CN105469163A - Similar day selection method used for photovoltaic power station power prediction - Google Patents

Abstract

The invention relates to a similar day selection method used for photovoltaic power station power prediction. The similar day selection method comprises the following steps: determining a meteorological information vector; and normalizing and calculating a correlation coefficient, calculating a subjective weight and an objective weight, and finally obtaining the similarity on each history day by a prediction day, wherein the history data with a large similarity numerical value is the similar day. The similar day selection method provides accurate basic data for photovoltaic output forecasting on the basis of the combination of the association coefficient and the combined weight.

Description

A kind of system of selection of the similar day for predicting power of photovoltaic plant

Technical field

The present invention relates to photovoltaic power generation power prediction field, be specifically related to the individual system of selection for the similar day of predicting power of photovoltaic plant.

Background technology

Along with the increase year by year of domestic photovoltaic power generation grid-connecting scale, the research both at home and abroad for photovoltaic power generation grid-connecting problem is more and more deep, but the research of photovoltaic power generation output forecasting does not still reach gratifying degree.The output power of photovoltaic generating system is subject to the impact of the multiple meteorologic factors such as solar irradiation intensity, exposure time, temperature, air pressure, humidity, has certain time variation and randomness, how to improve precision of prediction and still face larger difficulty.

In photovoltaic generation prediction, is usually called prediction day that day of exerting oneself by prognoses system, the some skies before this day is called the history day of prediction day, within the most close one day, be called Japan-China for history " similar day " with prediction day Weather information.It is generally acknowledged, the generated energy data of generated energy data to prediction day of similar day have important reference significance.

At present, the selection of similar day general rule of thumb or " Euclidean distance " result of calculation choose, the difficult quality guarantee of the similar day chosen, also have and choose similar day by degree of association method, but what traditional degree of association method adopted when choosing similar day is wait power process, and in fact each factor influence degree of exerting oneself to photovoltaic is not identical, therefore must give different weights as the case may be.How to choose and improve photovoltaic power generation output forecasting precision based on similar day comparatively accurately there is important researching value.

Summary of the invention

The object of this invention is to provide the method for a kind of accurate selection for the similar day of predicting power of photovoltaic plant.

The present invention adopts following technical scheme:

For a system of selection for the similar day of predicting power of photovoltaic plant, comprise the steps:

1) from prediction day, reverse several history days selecting season identical with predicting day, same weather pattern in historical data base, the weather information vector predicting day and each history day is obtained respectively; Described weather information vector comprises following weather information component of a vector:

2) respectively by step 1) component of weather information vector that obtains adopts range method method to be normalized;

3) correlation coefficient method calculation procedure 1 is adopted) the weather information component of a vector of described prediction day is to step 2) correlation coefficient of the weather information component of a vector of each history day after the normalization that obtains;

4) analytical hierarchy process calculation procedure 3 is adopted) the subjective weight of correlation coefficient that obtains;

5) entropy assessment calculation procedure 3 is adopted) objective weight of correlation coefficient that obtains;

6) by step 4) the subjective weight that obtains and step 5) objective weight that obtains carries out linear weighted function and obtains combining weights, utilizing correlation coefficient and combining weights calculation procedure 1) described prediction day, the history day that similarity numerical value is large was similar day to the similarity of each history day.

Further, described step 1) in weather information component of a vector determine by the following method:

The determination of the weather information vector of prediction day: obtain solar radiation in 12 hours next day from meteorological department and, according to average strength, air pressure mean value, relative humidity mean value, in a few days maximum temperature, in a few days minimum temperature and mean daily temperature, form weather information vector X:

X＝[S _aver，P _aver，h _aver，T _h，T _l，T _aver](1)

In formula (1), S _averbe that in 12 hours, average strength is shone in solar radiation; P _averfor air pressure mean value; h _averfor relative humidity mean value; T _h, T _l, T _averbe respectively in a few days maximum temperature, minimum temperature and medial temperature;

The determination of the weather information vector of history day: from prediction day, reverse several history days selecting same weather pattern in same season from historical data base, is made up of the weather information vector of each history day formula (1).

Further, step 2) in method for normalizing be:

Following formula is adopted to be normalized each weather information component of a vector:

x _j(i)＝[y _j(i)-m(i)]/[M(i)-m(i)](2)

Y in formula _ji () is the weather information component of a vector before normalization, x _ji () is the numerical value after normalization; M (i) and M (i) is respectively minimum value and the maximal value of i-th component of the weather information vector before normalization.

Further, step 3) described in correlation coefficient obtain by the following method:

Suppose that weather information vector has n component, the weather information vector of the prediction day after normalization and a jth history day is respectively x ₀and x _j, then have:

x ₀＝[x ₀(1)，x ₀(2)，…，x ₀(n)] ^T(3)

x _j＝[x _j(1)，x _j(2)，…，x _j(n)] ^T(4)

X ₀to x _jat the correlation coefficient of a kth component be:

${\mathrm{\ϵ}}_j\left(k\right)=\frac{\underset{j}{\min }\underset{k}{\min }\left|x_0\left(k\right)-x_j\left(k\right)\right|+\mathrm{\ρ}\underset{j}{\max }\underset{k}{\max }\left|x_0\left(k\right)-x_j\left(k\right)\right|}{\left|x_0\left(k\right)-x_j\left(k\right)\right|+\mathrm{\ρ}\underset{j}{\max }\underset{k}{\max }\left|x_0\left(k\right)-x_j\left(k\right)\right|}---\left(5\right)$

In formula (5): ρ is resolution ratio.ρ value generally gets 0.5.

Further, step 4) described in subjective weight obtain by the following method:

A) the judgment matrix A of weather information vector is determined:

$A=\left[\begin{array}{cccc}{a}_{11}& a_{12}& ...& a_{1n}\\ a_{21}& a_{22}& ...& a_{2n}\\ ...& ...& ...& ...\\ a_{n1}& a_{n2}& ...& a_{nn}\end{array}\right]=\left[\begin{array}{cccc}{\mathrm{\ω}}_1/{\mathrm{\ω}}_1& {\mathrm{\ω}}_1/{\mathrm{\ω}}_2& ...& {\mathrm{\ω}}_1/{\mathrm{\ω}}_n\\ {\mathrm{\ω}}_2/{\mathrm{\ω}}_1& {\mathrm{\ω}}_2/{\mathrm{\ω}}_2& ...& {\mathrm{\ω}}_2/{\mathrm{\ω}}_n\\ ...& ...& ...& ...\\ {\mathrm{\ω}}_n/{\mathrm{\ω}}_1& {\mathrm{\ω}}_n/{\mathrm{\ω}}_2& ...& {\mathrm{\ω}}_n/{\mathrm{\ω}}_n\end{array}\right]---\left(6\right)$

In formula (6), a _ijrepresent the relative importance of i-th target and a jth target;

A _ij=ω _i/ ω _jit is 1,3,5,7,9 that element ω scale gets Satty scale; As shown in table 1 below.

Table 1 judgment matrix scale value and corresponding implication

Scale value	Scale implication
		a ij＝1	Index ω iRelative indicatrix ω jNo less important
a ij＝3	Index ω iRelative indicatrix ω jImportant a little
		a ij＝5	Index ω iRelative indicatrix ω jObviously important
a ij＝7	Index ω iRelative indicatrix ω jStrongly important
		a ij＝9	Index ω iRelative indicatrix ω jExtremely important

B) consistency check is carried out:

$CI=\frac{{\mathrm{\λ}}_{\max }-n}{n-1}---\left(7\right)$

In formula (7), λ _maxfor the Maximum characteristic root of matrix A, n is the dimension of matrix A;

Can think that judgment matrix A has consistance when meeting formula (8).Wherein the value of RI is in table 2.

$CR=\frac{CI}{RI}<0.10---\left(8\right)$

The explanation of table 2RI value

n	1	2	3	4	5	6	7	8	9	10
											RI	0.00	0.00	0.58	0.90	1.12	1.24	1.32	1.41	1.45	1.49

C) subjective weight is asked for:

First the product M of each row element of judgment matrix A is calculated _i:

$M_i=\underset{j=1}{\overset{n}{\&Pi;}}{a}_{ij},i=1,2,...n---\left(9\right)$

Then M is calculated _in th Root

${\stackrel{\&OverBar;}{W}}_i=\sqrt[n]{M_i}---\left(10\right)$

Finally to vector normalization (normalized):

$W_i=\frac{{\stackrel{\&OverBar;}{W}}_i}{\underset{j=1}{\overset{n}{\&Sigma;}}\stackrel{\&OverBar;}{W_i}},i=1,2,\mathrm{...}n---\left(11\right)$

Then W _a=[W ₁, W ₂..., W _n] ^tbe required subjective weight.

Further, step 5) described in objective weight obtain by the following method:

Adopt Information Entropy determination objective weight:

$E_i=-K\underset{j=1}{\overset{n}{\&Sigma;}}{p}_{ij}{\mathrm{lnp}}_{ij}---\left(12\right)$

$W_S=\frac{1-E_i}{\underset{i=1}{\overset{m}{\&Sigma;}}(1-E_i)}---\left(13\right)$

In formula (12), r _ijfor the element of judgment matrix R, E _ibe the information entropy of i-th index, i=1,2 ..., m is index number, j=1,2 ..., n is evaluation object number, COEFFICIENT K=1/lnn.

Further, step 6) described in similarity obtain by the following method:

The weather information component of a vector of integrated forecasting day, to the correlation coefficient of the weather information component of a vector of each history day, draws whole x ₀to x _jsimilarity be:

$r_i=\underset{k=1}{\overset{n}{\&Sigma;}}{\mathrm{\&epsiv;}}_i\left(k\right)\&CenterDot;{\mathrm{\&lambda;}}_k---\left(14\right)$

In formula (14), λ _kfor the weight coefficient of each component, its computing formula is as follows:

λ _k＝k _AW _A+k _SW _S(15)

In formula (15), k _a, k _sbe respectively the subjective and objective preference coefficient of linear weighted function, the equal value of main, objective preference coefficient is 0.5, λ _kfor combining weights, W _afor subjective weight, W _sfor objective weight.

Beneficial effect of the present invention is: the multiple meteorologic factor such as solar irradiation intensity, exposure time, temperature, air pressure, humidity affecting the output power of photovoltaic generating system considers by the present invention, adopt non-authority processing method of Denging, give different weights as the case may be, propose the similar day system of selection combined with combining weights based on correlation coefficient, eliminate largely each meteorologic factor exist time variation and randomness on the impact of precision of prediction, for photovoltaic power generation output forecasting provides basic data comparatively accurately.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the inventive method.

Embodiment

Below in conjunction with Fig. 1 and embodiment, the present invention is described in detail.Scope is not limited to embodiment, and those skilled in the art make any change and also belong to the scope of protection of the invention in the scope that claim limits.

Embodiment

Below for somewhere photo-voltaic power generation station, the system of selection of similar day is described.

One, weather information vector is determined:

Obtain solar radiation in 12 hours next day from meteorological department and shine average strength, air pressure mean value, relative humidity mean value, in a few days maximum temperature, in a few days minimum temperature and mean daily temperature.

From prediction day, reverse several history days selecting same weather pattern in same season from historical data base, in 12 hours that obtain each history Japan and China, solar radiation is according to average strength, air pressure mean value, relative humidity mean value, in a few days maximum temperature, in a few days minimum temperature and mean daily temperature.

About same weather pattern, this sentences fine day day type is example, and from prediction day, the reverse fine day history day selecting some in same season from historical data base, form weather information vector set, the system of selection of other types day is similar.

Data are shown in Table 3.In table 1, August 2 was prediction day, and other are history day.

Table 3 weather information related data

Each weather information in table 3 brings the weather information vector being made up of prediction day and each history day formula (1) into, and each weather information vector forms weather information vector set.

Two, each component normalized of weather information vector:

Formula (2) is utilized to adopt range method to be normalized to the weather information vector determined.

Three, correlation coefficient is asked for:

By the weather information component of a vector of formula (3), (4), (5) computational prediction day to the correlation coefficient of the weather information component of a vector of each history day after normalization, result is as shown in table 4:

The correlation coefficient of day to history day each weather information component of a vector predicted by table 4

8-01	0.62	0.87	0.83	0.57	0.63	0.65
							7-29	0.73	0.67	0.63	0.33	0.41	0.33
7-28	0.36	0.65	0.33	0.57	0.58	0.57
							7-27	0.40	0.49	0.45	0.57	0.48	0.52
7-26	0.33	0.35	0.63	0.73	0.37	0.47
							7-25	0.42	0.45	0.45	0.63	0.33	0.40
7-23	0.40	0.47	0.56	0.73	0.38	0.40

Four, subjective weights W is calculated _a:

A) the judgment matrix A of weather information vector is determined by formula (6), as shown in table 5:

The each component judgment matrix of table 5

	S aver	P aver	h aver	T k	T l	T aver
							S aver	1	5	5	3	3	3
P aver	1/5	1	1	1/3	1/3	1/3
							h aver	1/5	1	1	1/3	1/3	1/3
T k	1/3	3	3	1	1	1
							T l	1/3	3	3	1	1	1
T aver	1/3	3	3	1	1	1

$A=\left[\begin{array}{cccccc}1& 5& 5& 3& 3& 3\\ \frac{1}{5}& 1& 1& \frac{1}{3}& \frac{1}{3}& \frac{1}{3}\\ \frac{1}{5}& 1& 1& \frac{1}{3}& \frac{1}{3}& \frac{1}{3}\\ \frac{1}{3}& 3& 3& 1& 1& 1\\ \frac{1}{3}& 3& 3& 1& 1& 1\\ \frac{1}{3}& 3& 3& 1& 1& 1\end{array}\right].$

B) consistency check is carried out by formula (7), (8): λ _max=6.0581, CI=0.0116.As shown in Table 2, as n=6, RI=1.24; Calculate CR=0.009 by formula (8) again, be less than 0.10, namely judgment matrix A has consistance.

C) subjective weight is asked for:

Calculated by formula (9), (10), (11): $W_A=\left[\begin{array}{c}0.3978\\ 0.0593\\ 0.0593\\ 0.1612\\ 0.1612\\ 0.1612\end{array}\right].$

(4) objective weight W is calculated _s:

Calculated by formula (12), E _i=[0.97890.97820.98100.98740.98660.9881].

Calculated by formula (13),

(5) similarity is calculated:

Combining weights λ is calculated by formula (15) _k=[0.21400.21010.19240.12780.13550.1202].

Similarity is calculated by formula (14) $\mathrm{\&gamma;}=\left[\begin{array}{c}0.6769\\ 0.5485\\ 0.4891\\ 0.4725\\ 0.4535\\ 0.4428\\ 0.4733\end{array}\right].$

According to result of calculation, what similarity was maximum is 0.6769, and namely August 1 was similar day.

According to the above embodiments, the present invention is described in detail.It should be noted that, above embodiment is only illustratively invented.Under the prerequisite not departing from spirit of the present invention and essence, those skilled in the art can design multiple alternative of the present invention and improvement project, and it all should be understood to be within protection scope of the present invention.

Claims (7)

1. for a system of selection for the similar day of predicting power of photovoltaic plant, it is characterized in that, comprise the steps:

1) from prediction day, reverse several history days selecting season identical with predicting day, same weather pattern in historical data base, the weather information vector predicting day and each history day is obtained respectively;

2) respectively by step 1) component of weather information vector that obtains adopts range method method to be normalized;

3) correlation coefficient method calculation procedure 1 is adopted) the weather information component of a vector of described prediction day is to step 2) correlation coefficient of the weather information component of a vector of each history day after the normalization that obtains;

4) analytical hierarchy process calculation procedure 3 is adopted) the subjective weight of correlation coefficient that obtains;

5) entropy assessment calculation procedure 3 is adopted) objective weight of correlation coefficient that obtains;

6) by step 4) the subjective weight that obtains and step 5) objective weight that obtains carries out linear weighted function and obtains combining weights, utilizing correlation coefficient and combining weights calculation procedure 1) described prediction day, the history day that similarity numerical value is large was similar day to the similarity of each history day.

2. the system of selection of a kind of similar day for predicting power of photovoltaic plant according to claim 1, is characterized in that, described step 1) in weather information vector determine by the following method:

The determination of the weather information vector of prediction day: obtain solar radiation in 12 hours next day from meteorological department and, according to average strength, air pressure mean value, relative humidity mean value, in a few days maximum temperature, in a few days minimum temperature and mean daily temperature, form weather information vector X:

X＝[S _aver,P _aver,h _aver,T _h,T _l,T _aver](1)

In formula (1), S _averbe that in 12 hours, average strength is shone in solar radiation; P _averfor air pressure mean value; h _averfor relative humidity mean value; T _h, T _l, T _averbe respectively in a few days maximum temperature, minimum temperature and medial temperature;

The determination of the weather information vector of history day: from prediction day, reverse several history days selecting same weather pattern in same season from historical data base, is made up of the weather information vector of each history day formula (1).

3. the system of selection of a kind of similar day for predicting power of photovoltaic plant according to claim 2, is characterized in that, step 2) in method for normalizing be:

Following formula is adopted to be normalized each weather information component of a vector:

x _j(i)＝[y _j(i)-m(i)]/[M(i)-m(i)](2)

Y in formula _ji () is the weather information component of a vector before normalization, x _ji () is the numerical value after normalization; M (i) and M (i) is respectively minimum value and the maximal value of i-th component of the weather information vector before normalization.

4. the system of selection of a kind of similar day for predicting power of photovoltaic plant according to claim 3, is characterized in that, step 3) described in correlation coefficient obtain by the following method:

Suppose that weather information vector has n component, the weather information vector of the prediction day after normalization and a jth history day is respectively x ₀and x _j, then have:

x ₀＝[x ₀(1),x ₀(2),…,x ₀(n)] ^T(3)

x _j＝[x _j(1),x _j(2),…,x _j(n)] ^T(4)

X ₀to x _jat the correlation coefficient of a kth component be:

${\mathrm{\&epsiv;}}_j\left(k\right)=\frac{\underset{j}{\min }\underset{k}{\min }|x_0\left(k\right)-x_j\left(k\right)|+\mathrm{\&rho;}\underset{j}{\max }\underset{k}{\max }|x_0\left(k\right)-x_j\left(k\right)|}{|x_0\left(k\right)-x_j\left(k\right)|+\mathrm{\&rho;}\underset{j}{\max }\underset{k}{\max }|x_0\left(k\right)-x_j\left(k\right)|}---\left(5\right)$

In formula (5): ρ is resolution ratio.

5. the system of selection of a kind of similar day for predicting power of photovoltaic plant according to claim 4, is characterized in that, step 4) described in subjective weight obtain by the following method:

A) the judgment matrix A of weather information vector is determined:

$A=\left[\begin{array}{cccc}{a}_{11}& a_{12}& ...& a_{1n}\\ a_{21}& a_{22}& ...& a_{2n}\\ ...& ...& ...& ...\\ a_{n1}& a_{n2}& ...& a_{nn}\end{array}\right]=\left[\begin{array}{cccc}{\mathrm{\&omega;}}_1/{\mathrm{\&omega;}}_1& {\mathrm{\&omega;}}_1/{\mathrm{\&omega;}}_2& ...& {\mathrm{\&omega;}}_1/{\mathrm{\&omega;}}_n\\ {\mathrm{\&omega;}}_2/{\mathrm{\&omega;}}_1& {\mathrm{\&omega;}}_2/{\mathrm{\&omega;}}_2& ...& {\mathrm{\&omega;}}_2/{\mathrm{\&omega;}}_n\\ ...& ...& ...& ...\\ {\mathrm{\&omega;}}_n/{\mathrm{\&omega;}}_1& {\mathrm{\&omega;}}_n/{\mathrm{\&omega;}}_2& ...& {\mathrm{\&omega;}}_n/{\mathrm{\&omega;}}_n\end{array}\right]---\left(6\right)$

In formula (6), a _ijrepresent the _ithe relative importance of individual target and a jth target;

A _ij=ω _i/ ω _jit is 1,3,5,7,9 that element ω scale gets Satty scale;

B) consistency check is carried out:

$CI=\frac{{\mathrm{\&lambda;}}_{max}-n}{n-1}---\left(7\right)$

In formula (7), λ _maxfor the Maximum characteristic root of matrix A, n is the dimension of matrix A;

Can think that judgment matrix A has consistance when meeting formula (8):

$CR=\frac{CI}{RI}<0.10---\left(8\right)$

C) subjective weight is asked for:

First the product M of each row element of judgment matrix A is calculated _i

$M_i=\underset{j=1}{\overset{n}{\&Pi;}}{a}_{ij},i=1,2,...n---\left(9\right)$

Then M is calculated _in th Root

$\stackrel{\&OverBar;}{W_i}=\sqrt[n]{M_i}---\left(10\right)$

Finally to vector normalization (normalized):

$W_i=\frac{\stackrel{\&OverBar;}{W_i}}{\underset{j=1}{\overset{n}{\&Sigma;}}\stackrel{\&OverBar;}{W_i}},i=1,2,...n---\left(11\right)$

Then W _a=[W ₁, W ₂..., W _n] ^tbe required subjective weight.

6. the system of selection of a kind of similar day for predicting power of photovoltaic plant according to claim 5, is characterized in that, step 5) described in objective weight obtain by the following method:

Adopt Information Entropy determination objective weight:

$E_i=-K\underset{j=1}{\overset{n}{\&Sigma;}}{p}_{ij}\ln p_{ij}---\left(12\right)$

$W_S=\frac{1-E_i}{\underset{i=1}{\overset{m}{\&Sigma;}}(1-E_i)}---\left(13\right)$

In formula (12), r _ijfor the element of judgment matrix R, E _ibe the information entropy of i-th index, i=1,2 ..., m is index number, j=1,2 ..., n is evaluation object number, COEFFICIENT K=1/lnn.

7. the system of selection of a kind of similar day for predicting power of photovoltaic plant according to claim 6, is characterized in that, step 6) described in similarity obtain by the following method:

The weather information component of a vector of integrated forecasting day, to the correlation coefficient of the weather information component of a vector of each history day, draws whole x ₀to x _jsimilarity be:

$r_i=\underset{k=1}{\overset{n}{\&Sigma;}}{\mathrm{\&epsiv;}}_i\left(k\right)\&CenterDot;{\mathrm{\&lambda;}}_k---\left(14\right)$

In formula (14), λ _kfor the weight coefficient of each component, its computing formula is as follows:

λ _k＝k _AW _A+k _SW _S(15)

In formula (15), k _a, k _sbe respectively the subjective and objective preference coefficient of linear weighted function, λ _kfor combining weights, W _afor subjective weight, W _sfor objective weight.