CN107194495A - A kind of longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data - Google Patents
A kind of longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data Download PDFInfo
- Publication number
- CN107194495A CN107194495A CN201710266598.2A CN201710266598A CN107194495A CN 107194495 A CN107194495 A CN 107194495A CN 201710266598 A CN201710266598 A CN 201710266598A CN 107194495 A CN107194495 A CN 107194495A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- photovoltaic power
- value
- day
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention belongs to Solar use research field, more particularly to a kind of longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data, history photovoltaic power data first against Various Seasonal carry out statistical analysis, obtain 6 statistical indicators of photovoltaic power under corresponding season.Above-mentioned statistical indicator is clustered using Euclidean distance, the similar day matrix in the case of Various Seasonal different weather is obtained, passes through one-dimensional Normal Cloud Generator formation similar day typical curve and distributed area.Following 24 hours photovoltaic power longitudinally prediction of corresponding moment is realized using Markov chain theory, the predicted value is merged with similar day typical curve distributed area, new predicted value is formed.Finally new predicted value is weighted with the predicted value that persistence forecasting method is obtained by one-dimensional backward cloud generator and merged, realizes that the photovoltaic power based on the similar cloud fusion of historical data is longitudinally predicted, further improves the precision of prediction of photovoltaic power.
Description
Technical field
It is more particularly to a kind of to be indulged based on the photovoltaic power that historical data is excavated the invention belongs to Solar use research field
To Forecasting Methodology.
Background technology
Compared with water energy, wind energy, geothermal energy, biological energy source etc., solar energy is so that it protrudes exclusive advantage and turns into people's weight
Depending on focus.Abundant solar radiant energy is inexhaustible, nexhaustible, and photovoltaic power generation apparatus noiseless, it is pollution-free, cheap,
Scaleable, it is easy to human freedom, extensively utilization.According to statistics, solar energy reaches the energy on ground each second and is up to 800,000 kilowatts,
If the solar energy of earth surface 0.1% is switched to electric energy, and number turnover is 5%, then can be up to 5.6 × 1,012,000 per annual electricity generating capacity
Watt-hour, equivalent to 40 times of world's total energy consumption.Therefore, photovoltaic generation, which is enjoyed, favors and is used widely.However, multiple by the external world
The influence of miscellaneous uncertain factor, the shortcomings of photovoltaic generation has randomness, fluctuation, intermittence, uncertainty, and photovoltaic are exported
Also nonlinear relation is presented with factor of influence in power, and it is often extremely unstable that this, which results in photovoltaic generation power, to power network
It is economical, safe and stable operation cause serious influence and threat.
Realize the prediction of photovoltaic plant power, it will help dispatching of power netwoks department overall arrangement normal power supplies and photovoltaic generation
Cooperation, in time adjustment operation plan, reasonable arrangement power system operating mode.At present, photovoltaic power prediction mainly can be with
It is divided into statistical method and the class of Artificial Neural Network two.Statistical method be by historical data carry out statistical analysis,
Its inherent law is found out using probability theory and for predicting;And Artificial Neural Network using sample data as input, through machine
Device training test study, sets up forecast model to be predicted future.Both the above method is obtained in terms of photovoltaic power prediction
Application, but its having some limitations property of method are arrived, such as, for regular and periodically stronger data message, both are pre-
Survey method can reach higher precision of prediction, but photovoltaic generation is the features such as have randomness, fluctuation, with both approaches,
Effect is often very poor, it is impossible to meet functional need.Therefore, a kind of method that can carry out photovoltaic power reliable prediction is found to have
There is important practical value.
The content of the invention
In order to solve the above problems, a kind of longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data of the present invention, bag
Include:
Step 1, collection photovoltaic plant power output data, sampling time interval is 15min;Choose under 1 year certain season
D days history photovoltaic power data carry out statistical analysis, obtain exerted oneself under corresponding season average, standard deviation, the coefficient of variation, peak
Degree, the degree of bias and 6 statistical indicators of summation of exerting oneself, and be normalized;
Step 2, using Euclidean distance D days history photovoltaic power data are classified, form n different similar day squares
Battle array c (m), wherein, 1≤m≤n;
Step 3, similar day matrix c (m) is calculated in D days using one-dimensional Normal Cloud Generator from early 7 points to 7 photovoltaic work(of evening
Cloud numerical characteristic (the E of ratex, En, He) formed similar day typical curve and by when water dust distributed area;Wherein, Ex- expect, En-
Entropy, He- super entropy.
Step 4, the M days photovoltaic power datas progress state demarcation that will be predicted under correspondence season day, utilize absolute profile
Markov chain theory calculates transfer frequency matrix and transition probability matrix, selection correspondence t transition probability matrix
Initial state distribution and probability vector are calculated, the probability distribution of prediction time is tried to achieve, sets up and is based on markovian light
The longitudinal forecast model of power is lied prostrate, predicted value is obtained;Wherein,When for certain day, t status i was transferred to the t up to after k days
Carve status j state transition probability matrix;
Step 5, reading weather forecast information determine the similar day typical curve generic of day to be predicted, according to step 3
Obtained similar day typical curve and by when water dust distributed area the predicted value obtained in step 4 is merged, merged
New predicted value afterwards, is weighted by one-dimensional backward cloud generator with the predicted value that persistence forecasting method is obtained and merged, realize base
Longitudinally predicted in the photovoltaic power of the similar cloud fusion of historical data, and then improve photovoltaic power precision of prediction.
Step 6, will predict the outcome is compared with measured value, calculates predicated error.
Statistical indicator in the step 1 includes:
(1) average of exerting oneself PmeanRepresent the Relatively centralized position of the output of power station level in T point time serieses;
PiThe photovoltaic plant output power value under the time point is represented, i=1,2 ..., T, T represents that time of day sequence is total
Data are counted;
(2) standard deviation SPThe size for horizontal degree of variation of being exerted oneself in the common T points time series of expression, was reflected under each time point
The dispersion degree exerted oneself;
(3) coefficient of variation CpTo weigh the statistic for horizontal degree of variation of being exerted oneself in common T points time series, when for two
Or the time series degree of variation under multiple granularities is when comparing, if average value is identical, directly compared with standard deviation;
(4) degree of bias QPSpy for probability distribution density curve in common T points time series relative to the asymmetric degree of average value
Levy the relative length of number, i.e. density function curve afterbody;
(5) kurtosis KPRepresent steep of the overall distribution density curve near its peak value in common T points time series;
(6) summation of exerting oneself PsumRepresent that same day photovoltaic plant is exerted oneself the summation of time series.
Enter row distance cluster to above-mentioned 6 statistical indicators using Euclidean distance in the step 2, obtain the phase of photovoltaic power
Like day matrix c (m), wherein, 1≤m≤n, here n=5;C (1)-fine day, c (2)-fine with occasional clouds, c (3)-cloudy or thunder shower,
C (4)-wet weather, c (5)-bad weather.
In the step 3, the algorithm of one-dimensional Normal Cloud Generator is
Input sizing concept A three numerical characteristic value (Ex, En,He) and water dust number q;
(1) it is E to produce a desired valuex, variance is EnNormal random number xi;
(2) it is E to produce a desired valuen, variance is HeNormal random number E "n;
(3) calculate
(4) (x is madei,yi) for water dust, it is the Linguistic Value that represents of the cloud once implementing quantitatively, wherein
xiFor qualitativing concept in domain this time corresponding numerical value, yiMeasured for the degree that belongs to this Linguistic Value;
(5) repeat step (1) arrives step (4), and the water dust number of number required is met until producing;
(6) similar day typical curve and distributed area are determined;
(7) quantitative values and each water dust for exporting q water dust represent concept A degree of certainty.
The detailed process of the step 4 is
(1) the line number R for obtaining M days photovoltaic power data a under correspondence season prediction day is that number of days represents every with columns N, N
Its sampling number;
(2) photovoltaic power data a maximum P is takenmaxIt is used as upper limit cX, minimum value PminIt is used as lower limit c0, by [c0,cX]
Number X deciles in interval, set up Markovian state's collection S:[c0,c1],[c1,c2],…,[cx-1,cX];Wherein, c0,…,cX∈
S, is X state included in state space S;c(t)It is c for t state in which0,c1,…,cXOne of.
(3) M days synchronization data are carried out with status switch statistics, and calculating obtains transfer frequency matrix fij(k) and turn
Move probability matrix Pij(k);
(4) initial state distribution and probability vector, Conjugative tiansfer probability matrix P are calculatedij(k) prediction time, is tried to achieve
Probability distribution;
P (k)=P0Pk
Wherein, P0For initial state probability vector;PkFor the probability matrix shifted by k next states;P (k) is at k-th
The state probability prediction at moment.
(5) predicted value is extracted, predicted value status space is the corresponding state space of maximum in gained probability distribution,
Predicted value is the average value in the space;
(6) the similar day typical curve generic that weather forecast information determines day to be predicted is read, is obtained according to step 3
Similar day typical curve and distributed area the predicted value obtained in step 4 is merged, the new predicted value after being merged.
The process of one-dimensional backward cloud generator and Weighted Fusion algorithm is in the step 5
1) one-dimensional backward cloud generator algorithm
The quantitative values and each water dust of q water dust of input represent the degree of certainty (x of concepti,yi);
(1) by xiCalculate the sample average of this group of dataSingle order sample Absolute Central MomentSample variance
(2) it must can be expected by (1)
(3) while entropy can be obtained by sample average
(4) it can be obtained by the entropy in the sample variance in (1) and (3)
Export the desired value E for the qualitativing concept A that this q water dust is representedx', entropy En' and super entropy He′;
2) persistence forecasting method:When lead is p, still using current measured value as the predicted value of future time, i.e. P (t+p)
=P (t);
3) Weighted Fusion algorithmic formula:
Wherein,For the photovoltaic power predicted value after fusion, piFor weights,For the standard deviation after fusion.
Predicated error RMS calculation formula in the step 6 are
Wherein, PPrediction(i) it is predicted value, PActual measurement(i) it is measured value.
Beneficial effect
The present invention proposes the longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data, this method by the predicted value with
Similar day typical curve distributed area is merged, and forms new predicted value, and new predicted value and persistence forecasting method are obtained
Predicted value is weighted fusion by one-dimensional backward cloud generator, realizes that the photovoltaic power based on the similar cloud fusion of historical data is indulged
To prediction, the precision of prediction of photovoltaic power is improved.The inventive method is for regular and periodically stronger data message, energy
Higher precision of prediction is reached, the features such as there is randomness, fluctuation for photovoltaic generation, effect of the invention is also fine, it is full
Sufficient functional need.
Brief description of the drawings
Fig. 1 is that the photovoltaic power merged based on similar cloud longitudinally predicts flow chart;
Fig. 2 is similar day typical curve and day part water dust distribution map;
Wherein, Fig. 2 a are similar day typical curve and distributed area;Fig. 2 b are 7 points of similar day typical curve to late 7 clouds
Drip distribution map;Fig. 2 c are 24 hours water dust distribution maps of similar day typical curve;
The one-dimensional reverse cloud model Weighted Fusion block diagrams of Fig. 3;
Fig. 4 is the prediction curve and accuracy comparison figure of the invention with conventional method;Wherein, Fig. 4 a are that photovoltaic power predicts bent
Line;Fig. 4 b are photovoltaic power predicated error curve.
Embodiment
The present invention proposes a kind of longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data below in conjunction with the accompanying drawings, right
Embodiment elaborates.
Fig. 1 is that the photovoltaic power merged based on similar cloud longitudinally predicts flow chart, as shown in the figure:
Step 1, collection photovoltaic plant power output data, sampling time interval 15min.Choose under 1 year certain season
History photovoltaic power data carry out statistical analysis within D days, obtain average, standard deviation, the coefficient of variation, kurtosis, the degree of bias under corresponding season
With 6 statistical indicators such as summation of exerting oneself, and it is normalized.Statistical characteristic value is calculated and method for normalizing is as follows:
(1) average of exerting oneself Pmean:Average of exerting oneself describes the Relatively centralized position of the output of power station level in T point time serieses
Put.
In formula, T represents that time of day sequence total data is counted, PiRepresent the photovoltaic plant power output under the time point
Value.
(2) standard deviation SP:The size for horizontal degree of variation of being exerted oneself in common T points time series is described, each time is reflected
The dispersion degree exerted oneself under point.
(3) coefficient of variation Cp:The statistic for horizontal degree of variation of being exerted oneself in common T points time series is weighed, when for two
Or the time series degree of variation under multiple granularities is when comparing, if average value is identical, directly it can be compared with standard deviation.
(4) degree of bias QP:Probability distribution density curve is relative to the asymmetric journey of average value in the common T points time series of degree of bias sign
The characteristic of degree, intuitively appears to the relative length of density function curve afterbody.
(5) kurtosis KP:It is precipitous near its peak value that kurtosis reflects overall distribution density curve in common T points time series
Degree.
(6) summation of exerting oneself Psum:Summation of exerting oneself reflects same day photovoltaic plant and exerted oneself the summation of time series, due to granularity
Exerted oneself under difference, each granularity significantly different.
(7) above-mentioned clustering target is normalized, its calculation formula is as follows:
Wherein, xminAnd xmaxRespectively the minimum and maximum value of sample array, ymin=-1, ymax=1.At normalization
After reason, data are all fallen within interval [- 1,1].
Step 2, enter using Euclidean distance row distance cluster to above-mentioned 6 statistical indicators, obtain the similar day of photovoltaic power
Matrix c (m), wherein, 1≤m≤n, here n=5.C (1)-fine day, c (2)-fine with occasional clouds, c (3)-cloudy or thunder shower, c
(4)-overcast and rainy, 5 kinds of weather conditions such as c (5)-bad weather.Calculation formula is as follows:
Step 3, similar day matrix c (k) is calculated in D days using one-dimensional Normal Cloud Generator from early 7 points to 7 photovoltaic work(of evening
Cloud numerical characteristic (the E of ratex, En,He), formed similar day typical curve and by when water dust distributed area.As shown in Fig. 2 Fig. 2 a are
Similar day typical curve and distributed area figure, 2b be 7 points of similar day typical curve to late 7 water dust distribution maps, Fig. 2 c are similar
Day 24 hours water dust distribution maps of typical curve.
The algorithm of one-dimensional Normal Cloud Generator is realized as follows:
Input:Represent sizing concept A three numerical characteristic value (Ex, En,He) and water dust number q;
Output:The quantitative values of q water dust, and each water dust represent concept A degree of certainty;
(1) it is E to produce a desired valuex, variance is EnNormal random number xi;
(2) it is E to produce a desired valuen, variance is HeNormal random number E "n;
(3) calculate:
(4) (x is madei,yi) for water dust, it is the Linguistic Value that represents of the cloud once implementing quantitatively, wherein
xiFor qualitativing concept in domain this time corresponding numerical value, yiMeasured for the degree that belongs to this Linguistic Value;
(5) repeat step (1) arrives step (4), and the water dust number of number required is met until producing.
(6) similar day typical curve and distributed area are determined.
Step 4, the M days photovoltaic power datas progress state demarcation that will be predicted under correspondence season day, utilize absolute profile
Markov chain theory calculates transfer frequency matrix and transition probability matrix, selection correspondence t transition probability matrix
Initial state distribution and probability vector are calculated, the probability distribution of prediction time is tried to achieve, sets up and is based on markovian light
Lie prostrate the longitudinal forecast model of power.Wherein,The t status j up to after k days is transferred to for certain day t status i
State transition probability matrix.Longitudinally prediction process is as follows for Markov Chain based on absolute profile:
(1) line number R (number of days) and the columns N for obtaining M days photovoltaic power data a under correspondence season prediction day (is adopted daily
Number of samples);
(2) photovoltaic power data a maximum P is takenmaxIt is used as upper limit cX, minimum value PminIt is used as lower limit c0, by [c0,cX]
Number X deciles in interval, set up Markovian state's collection S:[c0,c1],[c1,c2],…,[cx-1,cX];Wherein, c0,…,cX∈
S, is X state included in state space S;c(t)It is c for t state in which0,c1,…,cXOne of.
(3) M days synchronization data are carried out with status switch statistics, and calculating obtains transfer frequency matrix fij(k) and turn
Move probability matrix Pij(k);
(4) initial state distribution and probability vector, Conjugative tiansfer probability matrix P are calculatedij(k), carry out as the following formula pre-
Survey and calculate, try to achieve the probability distribution of prediction time;
P (k)=P0Pk
Wherein, P0For initial state probability vector;PkFor the probability matrix shifted by k next states;P (k) is at k-th
The state probability prediction at moment.
(5) predicted value is extracted, and predicted value status space is the corresponding state space of maximum in gained probability distribution,
Predicted value is the average value in the space.
(6) the similar day typical curve generic that weather forecast information determines day to be predicted is read, is obtained according to step 3
Similar day typical curve and distributed area the predicted value obtained in step 4 is merged, the new predicted value after being merged.
It is similar that step 5, predicted value and step 3 that the predicted value obtained in step 4, persistence forecasting method are obtained are obtained
Day typical curve and distributed area, fusion is weighted by one-dimensional backward cloud generator, realizes the light based on the fusion of similar cloud
Lie prostrate power longitudinally to predict, and then improve photovoltaic power precision of prediction.One-dimensional backward cloud generator and Weighted Fusion algorithm is realized such as
Under:
1) one-dimensional backward cloud generator algorithm is realized as follows:
Input:The quantitative values and each water dust of q water dust represent the degree of certainty (x of concepti,yi);
Output:The desired value E for the qualitativing concept A that this q water dust is representedx', entropy En' and super entropy He′;
(1) by xiCalculate the sample average of this group of dataSingle order sample Absolute Central MomentSample variance
(2) it must can be expected by (1)
(3) while entropy can be obtained by sample average
(4) it can be obtained by the entropy in the sample variance in (1) and (3)
2) persistence forecasting method:When lead is p, still using current measured value as the predicted value of future time, i.e. P (t+p)
=P (t).
3) as shown in figure 3, Weighted Fusion algorithmic formula:
Wherein, ExFor the photovoltaic power predicted value after fusion, piFor weights, EnFor the standard deviation after fusion.
Step 6:Calculate relative average error formula:
Wherein, PPrediction(i) it is predicted value, PActual measurement(i) it is measured value.
As shown in figure 4, longitudinally being predicted based on the Markov Chain photovoltaic power that similar cloud is merged for the present invention is respectively adopted
Method is entered with the longitudinal Forecasting Methodology of existing Markov Chain photovoltaic power to certain photovoltaic plant June and photovoltaic power data in July
Row prediction, Fig. 4 a are photovoltaic power prediction curve, and Fig. 4 b are that photovoltaic power predicated error curve prediction result is shown:The present invention's
Prediction relative average error obtained by method is significantly less than existing method.
Claims (7)
1. a kind of longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data, it is characterised in that including:
Step 1, collection photovoltaic plant power output data, the D days history photovoltaic power data chosen under upper 1 year certain season are entered
Row statistical analysis, obtains exerted oneself under corresponding season average, standard deviation, the coefficient of variation, kurtosis, the degree of bias and 6 statistics of summation of exerting oneself
Index, and be normalized;
Step 2, using Euclidean distance D days history photovoltaic power data are classified, form n different similar day matrix c
(m), wherein, 1≤m≤n;
Step 3, similar day matrix c (m) is calculated in D days using one-dimensional Normal Cloud Generator from early 7 points to late 7 photovoltaic powers
Cloud numerical characteristic (Ex, En,He) formed similar day typical curve and by when water dust distributed area;Wherein, Ex- expect, En- entropy, He-
Super entropy;
Step 4, the M days photovoltaic power datas progress state demarcation that will be predicted under correspondence season day, utilize the Ma Er of absolute profile
Can husband's chain theoretical calculation transfer frequency matrix and transition probability matrix, selection correspondence t transition probability matrixCalculate
Initial state distribution and probability vector, try to achieve the probability distribution of prediction time, set up and be based on markovian photovoltaic work(
Rate longitudinal direction forecast model, obtains predicted value;Wherein,It is transferred to for certain day t status i up to after k days residing for t
State j state transition probability matrix;
Step 5, reading weather forecast information determine the similar day typical curve generic of day to be predicted, are obtained according to step 3
Similar day typical curve and by when water dust distributed area the predicted value obtained in step 4 is merged, after being merged
New predicted value, is weighted by one-dimensional backward cloud generator with the predicted value that persistence forecasting method is obtained and merged, and is realized and is based on going through
The photovoltaic power of the similar cloud fusion of history data is longitudinally predicted, and then improves photovoltaic power precision of prediction;
Step 6, will predict the outcome is compared with measured value, calculates predicated error.
2. method according to claim 1, it is characterised in that the statistical indicator in the step 1 includes:
(1) average of exerting oneself PmeanRepresent the Relatively centralized position of the output of power station level in T point time serieses;
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mi>m</mi>
<mi>e</mi>
<mi>a</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>T</mi>
</munderover>
<msub>
<mi>P</mi>
<mi>i</mi>
</msub>
</mrow>
PiThe photovoltaic plant output power value under the time point is represented, i=1,2 ..., T, T represents the total strong point of time of day sequence
Number;
(2) standard deviation SPThe size for horizontal degree of variation of being exerted oneself in the common T points time series of expression, reflects and is exerted oneself under each time point
Dispersion degree;
<mrow>
<msub>
<mi>S</mi>
<mi>P</mi>
</msub>
<mo>=</mo>
<msqrt>
<mrow>
<mfrac>
<mn>1</mn>
<mi>T</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>T</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>P</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>P</mi>
<mrow>
<mi>m</mi>
<mi>e</mi>
<mi>a</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
(3) coefficient of variation CpTo weigh the statistic for horizontal degree of variation of being exerted oneself in common T points time series, when for two or many
When time series degree of variation under individual granularity compares, if average value is identical, directly compared with standard deviation;
<mrow>
<msub>
<mi>C</mi>
<mi>p</mi>
</msub>
<mo>=</mo>
<mfrac>
<msub>
<mi>S</mi>
<mi>P</mi>
</msub>
<msub>
<mi>P</mi>
<mrow>
<mi>m</mi>
<mi>e</mi>
<mi>a</mi>
<mi>n</mi>
</mrow>
</msub>
</mfrac>
</mrow>
(4) degree of bias QPCharacteristic for probability distribution density curve in common T points time series relative to the asymmetric degree of average value,
That is the relative length of density function curve afterbody;
<mrow>
<msub>
<mi>Q</mi>
<mi>P</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>T&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>T</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>P</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>P</mi>
<mrow>
<mi>m</mi>
<mi>e</mi>
<mi>a</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>3</mn>
</msup>
</mrow>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
<mo>(</mo>
<mi>T</mi>
<mo>-</mo>
<mn>2</mn>
<mo>)</mo>
<msub>
<mi>S</mi>
<mi>P</mi>
</msub>
</mrow>
</mfrac>
</mrow>
(5) kurtosis KPRepresent steep of the overall distribution density curve near its peak value in common T points time series;
<mrow>
<msub>
<mi>K</mi>
<mi>P</mi>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>T</mi>
</msubsup>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>P</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>P</mi>
<mrow>
<mi>m</mi>
<mi>e</mi>
<mi>a</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>4</mn>
</msup>
</mrow>
<mrow>
<mo>(</mo>
<mi>T</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
<msub>
<mi>S</mi>
<mi>P</mi>
</msub>
</mrow>
</mfrac>
</mrow>
(6) summation of exerting oneself PsumRepresent that same day photovoltaic plant is exerted oneself the summation of time series.
3. method according to claim 1, it is characterised in that referred in the step 2 using Euclidean distance to above-mentioned 6 statistics
Mark into row distance cluster, obtain the similar day matrix c (m) of photovoltaic power, wherein, 1≤m≤n, here n=5;C (1)-fine day, c
(2)-fine with occasional clouds, c (3)-cloudy or thunder shower, c (4)-wet weather, c (5)-bad weather.
4. method according to claim 1, it is characterised in that in the step 3, the algorithm of one-dimensional Normal Cloud Generator is
Input sizing concept A three numerical characteristic value (Ex, En,He) and water dust number q;
(1) it is E to produce a desired valuex, variance is EnNormal random number xi;
(2) it is E to produce a desired valuen, variance is HeNormal random number E "n;
(3) calculate
(4) (x is madei,yi) for water dust, it is the Linguistic Value that represents of the cloud once implementing quantitatively, wherein xiFor
Qualitativing concept this time corresponding numerical value, y in domainiMeasured for the degree that belongs to this Linguistic Value;
(5) repeat step (1) arrives step (4), and the water dust number of number required is met until producing;
(6) similar day typical curve and distributed area are determined;
(7) quantitative values and each water dust for exporting q water dust represent concept A degree of certainty.
5. method according to claim 1, it is characterised in that the detailed process of the step 4 is
(1) the line number R for obtaining M days photovoltaic power data a under correspondence season prediction day is that number of days represents to adopt daily with columns N, N
Number of samples;
(2) photovoltaic power data a maximum P is takenmaxIt is used as upper limit cX, minimum value PminIt is used as lower limit c0, by [c0,cX] interval interior
Number X deciles, set up Markovian state collection S:[c0,c1],[c1,c2],…,[cx-1,cX];Wherein, c0,…,cX∈ S, are shape
X state included in state space S;c(t)It is c for t state in which0,c1,…,cXOne of;
(3) M days synchronization data are carried out with status switch statistics, and calculating obtains transfer frequency matrix fij(k) it is general with transfer
Rate matrix Pij(k);
(4) initial state distribution and probability vector, Conjugative tiansfer probability matrix P are calculatedij(k) the general of prediction time, is tried to achieve
Rate is distributed;
P (k)=P0Pk
Wherein, P0For initial state probability vector;PkFor the probability matrix shifted by k next states;P (k) is k-th of moment
State probability prediction;
(5) predicted value is extracted, predicted value status space is the corresponding state space of maximum in gained probability distribution, prediction
It is worth the average value for the space;
(6) the similar day typical curve generic that weather forecast information determines day to be predicted, the phase obtained according to step 3 are read
The predicted value obtained in step 4 is merged like day typical curve and distributed area, the new predicted value after being merged.
6. method according to claim 1, it is characterised in that one-dimensional backward cloud generator and Weighted Fusion in the step 5
The process of algorithm is
1) one-dimensional backward cloud generator algorithm
The quantitative values and each water dust of q water dust of input represent the degree of certainty (x of concepti,yi);
(1) by xiCalculate the sample average of this group of dataSingle order sample Absolute Central MomentSample variance
(2) it must can be expected by (1)
(3) while entropy can be obtained by sample average
(4) it can be obtained by the entropy in the sample variance in (1) and (3)
Export the desired value E for the qualitativing concept A that this q water dust is representedx', entropy En' and super entropy He′;
2) persistence forecasting method:When lead is p, still using current measured value as the predicted value of future time, i.e. P (t+p)=P
(t);
3) Weighted Fusion algorithmic formula:
<mrow>
<msub>
<mi>E</mi>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>p</mi>
</msub>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</msubsup>
<msub>
<mi>p</mi>
<mi>i</mi>
</msub>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>i</mi>
</msub>
</mrow>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</msubsup>
<msub>
<mi>p</mi>
<mi>i</mi>
</msub>
</mrow>
</mfrac>
</mrow>
<mrow>
<msub>
<mi>p</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<msup>
<mrow>
<mo>(</mo>
<msubsup>
<mi>E</mi>
<mrow>
<mi>n</mi>
<mi>i</mi>
</mrow>
<mo>&prime;</mo>
</msubsup>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mfrac>
</mrow>
<mrow>
<msub>
<mi>E</mi>
<mrow>
<mi>n</mi>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>p</mi>
</msub>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<msqrt>
<mrow>
<msubsup>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>l</mi>
</msubsup>
<msub>
<mi>p</mi>
<mi>i</mi>
</msub>
</mrow>
</msqrt>
</mfrac>
</mrow>
Wherein,For the photovoltaic power predicted value after fusion, piFor weights,For the standard deviation after fusion.
7. method according to claim 1, it is characterised in that the predicated error RMS calculation formula in the step 6 are
Wherein, PPrediction(i) it is predicted value, PActual measurement(i) it is measured value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710266598.2A CN107194495B (en) | 2017-04-21 | 2017-04-21 | Photovoltaic power longitudinal prediction method based on historical data mining |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710266598.2A CN107194495B (en) | 2017-04-21 | 2017-04-21 | Photovoltaic power longitudinal prediction method based on historical data mining |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107194495A true CN107194495A (en) | 2017-09-22 |
CN107194495B CN107194495B (en) | 2020-05-12 |
Family
ID=59872298
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710266598.2A Expired - Fee Related CN107194495B (en) | 2017-04-21 | 2017-04-21 | Photovoltaic power longitudinal prediction method based on historical data mining |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107194495B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108900622A (en) * | 2018-07-10 | 2018-11-27 | 广州智能装备研究院有限公司 | Data fusion method, device and computer readable storage medium based on Internet of Things |
CN108921339A (en) * | 2018-06-22 | 2018-11-30 | 南京工程学院 | Genetic Support Vector Machine photovoltaic power interval prediction method based on quantile estimate |
CN109038646A (en) * | 2018-06-29 | 2018-12-18 | 中国电力科学研究院有限公司 | A kind of photovoltaic power generation scene build method and system |
CN109086928A (en) * | 2018-07-27 | 2018-12-25 | 福州大学 | Photovoltaic plant realtime power prediction technique based on SAGA-FCM-LSSVM model |
CN110110912A (en) * | 2019-04-26 | 2019-08-09 | 华北电力大学 | A kind of photovoltaic power multi-model interval prediction method |
CN110265996A (en) * | 2019-02-26 | 2019-09-20 | 国网吉林省电力有限公司 | It is a kind of suitable for photovoltaic/wind power prediction time characteristic scale modeling method |
CN111240282A (en) * | 2019-12-31 | 2020-06-05 | 联想(北京)有限公司 | Process optimization method, device, equipment and computer readable storage medium |
CN112734125A (en) * | 2021-01-15 | 2021-04-30 | 国网山西省电力公司晋城供电公司 | Photovoltaic output prediction method and device and electronic equipment |
CN114881341A (en) * | 2022-05-19 | 2022-08-09 | 中山大学 | Photovoltaic power prediction method and system based on reverse prediction historical data set |
CN115511220A (en) * | 2022-11-02 | 2022-12-23 | 河海大学 | Ultra-short-term solar radiation prediction method and system based on cross-mode attention mechanism |
CN116722544A (en) * | 2023-08-02 | 2023-09-08 | 北京弘象科技有限公司 | Distributed photovoltaic short-term prediction method and device, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103489038A (en) * | 2013-09-09 | 2014-01-01 | 国家电网公司 | Photovoltaic ultra-short-term power prediction method based on LM-BP neural network |
CN104616078A (en) * | 2015-02-03 | 2015-05-13 | 河海大学 | Spiking neural network based photovoltaic system generation power prediction method |
CN105139080A (en) * | 2015-08-04 | 2015-12-09 | 国家电网公司 | Improved photovoltaic power sequence prediction method based on Markov chain |
CN105260618A (en) * | 2015-11-04 | 2016-01-20 | 华北电力大学 | Cloud model-based energy storage system typical-curve mining method |
CN106529814A (en) * | 2016-11-21 | 2017-03-22 | 武汉大学 | Distributed photovoltaic ultra-short-term forecasting method based on Adaboost clustering and Markov chain |
-
2017
- 2017-04-21 CN CN201710266598.2A patent/CN107194495B/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103489038A (en) * | 2013-09-09 | 2014-01-01 | 国家电网公司 | Photovoltaic ultra-short-term power prediction method based on LM-BP neural network |
CN104616078A (en) * | 2015-02-03 | 2015-05-13 | 河海大学 | Spiking neural network based photovoltaic system generation power prediction method |
CN105139080A (en) * | 2015-08-04 | 2015-12-09 | 国家电网公司 | Improved photovoltaic power sequence prediction method based on Markov chain |
CN105260618A (en) * | 2015-11-04 | 2016-01-20 | 华北电力大学 | Cloud model-based energy storage system typical-curve mining method |
CN106529814A (en) * | 2016-11-21 | 2017-03-22 | 武汉大学 | Distributed photovoltaic ultra-short-term forecasting method based on Adaboost clustering and Markov chain |
Non-Patent Citations (3)
Title |
---|
YING-ZI LI等: "Forecast of Power Generation for Grid-Connected Photovoltaic System Based on Markov Chain", 《2009 ASIA-PACIFIC POWER AND ENERGY ENGINEERING CONFERENCE》 * |
尹星露: "基于预测有效度和马尔科夫- 云模型的母线负荷预测模型筛选与变权重组合预测", 《电力自动化设备》 * |
李燕斌等: "灰色-马尔可夫链统计组合模型在光伏发电功率短期预测中的应用", 《电测与仪表》 * |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108921339B (en) * | 2018-06-22 | 2021-08-10 | 南京工程学院 | Quantile regression-based photovoltaic power interval prediction method for genetic support vector machine |
CN108921339A (en) * | 2018-06-22 | 2018-11-30 | 南京工程学院 | Genetic Support Vector Machine photovoltaic power interval prediction method based on quantile estimate |
CN109038646A (en) * | 2018-06-29 | 2018-12-18 | 中国电力科学研究院有限公司 | A kind of photovoltaic power generation scene build method and system |
CN109038646B (en) * | 2018-06-29 | 2022-11-25 | 中国电力科学研究院有限公司 | Photovoltaic power generation scene construction method and system |
CN108900622A (en) * | 2018-07-10 | 2018-11-27 | 广州智能装备研究院有限公司 | Data fusion method, device and computer readable storage medium based on Internet of Things |
CN108900622B (en) * | 2018-07-10 | 2021-04-09 | 广州智能装备研究院有限公司 | Data fusion method and device based on Internet of things and computer readable storage medium |
CN109086928A (en) * | 2018-07-27 | 2018-12-25 | 福州大学 | Photovoltaic plant realtime power prediction technique based on SAGA-FCM-LSSVM model |
CN109086928B (en) * | 2018-07-27 | 2021-11-30 | 福州大学 | Photovoltaic power station real-time power prediction method based on SAGA-FCM-LSSVM model |
CN110265996A (en) * | 2019-02-26 | 2019-09-20 | 国网吉林省电力有限公司 | It is a kind of suitable for photovoltaic/wind power prediction time characteristic scale modeling method |
CN110265996B (en) * | 2019-02-26 | 2023-04-07 | 国网吉林省电力有限公司 | Time characteristic scale modeling method suitable for photovoltaic/wind power prediction |
CN110110912A (en) * | 2019-04-26 | 2019-08-09 | 华北电力大学 | A kind of photovoltaic power multi-model interval prediction method |
CN111240282A (en) * | 2019-12-31 | 2020-06-05 | 联想(北京)有限公司 | Process optimization method, device, equipment and computer readable storage medium |
CN111240282B (en) * | 2019-12-31 | 2021-12-24 | 联想(北京)有限公司 | Process optimization method, device, equipment and computer readable storage medium |
CN112734125A (en) * | 2021-01-15 | 2021-04-30 | 国网山西省电力公司晋城供电公司 | Photovoltaic output prediction method and device and electronic equipment |
CN114881341A (en) * | 2022-05-19 | 2022-08-09 | 中山大学 | Photovoltaic power prediction method and system based on reverse prediction historical data set |
CN114881341B (en) * | 2022-05-19 | 2023-04-07 | 中山大学 | Photovoltaic power prediction method and system based on reverse prediction historical data set |
CN115511220A (en) * | 2022-11-02 | 2022-12-23 | 河海大学 | Ultra-short-term solar radiation prediction method and system based on cross-mode attention mechanism |
CN115511220B (en) * | 2022-11-02 | 2023-06-13 | 河海大学 | Ultra-short-term solar radiation prediction method and system based on cross-modal attention mechanism |
CN116722544A (en) * | 2023-08-02 | 2023-09-08 | 北京弘象科技有限公司 | Distributed photovoltaic short-term prediction method and device, electronic equipment and storage medium |
CN116722544B (en) * | 2023-08-02 | 2023-10-20 | 北京弘象科技有限公司 | Distributed photovoltaic short-term prediction method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN107194495B (en) | 2020-05-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107194495A (en) | A kind of longitudinal Forecasting Methodology of photovoltaic power excavated based on historical data | |
CN106295899B (en) | Wind power probability density Forecasting Methodology based on genetic algorithm Yu supporting vector quantile estimate | |
CN105046374B (en) | A kind of power interval prediction technique based on core extreme learning machine model | |
CN109902874A (en) | A kind of micro-capacitance sensor photovoltaic power generation short term prediction method based on deep learning | |
CN108022001A (en) | Short term probability density Forecasting Methodology based on PCA and quantile estimate forest | |
CN109858673A (en) | A kind of photovoltaic generating system power forecasting method | |
CN107528350B (en) | A kind of wind power output typical scene generation method adapting to long -- term generation expansion planning | |
CN104463349A (en) | Photovoltaic generated power prediction method based on multi-period comprehensive similar days | |
CN105654207A (en) | Wind power prediction method based on wind speed information and wind direction information | |
CN103218674A (en) | Method for predicating output power of photovoltaic power generation system based on BP (Back Propagation) neural network model | |
CN104978611A (en) | Neural network photovoltaic power generation output prediction method based on grey correlation analysis | |
CN102102626A (en) | Method for forecasting short-term power in wind power station | |
CN102938562B (en) | Prediction method of total wind electricity power in area | |
CN107516145A (en) | A kind of multichannel photovoltaic power generation output forecasting method based on weighted euclidean distance pattern classification | |
CN104463356A (en) | Photovoltaic power generation power prediction method based on multi-dimension information artificial neural network algorithm | |
CN106875033A (en) | A kind of wind-powered electricity generation cluster power forecasting method based on dynamic self-adapting | |
CN103400204A (en) | Forecasting method for solar photovoltaic electricity generation amount based on SVM (support vector machine) - Markov combination method | |
CN103683274A (en) | Regional long-term wind power generation capacity probability prediction method | |
CN109934395A (en) | A kind of more small power station areas Short-Term Load Forecasting Method based on timesharing subregion meteorological data | |
CN107679687A (en) | A kind of photovoltaic output modeling method and Generation System Reliability appraisal procedure | |
CN106295857A (en) | A kind of ultrashort-term wind power prediction method | |
CN104820877A (en) | Photovoltaic system generation power prediction method based on cloud adaptive PSO-SNN | |
CN105846425A (en) | Economic dispatching method based on general wind power forecasting error model | |
CN110070228A (en) | BP neural network wind speed prediction method for neuron branch evolution | |
CN105303268A (en) | Wind power generation output power prediction method based on similarity theory |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200512 |
|
CF01 | Termination of patent right due to non-payment of annual fee |