined dispatching analyzed based on photovoltaic property Method.

The present invention analyzes photovoltaic property by building the photovoltaic plant that is made up of three kinds of indexs appraisement system of exerting oneself, and combines Exerting oneself feature in power station, builds the compensation optimizing model that compensation of hydropower photovoltaic plant is exerted oneself, thus realize water light day and combine tune Degree.

For solving above-mentioned technical problem, the present invention adopts the following technical scheme that:

One, a kind of water light day combined scheduling method analyzed based on photovoltaic property, including:

Step 1, according to the form of the history daily output sequence of photovoltaic plant, is divided into M special dictionary by history daily output sequence Type sequence, obtains the outer envelope curve of daily output of each type sequence；

Step 2, calculates the daily output clearness index R of each history daily output sequence, residual error of exerting oneself standard deviation D₂Residual with exerting oneself Difference Gini coefficient NG, daily output clearness indexThe residual error definition of exerting oneself in history daily output sequence each moment ForResidual error of exerting oneself standard deviation D₂With residual error Gini coefficient NG of exerting oneself according to residual error of exerting oneselfMeter Calculate；P_{Pv, d, t}The EIAJ of type sequence t belonging to history daily output sequence, according to outside the daily output of type sequence Envelope curve obtains；P_pv,tFor exerting oneself of history daily output sequence t；T represents the duration of exerting oneself of history daily output sequence； Residual error of exerting oneself for history daily output sequence t；

Described residual error Gini coefficient NG of exerting oneself is adopted and is obtained with the following method:

A () is by Beijing time coordinateBe converted to relative time coordinateFor history daily output sequence The residual error of exerting oneself of row t, ta=t-tr,Tr is the initial time of exerting oneself of history daily output sequence；

B () is to relative time coordinateZoom in and out to obtain scaling coordinateScaling coefficientT_dExert oneself duration for current history daily output sequence, T_d,stFor all history daily output sequences T_dMeansigma methods；

C () is to scaling coordinateCarry out interpolation and must scale standard coordinateSt.t represents scaling Moment after process, st.t=1,2 ..., T_d,st,

(d) standardization scaling standard coordinateIn exert oneself residual error

E () normalized is exerted oneself residual errorGini coefficient NG；

Step 3, the cluster of history daily output sequence, this step farther includes:

The D of 3.1 standardization each history daily output sequence₂And NG；

3.2 with D₂It is that history daily output sequence is once clustered by Cluster Assessment index with R；

3.3 manual analysis one time cluster results, with NG and D₂For Cluster Assessment index, be there is repetition in R value scope History daily output sequence class carries out secondary cluster respectively；If there is not R value scope to repeat, terminate.

Above-mentioned M is not more than 4.

The D of above-mentioned standardization each history daily output sequence₂And NG, use equation below to carry out:

$ST\·X_i=w_i\frac{X_i-\underset{i=1,\mathrm{2...}N}{\min }\left\{X_i\right\}}{\underset{i=1,\mathrm{2...}N}{\max }\left\{X_i\right\}-\underset{i=1,\mathrm{2...}N}{\min }\left\{X_i\right\}}$

Wherein, X_iFor i-th sample value, for D₂Or NG；ST·X_iFor the i-th sample value after standardization；N is Sample number, i.e. history daily output sequence quantity；w_iFor the residual error accumulation proportionality coefficient of exerting oneself of i-th history daily output sequence, Residual error of exerting oneself according to history daily output sequence each moment obtains.

Two, a kind of water light day combined scheduling method analyzed based on photovoltaic property, including:

Step 1, it is thus achieved that R, D that claim 1 gained each history daily output sequence class is corresponding₂, the scope of NG, examine Consider solar irradiation Changing Pattern and solar irradiation under each weather conditions photovoltaic plant to be exerted oneself affecting laws, according to each history day Sequence of exerting oneself apoplexy due to endogenous wind history daily output series modality, manually determines the weather pattern that each history daily output sequence class is corresponding；

Step 2, withBuilding, for object function, Remanent Model of exerting oneself, SUM ε represents compensation Remaining difference quadratic sum；ε (t) be moment t compensation more than poor, Actual for power station moment t is exerted oneself,P₀Representing force constant, water intaking power station is minimum Non-zero history is exerted oneself；Tr and td represents sunrise and sunset moment respectively, and T represents segment length when day is total；

Step 3, sorts out field of force scape and power station water situation in conjunction with photovoltaic plant, generates the combined dispatching system of prediction day System scene, this step farther includes:

3.1 obtain the weather pattern of prediction day according to weather forecast, obtain the history day predicting that day is corresponding according to weather pattern Sequence of exerting oneself class and R, D₂Scope with NG, it was predicted that day, corresponding history daily output sequence class was abbreviated as current class；

3.2 randomly choose a series of R value in the range of the R of current class, it is assumed that D₂Being zero, photovoltaic plant presses RP_pv,d,tGo out Power, scene of this being exerted oneself is exerted oneself scene as photovoltaic standard；P_pv,d,tEIAJ for current class t；

3.3 use sub-step (a)～(d) described method history each to current apoplexy due to endogenous wind daily output in claim 1 step 2 The residual error of exerting oneself in sequence each momentProcess, obtain the standardization in each moment and exert oneself residual error

3.4 in the range of the NG value of current class stochastic generation one NG value, obtain according to current apoplexy due to endogenous wind each history daily output sequence Obtain the residual error process of exerting oneself corresponding for current NG of stochastic generationI.e. standardization is exerted oneself residual errorChange with the moment Change process；

3.5 use the inverse process of sub-step (a)～(d) described method in claim 1 step 2 to process standardization exerts oneself Residual errorWillBe converted to the residual error process of exerting oneself of Beijing time coordinateSolve residual error process of exerting oneself Standard deviation D'₂；

3.6 at the D of current class₂Stochastic generation one D in the range of value₂, calculate current D₂Corresponding residual error process of exerting oneselfObtain photovoltaic plant to exert oneself scene

3.7: if P_pv,t≤P_pv,d,tAnd P_pv,t>=0, record current photovoltaic plant and exert oneself scene, perform step 3.8；Otherwise, The cycle-index of recording step 3.6, if cycle-index reaches preset times, performs step 3.4, if cycle-index is not up to Preset times, circulation performs step 3.6；

3.8 exert oneself from the photovoltaic plant of record according to weather forecast artificially selects the photovoltaic meeting current weather type scene Output of power station scene, i.e. predicts that the photovoltaic plant of day is exerted oneself scene；

3.9 must be predicted a day power station water situation, integrated water power station water situation and prediction day by River Basin Hydrology Prediction version Photovoltaic plant exert oneself scene generate water light combined dispatching scene；

Step 4, with water light combined dispatching scene as background, solves Remanent Model of exerting oneself, it is thus achieved that the water light associating of prediction day Scheduling optimized operation scheme.

Above-mentioned weather conditions include sunny type weather, cloudy type weather, sleet type weather and mixed type weather.

Above-mentioned steps 4 farther includes:

4.1 determine k_s0、k_s1、k_s2Span [k_min,k_max], k_min=1,P_zjFill for power station Machine is exerted oneself, P₀For going out force constant；

4.2 span [k_min,k_max] upper discrete k_s0、k_s1、k_s2, and obtain all combinations of discrete point；

4.3 calculate the actual P that exerts oneself in power station of prediction day day part by the period under each discrete point combines_sdz,t；

4.4 according to the actual P that exerts oneself in power station_sdz,tCalculate difference quadratic sum SUM ε more than the compensation under the combination of each discrete point；

4.5 exert oneself with photovoltaic with the day part power station under discrete point combination corresponding for minimum compensation remaining difference quadratic sum SUM ε Scheme is as water light combined dispatching optimized operation scheme.

In sub-step 4.5, if occurring, minimum compensates the situation of the remaining corresponding many prescriptions case of difference quadratic sum SUM ε, calculates each group Exert oneself in the lower power station of discrete point combination corresponding to scheme, with maximum power station exert oneself the discrete point of correspondence combine under each time Section power station and photovoltaic are exerted oneself scheme i.e. water light combined dispatching optimized operation scheme.

Compared to the prior art, the present invention has a characteristic that

Constructed model Physical Mechanism is clear and definite, and method is the most reliable, it is simple to uses and operates, and can be used for instructing electrical network real Border is dispatched.

Accompanying drawing explanation

Fig. 1 is the idiographic flow schematic diagram of the inventive method；

Fig. 2 is cluster result schematic diagram；

Fig. 3 is a kind of concrete application flow schematic diagram of the inventive method.

Specific embodiment

Below by example, and combine accompanying drawing, technical scheme is further elaborated with.

Step 1, obtains a day EIAJ process, i.e. daily output outsourcing line according to the history daily output sequence of photovoltaic plant.

This step is existing technological means in prior art, for ease of understanding, being embodied as of this step is provided below Process:

Step 1.1: history daily output sequence is divided into M quasi-representative sequence according to the form of history daily output sequence.

The form of history daily output sequence reflects the photovoltaic plant of each day and exerts oneself process, and photovoltaic plant exerted oneself process by the same day The impact of solar irradiation Changing Pattern, corresponding different photovoltaic plant is exerted oneself process by the most different weather conditions.The most on the same day The form of the history daily output sequence under vaporous condition is different, the form phase of the history daily output sequence under identical weather conditions Seemingly.Plesiomorphic history daily output sequence is classified as a quasi-representative sequence, and each type sequence is by vaporous for skies different for correspondence Condition.

Step 1.2: proceed as follows respectively for all kinds of type sequences:

Analyze the history daily output sequence corresponding to current type sequence, select the EIAJ in each moment, EIAJ with The day EIAJ process of the most current type sequence of change procedure in moment, i.e. daily output outsourcing line.

Step 2, builds and describes the appraisement system that photovoltaic plant is exerted oneself, and uses appraisement system to evaluate each history daily output sequence Row.

Appraisement system includes three kinds of indexs: daily output clearness index R, residual error of exerting oneself standard deviation D₂With residual error Geordie of exerting oneself Coefficient NG.Using three kinds of indexs in appraisement system to be evaluated each history daily output sequence respectively, these three index is fixed Justice is as follows:

(I) daily output clearness index R.

Daily output clearness index R is for characterizing the actual influence degree exerted oneself by weather conditions of photovoltaic plant, and its definition is such as Under:

$R=\frac{\underset{t=tr}{\overset{td}{\Σ}}{P}_{pv,t}}{\underset{t=tr}{\overset{td}{\Σ}}{P}_{pv,d,t}}---\left(1\right)$

In formula (1):

P_{Pv, d, t}Represent the EIAJ of photovoltaic plant t, type sequence t belonging to i.e. current history daily output sequence EIAJ, can be obtained by the outer envelope curve of daily output, unit: MW；

P_pv,tRepresent that the actual of photovoltaic plant t is exerted oneself, exerting oneself of i.e. current history daily output sequence t, unit: MW；

Tr, td are respectively exert oneself initial time and the end time of exerting oneself of current history daily output sequence.

Daily output clearness index R characterizes the influence degree that photovoltaic plant is exerted oneself by weather, the factors such as R is the biggest, weather Affecting the least, photovoltaic plant is exerted oneself just closer to photovoltaic plant EIAJ, then the actual total amount of exerting oneself of photovoltaic plant is the biggest.

(II) residual error of exerting oneself standard deviation D₂。

The residual error of exerting oneself of history daily output sequence tFor:

${\hat{P}}_{pv,t}={\mathrm{RP}}_{pv,d,t}-P_{pv,t}---\left(2\right)$

In formula (2), residual error of exerting oneselfI.e. consider actual the exerting oneself of current history daily output sequence t under weather conditions With its belonging to the residual error item of type sequence EIAJ, unit: MW；R is the daily output of current history daily output sequence Clearness index.

Residual error of exerting oneself standard deviation D₂Being used for characterizing the average degree that actual deviation standard of exerting oneself is exerted oneself, it is defined as follows:

$D_2=\sqrt{\frac{\underset{t=tr}{\overset{td}{\Σ}}{({\hat{P}}_{pv,t}-\stackrel{\‾}{{\hat{P}}_{v,t}})}^2}{td-tr-1}}---\left(3\right)$

In formula (3):

Residual error of exerting oneself for current history daily output sequence t；

Represent the residual error average of exerting oneself of current history daily output sequence, be taken as zero here；

Tr, td are respectively exert oneself initial time and the end time of exerting oneself of current history daily output sequence.

(III) residual error of exerting oneself Gini coefficient NG.

Due in history daily output sequence each day exert oneself initial time and end time of exerting oneself inconsistent, in this situation Under, directly use certain parameter to go description residual error shape of exerting oneself to be irrational.For describing residual error shape of exerting oneself more accurately, The present invention proposes a kind of Gini coefficient computational methods based on residual error scaling pretreatment of exerting oneself, and concrete grammar is as follows:

A Beijing time coordinate-system is converted to relative time coordinate-system by ().Coordinate former for one dayWill It is converted to relative time coordinateWherein, abscissa ta=t-tr, vertical coordinate Ta=1,2 ..., T_d, T_dDuration of exerting oneself for current history daily output sequence.

B () makes zoom factorT_d,stExerting oneself duration for standard, the most all history daily output sequences are exerted oneself duration T_dMeansigma methods；Use zoom factor α to relative time coordinateZoom in and out, obtain scaling coordinate

C () is according to scaling coordinateUtilize interpolation method must scale standard coordinate, be designated as St.t represent scaling process after moment, st.t=1,2 ..., T_d,st。

D () is exerted oneself residual error standardization:

Normalization scaling standard coordinateIn exert oneself residual errorSee formula (4):

$ST.{\hat{P}}_{pv,st,t}=\frac{{\hat{P}}_{pv,st,t}-{\hat{P}}_{pv,min}}{{\hat{P}}_{pv,\max }-{\hat{P}}_{pv,min}}---\left(4\right)$

In formula (4):

For residual error of exerting oneselfStandardized value；

Being respectively history daily output sequence each moment exerts oneself the maximum of residual error and minima.

Residual error of exerting oneself after (e) normalizedGini coefficient NG:

$NG=\underset{st.t=1}{\overset{T_{d,st}}{\Σ}}(w_{st.t}\·{\mathrm{\θ}}_{st.t})---\left(5\right)$

In formula (5):

w_st.tAnd θ_st.tIt is respectively accumulation proportionality coefficient and the accumulated time proportionality coefficient of the residual error of exerting oneself after standardization, wherein,

Step 3, clusters history daily output sequence based on index in appraisement system.

Step 3.1: the D to history daily output sequence₂It is standardized with NG.

The daily output clearness index R of each history daily output sequence, residual error of exerting oneself standard deviation D is can get by step 2₂With Residual error of exerting oneself Gini coefficient NG, the D to each history daily output sequence₂It is standardized with NG.Assume total N number of sample This, then i-th sample value X of variable X (NG or NG)_iStandardization formula as follows:

$ST\·X_i=w_i\frac{X_i-\underset{i=1,\mathrm{2...}N}{\min }\left\{X_i\right\}}{\underset{i=1,\mathrm{2...}N}{\max }\left\{X_i\right\}-\underset{i=1,\mathrm{2...}N}{\min }\left\{X_i\right\}}---\left(6\right)$

In formula (6):

ST·X_iFor the i-th sample value after standardization；

w_iResidual error accumulation proportionality coefficient of exerting oneself for i-th history daily output sequence.

Step 3.2: select R and D₂For Cluster Assessment index, history daily output sequence is once clustered.

The present embodiment uses hierarchical clustering method to cluster, and step is as follows:

Step 3.2.1: initialize, each sample standard deviation is classified as a class；

Step 3.2.2: calculate the distance between each two class, i.e. similarity between sample according to Cluster Assessment index；

Step 3.2.3: two closest classes are classified as a class；

Step 3.2.4: perform step 3.2.2～3.2.3, until completing cluster, terminates.

Step 3.3: cluster result of manual analysis, with NG and D₂For Cluster Assessment index, R value scope is existed The history daily output sequence class repeated carries out secondary cluster respectively.If there is not R value scope to repeat, directly perform step 3.4。

Fig. 2 is a cluster result, and wherein the R value scope of class 1 and 3 correspondence repeats, and i.e. needs to class 1 and 3 respectively Carry out secondary cluster.

Step 3.4: R, D that the most each history daily output sequence class is corresponding₂Scope with NG.

Step 3.5: consider that photovoltaic plant is exerted oneself shadow by the Changing Pattern of solar irradiation under different weather situation and solar irradiation Ring rule, analyze the weather pattern meeting each history daily output sequence class.In the present invention, according to weather conditions by each history Daily output sequence class is classified as the weather patterns such as sunny type weather, cloudy type weather, sleet type weather, mixed type weather.This Sub-step manually performs.

Step 4, sets up power station and the exert oneself Remanent Model of photovoltaic plant is fully compensated.

Obtain the weather pattern of forecast day according to weather forecast, obtain corresponding the going through of forecast day according to the cluster result of step 3 History daily output sequence class and R and D₂Scope, photovoltaic plant is exerted oneselfRP_pv,d,tFor definitiveness Part, it is believed that controlled composition；For the random partial affected by factors such as sky temperature, its average is 0, and can With according to R, D₂, NG generates corresponding photovoltaic plant and exerts oneself scene.

Setting up the power station for analyzing power station compensation ability and the exert oneself Remanent Model of photovoltaic plant is fully compensated, this is exerted oneself The object function of Remanent Model is minimum for compensating remaining difference quadratic sum SUM ε:

$\min SUM\mathrm{\ϵ}=\underset{t=1}{\overset{T}{\Σ}}{\left(\mathrm{\ϵ}\right(t\left)\right)}^2---\left(7\right)$

In formula (7), ε (t) be moment t compensation more than poor, represent the part that cannot supplement completely of power station；SUM ε is Compensating remaining difference quadratic sum, T is to predict segment length when day is total.

Compensating remaining difference ε (t) uses formula (8) to obtain:

$\mathrm{\&epsiv;}\left(t\right)=\left\{\begin{array}{cc}{P}_{sdz,t}-{\hat{P}}_{pv,t}-P_{s,t}& P_{sdz,t}<{\hat{P}}_{pv,t}+P_{s,t}\\ 0& P_{sdz,t}={\hat{P}}_{pv,t}+P_{s,t}\end{array}\right.---\left(8\right)$

In formula (8), P_sdz,tActual for power station moment t is exerted oneself, P_s,tCompensate photovoltaic for power station during moment t to exert oneself The anticipation output of generalization after residual error, its form is as follows:

$P_{s,t}=\left\{\begin{array}{c}{k}_{s0}\&CenterDot;P_0,t\&Element;\&lsqb;1,tr\&rsqb;\\ k_{s1}\&CenterDot;P_0,t\&Element;(tr,td)\\ k_{s2}\&CenterDot;P_0,t\&Element;\&lsqb;td,T\&rsqb;\end{array}\right.---\left(9\right)$

In formula (9), P₀Representing force constant, the minimum non-zero history in water intaking power station is exerted oneself；Tr and td be respectively sunrise and The sunset moment, T be day total time segment length, k_s0、k_s1、k_s2Represent the coefficient that three periods of stages exert oneself respectively.

By solving Remanent Model of exerting oneself, power station residual error of exerting oneself with photovoltaic superposes that to exert oneself be three sections of stepped processes of exerting oneself, Photovoltaic plant residue is exerted oneself and is exerted oneself for standard.Constraint in Remanent Model of exerting oneself includes the constraint of output of power station characteristic, restriction of exerting oneself Constraint, reservoir water balance, reservoir level (storage capacity) constraint, storage-capacity curve constraint, the constraint of level of tail water discharge relation Deng.These constraints are Constrained, do not repeat at this.

Step 5, generates photovoltaic plant and sorts out field of force scape, i.e. photovoltaic plant load process；In conjunction with power station water situation, Generate the combined dispatching system scenarios of prediction day.

This step farther includes sub-step:

Step 5.1: obtain the weather pattern of prediction day according to weather forecast, obtains prediction day according to weather pattern corresponding History daily output sequence class (being hereinafter abbreviated as " current class ") and R, D₂With NG value scope.

Step 5.2: generate photovoltaic standard and exert oneself scene, particularly as follows:

A series of R value is randomly choosed, it is assumed that D in the range of the R that prediction day is corresponding₂Being zero, now Gini coefficient NG is not Existing, photovoltaic plant presses RP_pv,d,tExert oneself, P_pv,d,tRepresent the EIAJ of current class t；This scene of exerting oneself is for ideal Photovoltaic plant under state is exerted oneself, and scene of this being exerted oneself is exerted oneself scene as photovoltaic standard.

Step 5.3: use residual error scaling pretreatment history each to the current apoplexy due to endogenous wind daily output sequence of exerting oneself described in step 2 each The residual error of exerting oneself in momentZoom in and out pretreatment, obtain the standardization in each moment and exert oneself residual error

Step 5.4: stochastic generation one NG value in the range of the NG value that prediction day is corresponding, according to current apoplexy due to endogenous wind each history day Sequence of exerting oneself obtains the residual error process of exerting oneself of the current NG value correspondence of stochastic generation, and residual error of exerting oneself the i.e. standardization of process is exerted oneself Residual errorChange procedure with the moment.

Step 5.5: willInterval the exerting oneself of tr to td is converted to according to the inverse process of residual error scaling pre-treatment step of exerting oneself Residual error processSolve residual error process of exerting oneselfStandard deviation D'₂。

Step 5.6: at the D that prediction day is corresponding₂Stochastic generation one D in the range of value₂, calculate residual error process of exerting oneselfThus obtain photovoltaic plant and exert oneself scene

Step 5.7: if P_pv,t≤P_pv,d,tAnd P_pv,t>=0, record current photovoltaic plant and exert oneself scene, perform step 5.8；No Then, the cycle-index of recording step 5.6, if cycle-index reaches preset times, perform step 5.4, if cycle-index is not Reaching preset times, circulation performs step 5.6.

Step 5.8: artificially select from the photovoltaic plant of record exerts oneself scene according to weather forecast and meet current weather type Photovoltaic plant exert oneself scene, i.e. predict that the photovoltaic plant of day is exerted oneself scene.

Step 5.9: obtained predicting day power station water situation, integrated water power station water by corresponding basin Plan on Hydrological Forecast Situation and prediction day photovoltaic plant exert oneself scene generate water light combined dispatching scene, water light combined dispatching scene include prediction Day power station water situation and photovoltaic plant exert oneself scene.

Step 6, with water light combined dispatching scene as background, solves power station and the exert oneself residual error mould of photovoltaic plant is fully compensated Type, obtains the water light combined dispatching optimized operation scheme of prediction day.

Step 6.1: determine k_s0、k_s1、k_s2Span [k_min,k_max], wherein, k_min=1,P_zjFor Power station installation is exerted oneself；

Step 6.2: at span [k_min,k_max] upper discrete k_s0、k_s1、k_s2, and obtain discrete point k_s0,i、k_s1,j、k_s2,m All combinations.

Discrete steps takes Δ k, discrete counts as Num, and each discrete point is expressed as follows:

$\left\{\begin{array}{c}{k}_{s0,i}=k_{min}+\mathrm{\&Delta;}k*i\\ k_{s1,j}=k_{\min }+\mathrm{\&Delta;}k*j\\ k_{s2,m}=k_{\min }+\mathrm{\&Delta;}k*m\end{array}\right.---\left(10\right)$

In formula (10), i, j, m represent that discrete point is numbered respectively, i, j, m ∈ [0, Num].

Step 6.3:t period photovoltaic plant is exerted oneself as RP_pv,d,t, exert oneself when being fully compensated in power station At each discrete point k_s0,i、k_s1,j、k_s2,mPredict that the power station of day day part is actual by period calculating respectively under combination to exert oneself P_sdz,t。

Step 6.4: according to the actual P that exerts oneself in power station_sdz,tCalculate difference quadratic sum SUM ε more than the compensation under the combination of each discrete point.

If P_sdz,t<P′_sdz,tTime, then ε (t)=P_sdz,t-P′_sdzz,t；Otherwise ε (t)=0, thus obtain predicting that difference is flat more than the compensation of day Fang He

Step 6.5: obtain minimum and compensate remaining discrete point combination corresponding for difference quadratic sum SUM ε, under the combination of this discrete point Day part power station and photovoltaic are exerted oneself scheme i.e. water light combined dispatching optimized operation scheme.

If occurring, minimum compensates the situation of the remaining corresponding many prescriptions case of difference quadratic sum SUM ε, calculates corresponding discrete of each prescription case Exert oneself in power station under some combination, the day part power station under the discrete point combination of correspondence of exerting oneself with maximum power station and photovoltaic Scheme of exerting oneself i.e. water light combined dispatching optimized operation scheme.

Exert oneself as P in power station₀[k_s0,itr+k_s1,j(td-tr-1)+k_s2,m(T-td+1)], k_s0,i、k_s1,j、k_s2,mRepresent from Discrete point in scatterplot combination.

The water light combined dispatching optimized operation scheme that this step is obtained can be used for instructing actual combined dispatching.

Claims (7)

1. a clustering method for the history daily output sequence of photovoltaic plant, is characterized in that, including:

Step 1, according to the form of the history daily output sequence of photovoltaic plant, is divided into M special dictionary by history daily output sequence Type sequence, obtains the outer envelope curve of daily output of each type sequence；

Step 2, calculates the daily output clearness index R of each history daily output sequence, residual error of exerting oneself standard deviation D₂Residual with exerting oneself Difference Gini coefficient NG, daily output clearness indexThe residual error definition of exerting oneself in history daily output sequence each moment ForResidual error of exerting oneself standard deviation D₂With residual error Gini coefficient NG of exerting oneself according to residual error of exerting oneselfMeter Calculate；P_{Pv, d, t}The EIAJ of type sequence t belonging to history daily output sequence, according to outside the daily output of type sequence Envelope curve obtains；P_pv,tFor exerting oneself of history daily output sequence t；T represents the duration of exerting oneself of history daily output sequence； Residual error of exerting oneself for history daily output sequence t；

Described residual error Gini coefficient NG of exerting oneself is adopted and is obtained with the following method:

A () is by Beijing time coordinateBe converted to relative time coordinate For history daily output sequence The residual error of exerting oneself of row t, ta=t-tr,Tr is the initial time of exerting oneself of history daily output sequence；

B () is to relative time coordinateZoom in and out to obtain scaling coordinateScaling coefficientT_dExert oneself duration for current history daily output sequence, T_d,stFor all history daily output sequences T_dMeansigma methods；

C () is to scaling coordinateCarry out interpolation and must scale standard coordinateSt.t represents scaling Moment after process, st.t=1,2 ..., T_d,st,

(d) standardization scaling standard coordinateIn exert oneself residual error

E () normalized is exerted oneself residual errorGini coefficient NG；

Step 3, the cluster of history daily output sequence, this step farther includes:

The D of 3.1 standardization each history daily output sequence₂And NG；

3.2 with D₂It is that history daily output sequence is once clustered by Cluster Assessment index with R；

3.3 manual analysis one time cluster results, with NG and D₂For Cluster Assessment index, be there is repetition in R value scope History daily output sequence class carries out secondary cluster respectively；If there is not R value scope to repeat, terminate.

2. the clustering method of the history daily output sequence of photovoltaic plant as claimed in claim 1, is characterized in that:

Described M is not more than 4.

3. the clustering method of the history daily output sequence of photovoltaic plant as claimed in claim 1, is characterized in that:

The D of described standardization each history daily output sequence₂And NG, use equation below to carry out:

$ST\&CenterDot;X_i=w_i\frac{X_i-\underset{i=1,2...N}{min}\left\{X_i\right\}}{\underset{i=1,2...N}{max}\left\{X_i\right\}-\underset{i=1,2...N}{min}\left\{X_i\right\}}$

Wherein, X_iFor i-th sample value, for D₂Or NG；ST·X_iFor the i-th sample value after standardization；N is Sample number, i.e. history daily output sequence quantity；w_iFor the residual error accumulation proportionality coefficient of exerting oneself of i-th history daily output sequence, Residual error of exerting oneself according to history daily output sequence each moment obtains.

4. the water light day combined scheduling method analyzed based on photovoltaic property, is characterized in that, including:

Step 1, it is thus achieved that R, D that claim 1 gained each history daily output sequence class is corresponding₂, the scope of NG, examine Consider solar irradiation Changing Pattern and solar irradiation under each weather conditions photovoltaic plant to be exerted oneself affecting laws, according to each history day Sequence of exerting oneself apoplexy due to endogenous wind history daily output series modality, manually determines the weather pattern that each history daily output sequence class is corresponding；

Step 2, withBuilding, for object function, Remanent Model of exerting oneself, SUM ε represents compensation Remaining difference quadratic sum；ε (t) be moment t compensation more than poor, Actual for power station moment t is exerted oneself,P₀Representing force constant, water intaking power station is minimum Non-zero history is exerted oneself；Tr and td represents sunrise and sunset moment respectively, and T represents segment length when day is total；

Step 3, sorts out field of force scape and power station water situation in conjunction with photovoltaic plant, generates the combined dispatching system of prediction day System scene, this step farther includes:

3.1 obtain the weather pattern of prediction day according to weather forecast, obtain the history day predicting that day is corresponding according to weather pattern Sequence of exerting oneself class and R, D₂Scope with NG, it was predicted that day, corresponding history daily output sequence class was abbreviated as current class；

3.2 randomly choose a series of R value in the range of the R of current class, it is assumed that D₂Being zero, photovoltaic plant presses RP_pv,d,tGo out Power, scene of this being exerted oneself is exerted oneself scene as photovoltaic standard；P_pv,d,tEIAJ for current class t；

3.3 use sub-step (a)～(d) described method history each to current apoplexy due to endogenous wind daily output in claim 1 step 2 The residual error of exerting oneself in sequence each momentProcess, obtain the standardization in each moment and exert oneself residual error

3.4 in the range of the NG value of current class stochastic generation one NG value, obtain according to current apoplexy due to endogenous wind each history daily output sequence Obtain the residual error process of exerting oneself corresponding for current NG of stochastic generationI.e. standardization is exerted oneself residual errorChange with the moment Change process；

3.5 use the inverse process of sub-step (a)～(d) described method in claim 1 step 2 to process standardization exerts oneself Residual errorWillBe converted to the residual error process of exerting oneself of Beijing time coordinateSolve residual error process of exerting oneself Standard deviation D'₂；

3.6 at the D of current class₂Stochastic generation one D in the range of value₂, calculate current D₂Corresponding residual error process of exerting oneselfObtain photovoltaic plant to exert oneself scene

3.7: if P_pv,t≤P_pv,d,tAnd P_pv,t>=0, record current photovoltaic plant and exert oneself scene, perform step 3.8；Otherwise, The cycle-index of recording step 3.6, if cycle-index reaches preset times, performs step 3.4, if cycle-index is not up to Preset times, circulation performs step 3.6；

3.8 exert oneself from the photovoltaic plant of record according to weather forecast artificially selects the photovoltaic meeting current weather type scene Output of power station scene, i.e. predicts that the photovoltaic plant of day is exerted oneself scene；

3.9 must be predicted a day power station water situation, integrated water power station water situation and prediction day by River Basin Hydrology Prediction version Photovoltaic plant exert oneself scene generate water light combined dispatching scene；

Step 4, with water light combined dispatching scene as background, solves Remanent Model of exerting oneself, it is thus achieved that the water light associating of prediction day Scheduling optimized operation scheme.

5. the water light day combined scheduling method analyzed based on photovoltaic property as claimed in claim 4, is characterized in that:

Described weather conditions include sunny type weather, cloudy type weather, sleet type weather and mixed type weather.

6. the water light day combined scheduling method analyzed based on photovoltaic property as claimed in claim 4, is characterized in that:

Step 4 farther includes:

4.1 determine k_s0、k_s1、k_s2Span [k_min,k_max], k_min=1,P_zjFill for power station Machine is exerted oneself, P₀For going out force constant；

4.2 span [k_min,k_max] upper discrete k_s0、k_s1、k_s2, and obtain all combinations of discrete point；

4.3 calculate the actual P that exerts oneself in power station of prediction day day part by the period under each discrete point combines_sdz,t；

4.4 according to the actual P that exerts oneself in power station_sdz,tCalculate difference quadratic sum SUM ε more than the compensation under the combination of each discrete point；

4.5 exert oneself with photovoltaic with the day part power station under discrete point combination corresponding for minimum compensation remaining difference quadratic sum SUM ε Scheme is as water light combined dispatching optimized operation scheme.

7. the water light day combined scheduling method analyzed based on photovoltaic property as claimed in claim 4, is characterized in that:

In sub-step 4.5, if occurring, minimum compensates the situation of the remaining corresponding many prescriptions case of difference quadratic sum SUM ε, calculates each group Exert oneself in the lower power station of discrete point combination corresponding to scheme, with maximum power station exert oneself the discrete point of correspondence combine under each time Section power station and photovoltaic are exerted oneself scheme i.e. water light combined dispatching optimized operation scheme.