CN104156777B - Low-cost photovoltaic power prediction method based on city weather forecasts - Google Patents

Abstract

The invention discloses a low-cost photovoltaic power prediction method based on city weather forecasts. The low-cost photovoltaic power prediction method comprises the steps of first obtaining penetration coefficients of extrasolar solar radiation through the atmosphere by calculating the extrasolar radiation intensity through a Sun-Earth model; then obtaining a weather forecast type classifier model of statistic penetration coefficients which can be obtained; then obtaining the statistic penetration coefficients of 14 days ago; conducting subtraction of the penetration coefficients and the obtained statistic penetration coefficients to obtain a difference value sequence; conducting multiply-accumulate of the different value sequence with a corresponding weight sequence to obtain a penetration coefficient error; correcting the statistic penetration coefficients under a weather type predicted on the same day through the penetration coefficient error; meanwhile, using a solar radiation intensity curve on the prediction day to reversely obtain a prediction solar radiation intensity curve; and obtaining a generated power prediction curve of a power station through the light intensity-power relation. The low-cost photovoltaic power prediction method based on the city weather forecasts can achieve real-time prediction based on the release of weather state forecasts and does not rely on high-cost numerical weather forecast products, and the cost of the generated power forecast is reduced.

Description

A kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast

Technical field

The present invention relates to generation of electricity by new energy electric powder prediction, more particularly, it relates to a kind of be based on city weather forecast Inexpensive photovoltaic power generation power prediction method.

Background technology

Energy due to photovoltaic generation derives from the sun, and intensity of solar radiation is affected by weather it is impossible to direct predict, only Can be by the indirect predictions of weather condition be realized with the prediction of photovoltaic generation, so when carrying out photovoltaic power generation power prediction, needing Will support based on weather forecast.Photovoltaic power generation power prediction method both domestic and external is all based on numerical weather forecast at present, Numerical weather forecast needs to buy customization to meteorological department, costly.Additionally, numerical weather forecast product data amount is big, and The learning method using is again complicated, often due to fast prediction that is computationally intensive and not enabling photovoltaic generation power.

Find by prior art documents, document《Least square method supporting vector machine is in photovoltaic power prediction Application [J]》(electric power network technique, 2011,07:One kind is proposed with satellite cloud picture as input data, by a young waiter in a wineshop or an inn in 54-59.) The method after taking advantage of SVMs to set up forecast model, photovoltaic power output being made to predict.The method needs substantial amounts of satellite cloud picture Data, and employ SVMs to be trained device, operand is than larger.Chinese Patent Application No. is： 201110369756.X, entitled：Based on the method for predicting output power of power generation in photovoltaic power station of meteorological element, mention in this application Method for predicting output power of power generation in photovoltaic power station to count as sample using the weather that local weather gathers, by BP neural network For training aids, inputted as the classification of photovoltaic power output prediction with numerical weather forecast.The method needs detailed meteorology to go through History data, is also required to the support of numerical weather forecast during prediction, do not consider to be measured with history with historical values weather forecast Weather difference to be counted is modified to future anticipation, and computationally intensive.

Content of the invention

In order to overcome above-mentioned technical problem, the present invention propose a kind of be independent of the high numerical weather forecast product of price and Meteorological historical data can achieve to photovoltaic power generation power prediction the inexpensive photovoltaic generation based on city weather forecast in detail Power prediction.

The technical solution adopted for the present invention to solve the technical problems is as described below：A kind of low based on city weather forecast Cost photovoltaic power generation power forecasting method is it is characterised in that comprise the steps：

Step (1), according to day ground model calculate history altitude of the sun change curve sequence { S_n(t) }, according to being the outer sun Radiation constant 1353w/m²Calculate intensity of solar radiation Curve Sequences { G outside department of history_n(t)}；Calculate history day measurement sun spoke Firing association's amount and the ratio of the Japanese outer solar radiation total amount of history, i.e. history day radiation penetrating coefficient sequence { α_n}；According to history city City's weather forecast sequence { w_nDay radiation penetrating coefficient is classified, showing that the radiation of the statistics under each weather typing penetrates is Number classification chart J (w_m)；

Step (2), according to the weather forecast sequence { w of first 14 days_jAnd by statistics radiation penetrating coefficient classification chart J (w_m) can Draw the statistic of classification penetrating coefficient sequence of first 14 daysBy the 14 days history penetrating coefficient sequences drawing with Practical Calculation {α_jMake difference calculate 14 days penetrating coefficient error sequence { θ_j, it is multiplied by weights sequence { Δ_jSue for peace afterwards, penetrate as prediction and be The round-off error θ of number；

Step (3), prediction the previous day target day forecast the same day weather pattern w_pDivided according to statistics radiation penetrating coefficient Class table J (w_m) obtain prediction classification penetrating coefficient α_p, add prediction classification penetrating coefficient α with round-off error θ_pShow that prediction penetrates Factor alpha_Δ；That calculate prediction target day is outer intensity of solar radiation curve G_pIt is multiplied by prediction penetrating coefficient α after (t)_ΔDraw the same day Intensity of solar radiation prediction curve F_p(t)；Intensity of illumination according to power station-power relation K (F), calculates power station generated energy Prediction curve P (t).

Is that outer intensity of solar radiation curve G (t) computational methods are as follows in step (1) and step (3)：

δ=sin^-1(0.39795cos(0.98563(N-173))) (1)

In formula, δ is the solar declination on the same day；N be from annual January 1 start calculate number of days；Geographical latitude for power station Degree；T is the hour angle of sidereal time；1353 is to be outer solar radiation constant；S (t) is altitude of the sun change curve, passes through transposition in formula And taking value more than 0, this part represents a day daytime part, whenMoment be the sunrise sunset moment.

In step (1), day radiates penetrating coefficient α_nCalculate function as follows：

In formula, F_nT () is measurement solar radiation variations curve.

In step (2), weather forecast w is divided into the weather pattern set of 33 classes according to the definition one of Chinese weather net {w_m, there is statistics radiation penetrating coefficient classification chart J (w_m) be：

In formula, α_iFor classification results；{ C } is to have w in historical set n_n=w_mSample set；Size ({ C }) is this sample The size of set.

In step (2), 14 days penetrating coefficient error sequence { θ_jBe：

In formula, α_j=J (w_j)

In step (2), the round-off error θ of prediction penetrating coefficient is：

θ=∑ θ_jΔ_j( 7)

In formula, Δ_jFor first 14 days weight coefficients, it is set to:

{Δ_j}={ 0.25,0.2,0.16,0.14,0.1,0.05,0.03,0.02,0.01,0.008,0.008,0 .008, 0.008,0.008}

In step (3), there is prediction penetrating coefficient α_ΔFor：

α_Δ=θ+α_p(8)

In formula, α_p=J (w_p).

In step (3), there is intensity of solar radiation prediction curve F_pT () is：

F_p(t)=α_ΔG_p(t). (9)

In step (3), by the intensity of illumination-power relation K (F) in known power station, there is power station generated energy prediction curve P T () is：

P (t)=K (F_p(t)). (10)

According to the present invention of said structure, its advantage is, not only method is easy for the present invention, and to hardware performance Require low, real-time estimate can be realized according to the issue of state of weather prediction, be independent of the high numerical weather forecast of price simultaneously Product, decreases the cost of generated power forecasting.

Brief description

Below in conjunction with the accompanying drawings and embodiment the present invention is described further.

Fig. 1 is a kind of inexpensive photovoltaic power generation power prediction procedure figure based on city weather forecast of the present invention；

Fig. 2 is a kind of Sino-Japan ground of inexpensive photovoltaic power generation power prediction procedure based on city weather forecast of the present invention Model SEM schemes；

Fig. 3 be the present invention a kind of based on the system in the inexpensive photovoltaic power generation power prediction procedure of city weather forecast Meter penetrating coefficient classification based training module SPCCT figure；

Fig. 4 be the present invention a kind of based on wearing in the inexpensive photovoltaic power generation power prediction procedure of city weather forecast Coefficient round-off error computing module PCCEC figure thoroughly；

Fig. 5 be the present invention a kind of based on defeated in the inexpensive photovoltaic power generation power prediction algorithmic procedure of city weather forecast Go out power classification prediction module OPP figure.

Specific embodiment

A kind of inexpensive photovoltaic power generation power prediction algorithm based on city weather forecast,

Process is as shown in Figure 1：In statistics penetrating coefficient classification based training module SPCCT, according to day ground model, calculate history too Yanggao County degree change curve sequence { S_n(t) }, according to solar radiation constant 1353w/m², calculate solar radiation variations outside department of history Curve Sequences { G_n(t)}；Calculate ratio history day measuring solar radiation total amount and the Japanese outer solar radiation total amount of history, that is, go through Shi radiates penetrating coefficient sequence { α_n, according to historical city weather forecast sequence { w_n, day radiation penetrating coefficient is carried out point Class, draws the statistics radiation penetrating coefficient classification chart J (w under each weather typing_m)；

In penetrating coefficient round-off error computing module PCCEC, according to the weather forecast sequence { w of first 14 days_jAnd by counting Radiation penetrating coefficient classification chart J (w_m) draw the statistic of classification penetrating coefficient sequence of first 14 daysBy drawing with Practical Calculation 14 days history penetrating coefficient sequence { α_jMake difference calculate 14 days penetrating coefficient error sequence { θ_jIt is multiplied by weights sequence { Δ_j} After sue for peace, as prediction penetrating coefficient round-off error θ.

In power output classification prediction module OPP, forecast the weather pattern w on the same day prediction the previous day target day_pAccording to Statistics radiation penetrating coefficient classification chart J (w_m) obtain prediction classification penetrating coefficient α_p, add that prediction classification penetrates with round-off error θ Factor alpha_pObtain predicting penetrating coefficient α_Δ；That calculate prediction target day is external radiation intensity curve G_pIt is multiplied by prediction after (t) and penetrates and be Number α_Δ, draw the intensity of solar radiation prediction curve F on the same day_p(t)；Intensity of illumination according to power station-power relation K (F), calculates Go out power station generated energy prediction curve P (t).

Day ground model SEM is as shown in Fig. 2 N is one day corresponding number of days in a year；6Solar declination is too Positive declination calculator；δ is solar declination；Geographic latitude for power station；7Solar elevation is altitude of the sun calculator；S T () is the altitude of the sun curve of output；8Phase shift is phase shifter；For phase offsetAltitude of the sun Curve；9Diurnal judgment is daytime determining device；Phase shift altitude of the sun curve for part on daytime；1353 (Wm^-2) it is solar radiation constant；G (t) is is outer intensity of solar radiation curve.

Because the eccentricity when earth revolves around the sun is very low, can be approximated to be circular motion, have yellow red angle to fix again Constant, according to one day number of days in a year, the solar declination on the same day can be calculated.Geographic latitude according to power station and the sun are red Latitude, can calculate the altitude of the sun of each moment t in the middle of a day, that is, have altitude of the sun curve S (t) of output, and wherein t adopts It is sidereal time hour angle as chronomere.By phase shifter 8Phase shift by S (t) phase shiftAfter haveNowIt is just daytime more than 0 part,It is night less than 0 part,When being sunrise sunset during equal to 0 Carve.During due to nightValue change to intensity of solar radiation not in all senses, so passing through daytime determining device 9Diurnal judgment sets to 0 night part and obtainsReacting condition solar radiation due to sun altitude When the change of intensity and direct projection altitude of the sun be 1, and be outer sunlight direct radiation intensity be constant 1353 (Wm^-2), it is thus regarded that being outer Intensity of solar radiation curve G (t) isBe outer sunlight direct radiation intensity be the long-pending of constant.

Statistics penetrating coefficient classification based training module SPCCT is as shown in figure 3,1-1SEM is day ground model；Geography for power station Latitude；{N_nIt is corresponding sky Number Sequence in a year of each day in historical data；{G_n(t) } become for solar radiation outside logical department of history Change Curve Sequences；{F_n(t) } measure solar radiation variations Curve Sequences for history；2Integral comparator is integration ratio Compared with device；{α_nRadiate penetrating coefficient sequence for history day；{α_jBe nearly 14 days history day radiation penetrating coefficient sequence；{w_nBe Historical city weather forecast sequence；4Trainer is statistical average grader；J(w_m) radiate penetrating coefficient classification chart for statistics.

By known geography information and temporal information, by day ground model 1-1SEM, calculate each sky in historical data right Answer is outer solar radiation variations curve.Calculate sun spoke outside whole day system using integral contrast device 2Integral comparator Ratio is done, this value describes solar radiation and counts penetration capacity to atmospheric day, within the next few days after the amount of penetrating and measurement solar radiation quantity Radiation penetrating coefficient.By weather history is forecast that type and history day radiate penetrating coefficient input value statistical average grader 4Trainer, draws the average penetration coefficient under various weather forecasts, as statistics radiation penetrating coefficient classification chart J (w_m).

Penetrating coefficient round-off error computing module PCCEC is as shown in figure 4, J (w_m) radiate penetrating coefficient classification chart for statistics； {w_jIt is nearly historical city weather forecast sequence；3Classifier is grader；Penetrate for the statistic of classifications of nearly 14 days and be Number Sequence；{α_jBe nearly 14 days history day radiation penetrating coefficient sequence；{θ_jIt is 14 days penetrating coefficient error sequences；{Δ_jBe 14 days weight coefficient sequence；6Multiplier accumulator is adder and multiplier；θ is prediction penetrating coefficient round-off error.

Grader 3Classifier is carried out to the weather forecast type of nearly 14 days according to statistics radiation penetrating coefficient classification chart Classification draws the statistic of classification penetrating coefficient sequence of 14 days{α_jAndThe history day radiation of nearly 14 days is drawn after making difference Penetrating coefficient sequence { θ_j}.Due to the difference according to time gap, predicated error is different to the relevance predicting the outcome, definition and It is worth 14 days weight coefficient sequence { Δs for 1_j, and radiate penetrating coefficient sequence { θ with the history day of nearly 14 days_jInput in the lump to Adder and multiplier 6Multiplier accumulator, obtains predicting penetrating coefficient round-off error θ, this value is used for revising during prediction Statistic of classification penetrating coefficient.

Power output classification prediction module OPP is as shown in figure 5, J (w_m) radiate penetrating coefficient classification chart for statistics；w_pFor front The weather pattern on one day forecast same day；3Classifier is grader；α_pFor prediction classification penetrating coefficient；θ penetrates for prediction Number round-off error；α_ΔFor predicting penetrating coefficient；It is the geographic latitude in power station；N_pFor prediction day corresponding number of days in a year；1- 2SEM is day ground model；G_p(t) be prediction target day be external radiation intensity curve；F_pT () predicts bent for intensity of solar radiation Line；5Mapping is mapper；K (F) is the intensity of illumination-power relation in power station；P (t) is power station generated energy prediction curve.

Grader 3Classifier is according to statistics radiation penetrating coefficient classification chart to short-range forecast type w_pCarry out point Class draws statistic of classification penetrating coefficient α_p, obtain predicting penetrating coefficient α after adding prediction penetrating coefficient round-off error θ_Δ, it is to be multiplied by The target Japanese external radiation intensity curve G being obtained by day ground model 1-2SEM_pT (), obtains intensity of solar radiation prediction curve F_p (t).Mapper is passed through according to intensity of illumination-power relation K (F) that power station provides, obtains power station generated energy prediction curve P (t).

Claims (8)

1. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast is it is characterised in that include following walking Suddenly：

Step (1), according to day ground model calculate history altitude of the sun change curve sequence { S_n(t) }, according to being that outer solar radiation is normal Number 1353w/m²Calculate intensity of solar radiation Curve Sequences { G outside department of history_n(t)}；Calculate history day measurement solar radiation total amount The ratio of outer solar radiation total amount Japanese with history, i.e. history day radiation penetrating coefficient sequence { α_n}；According to historical city weather Prediction sequence { w_nDay radiation penetrating coefficient is classified, draw the statistics radiation penetrating coefficient classification under each weather typing Table J (w_m)；

Step (2), according to the weather forecast sequence { w of first 14 days_jAnd by statistics radiation penetrating coefficient classification chart J (w_m) can draw The statistic of classification penetrating coefficient sequence of first 14 daysBy the 14 days history penetrating coefficient sequence { α drawing with Practical Calculation_j} Make difference and calculate 14 days penetrating coefficient error sequence { θ_j, it is multiplied by weights sequence { Δ_jSue for peace afterwards, as prediction penetrating coefficient Round-off error θ；

Step (3), prediction the previous day target day forecast the same day weather pattern w_pAccording to statistics radiation penetrating coefficient classification chart J (w_m) obtain prediction classification penetrating coefficient α_p, add prediction classification penetrating coefficient α with round-off error θ_pDraw prediction penetrating coefficient α_Δ；That calculate prediction target day is outer intensity of solar radiation curve G_pIt is multiplied by prediction penetrating coefficient α after (t)_ΔDraw the same day too Positive radiation intensity prediction curve F_p(t)；Intensity of illumination according to power station-power relation K (F), calculates power station generated energy prediction Curve P (t)；

Is that outer intensity of solar radiation curve G (t) computational methods are as follows in step (1) and step (3)：

δ=sin^-1(0.39795cos(0.98563(N-173))) (1)

$G\left(t\right)=\left\{\begin{array}{cc}1353\·S(t-\frac{\mathrm{\π}}{2})& S(t-\frac{\mathrm{\π}}{2})>0\\ 0& S(t-\frac{\mathrm{\π}}{2})\≤0\end{array}\right.---\left(3\right)$

In formula, δ is the solar declination on the same day；N be from annual January 1 start calculate number of days；Geographic latitude for power station；t Hour angle for the sidereal time；1353 is to be outer solar radiation constant；S (t) is altitude of the sun change curve, passes through to transplant and take in formula Value more than 0, this part represents a day daytime part, whenMoment be the sunrise sunset moment.

2. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast according to claim 1, its It is characterised by, in step (1), day radiates penetrating coefficient α_nCalculate function as follows：

${\mathrm{\α}}_n=\frac{{\∫}_{t=0}^{2\mathrm{\π}}{G}_n\left(t\right)}{{\∫}_{t=0}^{2\mathrm{\π}}{F}_n\left(t\right)}---\left(4\right)$

In formula, F_nT () is measurement solar radiation variations curve.

3. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast according to claim 1, its It is characterised by, in step (2), weather forecast w is divided into the weather pattern set of 33 classes according to the definition one of Chinese weather net {w_m, there is statistics radiation penetrating coefficient classification chart J (w_m) be：

$J\left(w_m\right)=\frac{\underset{i\∈\left\{C\right\}}{\Σ}{\mathrm{\α}}_i}{size\left(\right\{C\left\}\right)}---\left(5\right)$

In formula, α_iFor classification results；{ C } is to have w in historical set n_n=w_mSample set；Size ({ C }) is this sample set Size.

4. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast according to claim 1, its It is characterised by, in step (2), 14 days penetrating coefficient error sequence { θ_jBe：

$\left\{{\mathrm{\θ}}_j\right\}=\left\{{\mathrm{\α}}_j\right\}-\left\{{\mathrm{\α}}_j^{*}\right\}---\left(6\right)$

In formula, α_j=J (w_j).

5. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast according to claim 1, its It is characterised by, in step (2), the round-off error θ of prediction penetrating coefficient is：

θ=∑ θ_jΔ_j(7)

In formula, Δ_jFor first 14 days weight coefficients, it is set to:

{Δ_j}={ 0.25,0.2,0.16,0.14,0.1,0.05,0.03,0.02,0.01,0.008,0.008,0 .008, 0.008,0.008}.

6. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast according to claim 1, its It is characterised by, in step (3), thering is prediction penetrating coefficient α_ΔFor：

α_Δ=θ+α_p(8)

In formula, α_p=J (w_p).

7. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast according to claim 1, its It is characterised by, in step (3), thering is intensity of solar radiation prediction curve F_pT () is：

F_p(t)=α_ΔG_p(t). (9)

8. a kind of inexpensive photovoltaic power generation power prediction method based on city weather forecast according to claim 1, its It is characterised by, in step (3), by the intensity of illumination-power relation K (F) in known power station, thering is power station generated energy prediction curve P T () is：

P (t)=K (F_p(t)). (10).