CN109767353B - Photovoltaic power generation power prediction method based on probability distribution function - Google Patents
Photovoltaic power generation power prediction method based on probability distribution function Download PDFInfo
- Publication number
- CN109767353B CN109767353B CN201910031330.XA CN201910031330A CN109767353B CN 109767353 B CN109767353 B CN 109767353B CN 201910031330 A CN201910031330 A CN 201910031330A CN 109767353 B CN109767353 B CN 109767353B
- Authority
- CN
- China
- Prior art keywords
- power generation
- photovoltaic
- photovoltaic power
- generation power
- irradiance
- 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.)
- Active
Links
Images
Classifications
-
- 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
Landscapes
- Photovoltaic Devices (AREA)
Abstract
The invention relates to a photovoltaic power generation power prediction method based on a probability distribution function, which comprises the following steps: step 1: irradiance data of a photovoltaic power generation system to be predicted and corresponding power generation power data are collected; step 2: dividing the value range of irradiance into M intervals; in each interval, fitting by adopting a general distribution fitting method to obtain a photovoltaic power generation power probability density function and an accumulated distribution function corresponding to the interval; and step 3: for any known irradiance, finding a corresponding interval and a photovoltaic power generation power probability distribution density function to obtain a plurality of photovoltaic power generation power values corresponding to the known irradiance, wherein each photovoltaic power generation power value forms a photovoltaic output scene set; and 4, step 4: and calculating the average value of all photovoltaic power generation power values in the photovoltaic output scene set as a photovoltaic power generation power predicted value corresponding to the known irradiance. The method can predict the photovoltaic power generation power quickly and accurately, and has high practical application value.
Description
Technical Field
The invention belongs to the field of photovoltaic power generation, and particularly relates to a method for predicting power of a photovoltaic power generation system.
Background
The photovoltaic power generation has the advantages of no pollutant discharge, no fuel consumption, no limit on capacity scale, flexible application form, safety, reliability, simple maintenance and the like, thereby having wide development prospect. Due to the support of policies of countries around the world, photovoltaic power generation has been rapidly developed in recent years, and has been applied to a larger scale on a global scale. According to the newly published statistical data of the German solar energy society, 70GW of the global photovoltaic new installation machine is increased by about 30% in 2016 compared with 2015, and the total amount of the global photovoltaic installation machine reaches 300 GW.
A photovoltaic power generation system is a power generation system that converts solar energy into electric energy using a photovoltaic effect. The photovoltaic module comprises a photovoltaic module, a storage battery, a controller, an inverter and a bracket system. Photovoltaic power generation systems also typically incorporate small meteorological devices and some monitoring systems to measure and record meteorological and system performance data. The photovoltaic power generation system is basically as shown in figure 1.
The prediction of the generated power of the photovoltaic power generation system is an important work in the process of applying the photovoltaic power generation system, and in the prediction process, a photovoltaic power generation power model is required to be utilized, namely under the condition of a given set of meteorological prediction information, the photovoltaic power generation power model can be used for predicting the output power of the photovoltaic power generation system, and the prediction is an indispensable link of the photovoltaic power generation power. In addition, the photovoltaic power generation power model may be used to compare the expected output of a particular system to other possible system designs prior to purchasing the equipment. Finally, the photovoltaic power generation power model may also be used to determine if the system is operating on schedule, enabling the system operator to determine a maintenance schedule. Therefore, the research on the photovoltaic power generation power model has certain practical guiding significance, and the determination of the photovoltaic power generation model directly influences the effect generated after the photovoltaic power generation model is applied.
At present, the modeling methods of photovoltaic power generation power models at home and abroad mainly comprise a physical model and a statistical model. Physical modeling focuses on describing mathematical relationships between energy conversion devices (photovoltaic cells, controllers, inverters and the like) and operation control system modules in the process of converting solar energy into electric energy, and the effectiveness of the method depends on the degree of understanding of the internal composition of an object to be researched and the law followed by the internal composition and the accuracy of model parameters. The most important link influencing the accuracy of the physical model is the photovoltaic cell model, and the most common method is to establish a diode-based photovoltaic cell equivalent circuit model. The earliest model started from short circuit, open circuit voltage and diode ideality factor and proposed a three-parameter model consisting of a linear independent current source and a parallel diode. The method is simple and easy to operate, but has low precision. Furthermore, the literature proposes an improvement of the three-parameter model by means of a series resistance Rs, which is widely referred to as the Rs model, which is the most widely used model so far, but whose calculation accuracy is greatly reduced at higher temperatures. In this regard, the related literature adds an additional parallel resistor Rp on the basis of the Rs model, and extends the additional parallel resistor Rp into the Rp model. Although the accuracy of the model is improved, the parameter setting is increasingly difficult and the calculation amount is large along with the increasing of model parameters. Therefore, on the basis of the Rp five-parameter model, a simplified photovoltaic cell engineering model is provided, and the output characteristic of the photovoltaic cell can be obtained by the model only by four basic parameters provided by a photovoltaic cell manufacturer. The engineering model reduces the calculated amount of the model to a certain extent, but practice shows that the accuracy of the physical model is lower under the condition of low irradiance due to the influence of the efficiency of a photovoltaic power generation system.
Different from a physical model, the statistical model focuses more on the mathematical statistical law between the input and the output of the photovoltaic power generation system, and the analysis of the characteristics of each internal module and each internal influence factor is faded. The photovoltaic output characteristics are fitted based on actual operating data, and nonlinear mapping between input and output is reflected. Common statistical model modeling methods include neural networks, associated data, and the like: the literature adopts a neural network to evaluate the performance of the photovoltaic array, further utilizes a recurrent neural network to establish an output model of the photovoltaic power generation system, compares the model with a physical model modeling method, and indicates that the computational accuracy of the recurrent neural network model is superior to that of the physical model. Although the neural network method has high precision, the number of hidden layer neurons is large, and the determination of the optimal number of hidden layer neurons and the optimal synaptic weights requires a complex algorithm and a great deal of research work.
Therefore, the problems of low accuracy or complex algorithm and large workload of the existing photovoltaic power generation power prediction are caused based on the defects of the existing photovoltaic power generation power model.
Disclosure of Invention
The invention aims to provide a photovoltaic power generation power prediction method based on a probability distribution function, which has higher precision, is simpler and is easy to implement.
In order to achieve the purpose, the invention adopts the technical scheme that:
a photovoltaic power generation power prediction method based on a probability distribution function comprises the following steps:
step 1: irradiance data of a photovoltaic power generation system to be predicted and corresponding power generation power data are collected in historical data, and data preprocessing is carried out;
step 2: dividing the value range of irradiance into M intervals; in each interval, a general distribution fitting method is adopted to fit to obtain a photovoltaic power generation power probability density function and an accumulative distribution function corresponding to the interval, and the photovoltaic power generation power probability distribution density function corresponding to each interval forms a general distribution probability density function library;
and step 3: for any known irradiance, finding the corresponding interval and the photovoltaic power generation power probability distribution density function, performing inverse transformation sampling on the photovoltaic power generation power probability distribution density function to obtain a plurality of photovoltaic power generation power values corresponding to the known irradiance, wherein each photovoltaic power generation power value forms a photovoltaic output scene set;
and 4, step 4: and calculating the average value of all the photovoltaic power generation power values in the photovoltaic output scene set as the photovoltaic power generation power predicted value corresponding to the known irradiance.
In step 1, the data preprocessing includes deleting error data, eliminating useless data, and performing per unit on historical data.
In the step 2, the value range of irradiance is 0-1 p.u. and is equally divided into M intervals, and the width of each interval is 1/M of the value range of irradiance.
In the step 3, the value range of irradiance is divided into M intervals with different widths, and the power generation power data corresponding to each interval form a single-peak state.
In the step 2, the probability density function expression corresponding to the general distribution fitting method is
The cumulative distribution function expression is f (x) ═ 1+ e-α(x-γ))-βWherein alpha, beta and gamma are shape parameters respectively, and satisfy alpha > 0, beta > 0, - ∞ < gamma < + >.
In step 3, the method for inverse transforming samples comprises: introducing a random variable ZtBy using
Solving for the random variable ZtCorresponding standard normal distribution function value phi (Z)t) Reuse of Pt＝Fl-1(Φ(Zt) The value of the standard normal distribution function phi (Z) is obtainedt) Corresponding photovoltaic power generation power value, wherein Fl-1(U) is a cumulative distribution function FlThe inverse function of (c).
In the step 4, a weighted average value is obtained for each photovoltaic power generation power value in the photovoltaic output scene set and is used as a photovoltaic power generation power predicted value corresponding to the known irradiance.
And in the step 2, introducing a variable temperature correction coefficient to correct the photovoltaic power generation power probability density function and the accumulative distribution function.
The expression of the temperature correction coefficient is alpha ═ (P)t cal-Pt actual)/(Tt-Tt-1) Where α is the temperature correction coefficient, t is the time, Pt calIs a predicted value of photovoltaic power generation power at the time t, Pt actualIs the actual value of the photovoltaic output at the moment T, TtTemperature at time T, Tt-1Is the temperature at time t-1.
And for each interval, obtaining a plurality of temperature correction basic coefficients at different moments and under different irradiance conditions by taking the minimum root-mean-square error of the photovoltaic power generation power as a target, and taking the average value of the temperature correction basic coefficients as the temperature correction coefficient of the interval.
Due to the application of the technical scheme, compared with the prior art, the invention has the following advantages: the method can predict the photovoltaic power generation power quickly and accurately, and has high practical application value.
Drawings
FIG. 1 is a schematic diagram of the basic components of a photovoltaic power generation system.
FIG. 2 is a comparison graph of the fit of the generic, Gaussian and Weibull distributions.
Fig. 3 is an inverse transform sampling flow diagram.
FIG. 4 is a flow chart of a photovoltaic power generation power prediction method based on a probability distribution function.
FIG. 5 is a schematic diagram showing the large dispersion and the bimodal phenomenon of the historical data.
FIG. 6 is a schematic diagram of fitting effect after uneven binning improvement.
FIG. 7 is a diagram showing the calculation results of the models.
Detailed Description
The invention will be further described with reference to examples of embodiments shown in the drawings to which the invention is attached.
The first embodiment is as follows: a photovoltaic power generation power prediction method based on a probability distribution function comprises the following steps:
step 1: and collecting irradiance data of the photovoltaic power generation system to be predicted and corresponding power generation power data in historical data, and preprocessing the data.
Since the obtained raw data contains much and complicated information, the obtained historical data needs to be preprocessed first. The data preprocessing comprises error data deletion, useless data elimination and historical data per unit, and specifically comprises the following steps:
a) and extracting required data. Desired data, i.e., irradiance data, temperature data, and photovoltaic output data (i.e., generated power data) are extracted from the historical data.
b) The error data is deleted. Firstly, deleting error data with irradiance obviously not corresponding to photovoltaic output; secondly, due to measurement errors near sunrise and sunset times, some irradiance is negative, and these data are also erroneous data and should be deleted.
c) And eliminating useless data. Irradiance before sunrise and irradiance after sunrise is zero, photovoltaic output is also zero, and for establishing a photovoltaic output model, the data are useless data and should be removed.
d) And (5) per unit of historical data. After the error data and the useless data are eliminated, all historical data are converted into per unit values by taking the maximum irradiance value and the maximum photovoltaic output as basic values for the convenience of later research.
Step 2: the value range of irradiance is divided into M sections, wherein M is a positive integer, each section corresponds to an irradiance level and is called as a box in the text, and therefore the irradiance division section can also be called as a box. The common mode is that the value range of irradiance is equally divided into M intervals, and the width of each interval is 1/M of the value range of irradiance. And the value range of irradiance is 0-1, so the value range of irradiance corresponding to each interval is [0,1/M ], [1/M,2/M ], [2/M,3/M ], …, [ (M-1)/M,1 ]. The irradiance of the (M-1, 2, …, M) th level is [ (M-1)/M, M/M ].
The photovoltaic actual output original data set corresponding to the mth irradiance level: at each time interval in the raw data set, there is one irradiance value and one photovoltaic actual power output value. For each irradiance value belonging to the m-th level, a photovoltaic actual output value corresponding to the irradiance value belongs to, and a set formed by all the actual output values is called as a photovoltaic actual output original data set corresponding to the m-th irradiance level.
It is worth mentioning that M is a user-defined positive integer, and the more the available historical data is, the larger the value of M is, and vice versa. On the other hand, if the total amount of the historical data is fixed, the larger the value of M is, the smaller the number of samples in the photovoltaic actual output original data set corresponding to each box is, and the larger the number of samples is, otherwise. For the photovoltaic actual output original data set corresponding to each box, if the number of samples of the set is too small, the reliability of the statistical result is reduced, and the statistical significance is lost. Therefore, in order to guarantee the effect of statistical analysis in the following text, the number of samples in the photovoltaic actual output original data set corresponding to each box should be enough.
Because the whole data does not present a certain distribution characteristic, if the historical irradiance data and the photovoltaic output data are directly fitted without binning, the fitting effect is poor, the fitting precision is low, and therefore binning processing needs to be performed on the historical data.
After the data binning is completed to obtain the photovoltaic actual output original data set corresponding to each bin, data fitting needs to be carried out on the photovoltaic actual output original data set, namely the probability distribution of the photovoltaic actual output in each bin is established. In this embodiment, in each interval, a general distribution fitting method is adopted to fit and obtain a photovoltaic power generation power probability density function and an accumulated distribution function (probability distribution model) corresponding to the interval, so that the photovoltaic power generation power probability distribution density functions corresponding to the intervals form a general distribution probability density function library. The probability density function expression corresponding to the adopted general distribution fitting method is as follows:
the cumulative distribution function expression is:
F(x)＝(1+e-α(x-γ))-β (2)
wherein alpha, beta and gamma are shape parameters respectively and satisfy:
α＞0,β＞0,-∞＜γ＜+∞ (3)
the general distribution has the following excellent properties:
a) the universal distribution has three shape parameters, and PDF and CDF curves of the universal distribution can be flexibly deformed by adjusting the three shape parameters, so that the PDF and CDF curves of the universal distribution can approach the probability distribution of the actual photovoltaic output to the maximum extent, and therefore the universal distribution can accurately represent the probability distribution of the actual photovoltaic output under any irradiance condition.
b) The CDF of the general distribution and the inverse function of the CDF have analytic closed expressions, and the general distribution is used for representing the probability distribution of the photovoltaic output, so that the calculation process of the model can be simplified.
The invention selects two typical subcases (m is 7 and m is 11) to carry out the comparison analysis of the fitting effect, the fitting error is shown in table 1, and the fitting effect is shown in fig. 2.
TABLE 1 comparison of fitting errors for general, Gaussian and Weibull distributions
As can be seen from Table 1, the fitting error of the general distribution to the actual distribution is the smallest, the fitting accuracy is the highest, and the fitting effect is the best.
As can be seen from fig. 2, when the photovoltaic output is smaller, the actual distribution is biased to the left, and the thick tail phenomenon occurs; when the photovoltaic output is larger, the actual distribution of the photovoltaic output is deviated to the right, and a thick first phenomenon occurs. The Gaussian distribution is symmetrical distribution, so that the off-axis characteristic of actual distribution cannot be represented; although the weibull distribution has off-axis characteristics, the fitting effect is not good, and as shown in fig. 2(a), the weibull distribution cannot accurately represent the thick head and the thick tail phenomena of the actual distribution; in contrast, the off-axis characteristic and the thick head and tail phenomena of the actual distribution can be accurately represented due to the fact that the universal distribution has three parameters, and the universal distribution is more accurate in fitting to the peak value of the actual distribution.
And step 3: for any known irradiance, finding a corresponding interval and a photovoltaic power generation power probability distribution density function, performing inverse transformation sampling on the photovoltaic power generation power probability distribution density function to obtain a plurality of photovoltaic power generation power values corresponding to the known irradiance, wherein each photovoltaic power generation power value forms a photovoltaic output scene set.
Specifically, after the historical data is subjected to box-dividing fitting, for any known irradiation intensity, which box the irradiation intensity belongs to is judged, a general distribution function corresponding to the irradiation intensity is found, and inverse transformation sampling is performed on the general distribution by generating a large number of random numbers to obtain a scene set possibly worth of photovoltaic output. The basic method of inverse transforming samples is as follows:
suppose we need to be compliant with Pr(Pt≤p)＝Fl(P) sampling a random variable, i.e. from the random variable PtThe CDF of (1) is randomly sampled, and the inverse transform sampling is as follows:
Pt＝Fl -1(U),U～Unif[0,1] (4)
in the formula, Fl -1(U) is the cumulative probability distribution FlThe inverse function of (c); in the present invention, the cumulative probability distribution FlIs a universal distribution; unif [0, 1]]Is represented by [0, 1]]Are evenly distributed in between.
Introducing a random variable Z following standard normal distributiontThe standard deviation is 1, desirably 0. If random numbers complying with the standard normal distribution are generated in large quantities, the standard normal distribution function value set of these random numbers complies with [0, 1]]Are uniformly distributed. Then, the uniform distribution U in equation (4) can be represented by a standard normal distribution function value Φ (Z)t) And (4) replacing. Known random variable ZtFor random number of (2), for random variable PtThe following formula can be used for sampling:
by using
Finding the variable ZtCorresponding standard normal distribution function value phi (Z)t) Is recycled and reused
Pt＝Fl -1(Φ(Zt)) (6)
Calculating the function value phi (Z) of the standard normal distributiont) Corresponding photovoltaic power generation power value, wherein Fl -1(U) is a cumulative distribution function FlThe inverse function of (c).
The inverse transform sampling flow may be represented by fig. 3.
And 4, step 4: and calculating a weighted average value of all photovoltaic power generation power values in the photovoltaic output scene set as a photovoltaic power generation power predicted value (expected value of photovoltaic output) corresponding to the known irradiance.
A flowchart of the above photovoltaic power generation power prediction method based on the probability distribution function can be represented by fig. 4.
The photovoltaic power generation power prediction method based on the probability distribution function is improved as follows:
1. uneven box separation
In the process of performing bin fitting on the historical data, it is found that if uniform bin fitting is adopted, the historical data in some bins are more, but the data are more dispersive, and even bimodal phenomena occur, as shown in fig. 5. These phenomena all reduce the fitting accuracy to a large extent, so that it is necessary to perform quadratic binning, i.e. non-uniform binning, on these bins. That is, in step 3, the value range of irradiance is divided into M intervals with different widths, so that each corresponding generated power data in each interval forms a single-peak state. The fitting effect after the improved non-uniform binning is shown in fig. 6. As can be seen from FIGS. 5 and 6, after the non-uniform binning improvement is adopted, the historical data dispersibility is greatly reduced, and the double-peak phenomenon does not occur any more, so the fitting effect and the fitting accuracy are greatly improved. The fitting error before and after improvement is shown in table 2:
TABLE 2 comparison of fit errors between uniform binning and improved non-uniform binning
2. Variable coefficient temperature correction
The model only considers the influence of irradiance on photovoltaic output, and two main factors influencing the photovoltaic output are irradiance and component temperature, so a temperature correction link needs to be introduced, and the model is corrected by adopting a constant temperature correction coefficient in the existing method. The invention provides a method for correcting a photovoltaic power generation power probability density function and an accumulative distribution function by introducing a variable temperature correction coefficient, reflecting the calculation error of a model under different temperature conditions at the same time in history, and correcting a photovoltaic output model by using the calculation error, wherein the expression of the temperature correction coefficient is
α＝(Pt cal-Pt actual)/(Tt-Tt-1) (7)
In the above formula, α is the temperature correction coefficient, t is the time, Pt calIs a predicted value of the photovoltaic output model at the time t (namely a predicted value of the photovoltaic power generation power obtained by the photovoltaic power generation power prediction method based on the probability distribution function), Pt actualIs the actual value of the photovoltaic output at the moment T, TtTemperature at time T, Tt-1Is the temperature at time t-1.
Similarly, a box separation idea is adopted, for each interval, the minimum root mean square error of the photovoltaic power generation power is taken as a target, temperature correction basic coefficients at different moments and under different irradiance conditions are obtained, the average value of the temperature correction basic coefficients is taken as the temperature correction coefficient of the interval, namely the temperature correction coefficients in each box are different, so that the model correction precision is improved, and the temperature correction basic coefficients are called as temperature correction variable coefficients.
The specific correction flow is as follows:
a) and dividing the photovoltaic output model predicted value and the photovoltaic output actual value into M sections according to the irradiance per unit value, wherein each section has a plurality of photovoltaic output model predicted value and photovoltaic output actual value data pairs, and finding the temperature value at the corresponding moment.
b) Calculating the difference value between the photovoltaic output model predicted value and the photovoltaic output actual value caused by temperature change at the adjacent moment in each interval according to the formula (7), thereby calculating the temperature coefficient at the adjacent moment in each interval, and taking the average value of the temperature coefficients in the interval as the temperature correction coefficient alpha of the intervalavg。
c) For each prediction time, judging that the irradiance prediction value at the time is in the mth interval, obtaining the photovoltaic output model prediction value by using the probability distribution model of the interval, and then performing temperature correction on the prediction value according to the temperature difference of adjacent times and the formula (8):
in the formula (8), t is a time,
the predicted value of the photovoltaic output after the temperature correction at the time t + 1,
is the temperature correction coefficient of the mth interval (namely the average value of the temperature coefficients of the interval),
the predicted value of the photovoltaic output, T, is obtained by using the probability distribution model of the interval at the moment of T +1tTemperature at time T, Tt+1The temperature at the time t +1 is,
is the amount of correction.
Example results and analysis
The historical data adopted by the invention is derived from a roof photovoltaic power station, the installed capacity of the photovoltaic power station is 9.066MW, the photovoltaic power station consists of two photovoltaic modules and 18 inverters, and the basic parameters of the two photovoltaic modules are shown in a table 3:
TABLE 3 two basic parameters of photovoltaic modules
The invention respectively adopts a physical model for engineering and a probability model (uniform binning, non-uniform binning and variable coefficient temperature correction) based on general distribution to calculate the photovoltaic output, the calculation results of each model are shown in figure 7, and the calculation errors of each model are shown in table 4:
TABLE 4 errors calculated for each model
As can be seen from fig. 7 and table 4, when the irradiance is low, the calculation error of the engineering physical model is large, and the deviation between the calculated value and the actual value of the model is large in several hours after sunrise and before sunset, so that the calculation error of the model as a whole is large, and when the irradiance is high, the calculation accuracy of the engineering physical model is high. Compared with a physical model for engineering, the probability model based on general distribution has higher overall calculation precision and smaller overall calculation error: the calculation accuracy in low irradiance is greatly improved by uniform box separation, but the calculation accuracy near the photovoltaic output peak value is reduced, and the whole calculation error is greatly reduced; on the basis of uniform distribution, the calculation accuracy of low irradiance is further improved by non-uniform distribution, however, the calculation error near the photovoltaic output peak value is not improved, and the whole calculation error is slightly reduced; on the basis of uneven binning, the uneven binning and the temperature correction model not only further improve the calculation precision in the low irradiance, but also improve the calculation error near the photovoltaic output peak value, and greatly reduce the whole calculation error. The calculation time of all models is within 0.3s, and the requirement of rapidity is met.
The physical model calculation value for engineering is the product of the theoretical value of photovoltaic output and the efficiency of the photovoltaic power generation system, and because of factors such as dust, rainwater shelter from, the subassembly is not matched in series, the power loss of inverter, the power loss of direct current and alternating current part cable, the power loss of transformer and the precision of tracking system, the efficiency of the photovoltaic power generation system is usually only 0.8, so there are:
physical model calculation value for engineering (theoretical value of photovoltaic output) 0.8
However, at low irradiance, the actual efficiency of a photovoltaic power generation system tends to be much less than 0.8 for the following reasons:
a) the losses of photovoltaic modules, inverters, cables and the like are not linearly variable, and the loss is large at low irradiance;
b) when the radiation is very low, the situation that the voltage of the string is too low and part of inverters cannot be started exists;
c) the inverter has certain self-power consumption (the self-power consumption is very small along with the change of load), and the proportion of the self-power consumption to the generated energy is higher under the condition of low irradiation, so that the system efficiency is reduced.
Therefore, the actual value of the photovoltaic output at low irradiance is far smaller than the calculated value of the physical model for engineering. The probability model based on the universal distribution establishes the nonlinear relation between irradiance and temperature and actual photovoltaic output, and in the process of establishing the model, the efficiency of the photovoltaic power generation system is already contained in the model, so that the overall prediction precision is high.
In conclusion, the invention adopts the general distribution to fit the historical data, establishes the probability distribution of the photovoltaic power generation power in different irradiance intervals, provides the non-uniform box data fitting method and improves the precision of data fitting. The invention provides a variable coefficient temperature correction model for correcting photovoltaic output, aims at minimizing the root-mean-square error of photovoltaic power generation power, obtains temperature correction coefficients at different moments and under different irradiance conditions, and further improves the precision of the model.
According to the method, a probability distribution modeling method for representing the mapping relation of photovoltaic power generation power, irradiance and temperature is provided, the model does not need detailed parameters of each physical module, and does not relate to a complex solving algorithm, so that when the photovoltaic power generation power is predicted based on the model, the calculation speed is high, and compared with the traditional engineering physical model, the calculation precision is greatly improved; the model is improved by uneven binning and variable coefficient temperature correction, and the calculation results show that the improved model has rapidity and the calculation precision is further improved. Under the condition that a set of meteorological condition predicted values are known, the model can be used for accurately calculating the predicted value of the photovoltaic output under the condition, and the improvement of the model precision is an important link for realizing accurate prediction of the photovoltaic power generation power, so that the model has high practical application value.
The above embodiments are merely illustrative of the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and implement the present invention, and not to limit the protection scope of the present invention. All equivalent changes and modifications made according to the spirit of the present invention should be covered within the protection scope of the present invention.
Claims (10)
1. A photovoltaic power generation power prediction method based on a probability distribution function is characterized by comprising the following steps: the photovoltaic power generation power prediction method comprises the following steps:
step 1: irradiance data of a photovoltaic power generation system to be predicted and corresponding power generation power data are collected in historical data, and data preprocessing is carried out;
step 2: dividing the value range of irradiance into M intervals; in each interval, a general distribution fitting method is adopted to fit to obtain a photovoltaic power generation power probability density function and an accumulative distribution function corresponding to the interval, and the photovoltaic power generation power probability distribution density function corresponding to each interval forms a general distribution probability density function library;
and step 3: for any known irradiance, finding the corresponding interval and the photovoltaic power generation power probability distribution density function, performing inverse transformation sampling on the photovoltaic power generation power probability distribution density function to obtain a plurality of photovoltaic power generation power values corresponding to the known irradiance, wherein each photovoltaic power generation power value forms a photovoltaic output scene set;
and 4, step 4: and calculating the average value of all the photovoltaic power generation power values in the photovoltaic output scene set as the photovoltaic power generation power predicted value corresponding to the known irradiance.
2. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 1, wherein: in step 1, the data preprocessing includes deleting error data, eliminating useless data, and performing per unit on historical data.
3. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 1, wherein: in the step 2, the value range of irradiance is 0-1 p.u. and is equally divided into M intervals, and the width of each interval is 1/M of the value range of irradiance.
4. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 1, wherein: in the step 2, the value range of irradiance is divided into M intervals with different widths, and the power generation data corresponding to each interval form a single-peak state.
5. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 1, wherein: in the step 2, the probability density function expression corresponding to the general distribution fitting method is
The cumulative distribution function expression is f (x) ═ 1+ e-α(x-γ))-βWherein alpha, beta and gamma are shape parameters respectively, and satisfy alpha > 0, beta > 0, - ∞ < gamma < + >.
6. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 1, wherein: in step 3, the method for inverse transforming samples comprises: introducing a random variable ZtBy using
Solving for the random variable ZtCorresponding standard normal distribution function value phi (Z)t) Reuse of Pt＝Fl -1(Φ(Zt) The value of the standard normal distribution function phi (Z) is obtainedt) Corresponding photovoltaic power generation power value, wherein
Is a cumulative distribution boxNumber FlThe inverse function of (c).
7. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 1, wherein: in the step 4, a weighted average value is obtained for each photovoltaic power generation power value in the photovoltaic output scene set and is used as a photovoltaic power generation power predicted value corresponding to the known irradiance.
8. The photovoltaic power generation power prediction method based on the probability distribution function according to any one of claims 1 to 7, wherein: and in the step 2, introducing a variable temperature correction coefficient to correct the photovoltaic power generation power probability density function and the accumulative distribution function.
9. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 8, wherein: the expression of the temperature correction coefficient is alpha ═ (P)t cal-Pt actual)/(Tt-Tt-1) Where α is the temperature correction coefficient, t is the time, Pt calIs a predicted value of photovoltaic power generation power at the time t, Pt actualIs the actual value of the photovoltaic output at the moment T, TtTemperature at time T, Tt-1Is the temperature at time t-1.
10. The photovoltaic power generation power prediction method based on the probability distribution function according to claim 9, wherein: and for each interval, obtaining a plurality of temperature correction basic coefficients at different moments and under different irradiance conditions by taking the minimum root-mean-square error of the photovoltaic power generation power as a target, and taking the average value of the temperature correction basic coefficients as the temperature correction coefficient of the interval.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910031330.XA CN109767353B (en) | 2019-01-14 | 2019-01-14 | Photovoltaic power generation power prediction method based on probability distribution function |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910031330.XA CN109767353B (en) | 2019-01-14 | 2019-01-14 | Photovoltaic power generation power prediction method based on probability distribution function |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109767353A CN109767353A (en) | 2019-05-17 |
| CN109767353B true CN109767353B (en) | 2020-12-18 |
Family
ID=66452777
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910031330.XA Active CN109767353B (en) | 2019-01-14 | 2019-01-14 | Photovoltaic power generation power prediction method based on probability distribution function |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109767353B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111476411B (en) * | 2020-04-01 | 2023-10-31 | 阳光电源（上海）有限公司 | Photovoltaic power station generating capacity calculation method and server |
| CN111861259B (en) * | 2020-07-30 | 2022-07-15 | 中国大唐集团科学技术研究院有限公司华东电力试验研究院 | Load modeling method, system and storage medium considering time sequence |
| CN113570116B (en) * | 2021-07-02 | 2022-02-18 | 福建时代星云科技有限公司 | Method for estimating heating power of storage and charging station and terminal |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103208029A (en) * | 2013-03-11 | 2013-07-17 | 中国电力科学研究院 | Super-short-term power prediction method based on clearance model for photovoltaic power station |
| CN104657791A (en) * | 2015-02-28 | 2015-05-27 | 武汉大学 | Wind power plant group wind speed distribution prediction method based on correlation analysis |
| CN105023070A (en) * | 2015-08-12 | 2015-11-04 | 河海大学常州校区 | Output power prediction method of photovoltaic system |
| CN105490310A (en) * | 2016-01-25 | 2016-04-13 | 武汉大学 | Real-time dynamic economical scheduling method for generally-distributed electric power system containing wind power |
| CN105846425A (en) * | 2016-04-08 | 2016-08-10 | 江苏省电力试验研究院有限公司 | Economic dispatching method based on general wind power forecasting error model |
| CN105930671A (en) * | 2016-04-29 | 2016-09-07 | 武汉大学 | Improved versatile distribution and versatile mixture distribution models characterizing wind power probability distribution |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9638831B1 (en) * | 2011-07-25 | 2017-05-02 | Clean Power Research, L.L.C. | Computer-implemented system and method for generating a risk-adjusted probabilistic forecast of renewable power production for a fleet |
-
2019
- 2019-01-14 CN CN201910031330.XA patent/CN109767353B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103208029A (en) * | 2013-03-11 | 2013-07-17 | 中国电力科学研究院 | Super-short-term power prediction method based on clearance model for photovoltaic power station |
| CN104657791A (en) * | 2015-02-28 | 2015-05-27 | 武汉大学 | Wind power plant group wind speed distribution prediction method based on correlation analysis |
| CN105023070A (en) * | 2015-08-12 | 2015-11-04 | 河海大学常州校区 | Output power prediction method of photovoltaic system |
| CN105490310A (en) * | 2016-01-25 | 2016-04-13 | 武汉大学 | Real-time dynamic economical scheduling method for generally-distributed electric power system containing wind power |
| CN105846425A (en) * | 2016-04-08 | 2016-08-10 | 江苏省电力试验研究院有限公司 | Economic dispatching method based on general wind power forecasting error model |
| CN105930671A (en) * | 2016-04-29 | 2016-09-07 | 武汉大学 | Improved versatile distribution and versatile mixture distribution models characterizing wind power probability distribution |
Non-Patent Citations (2)
| Title |
|---|
| 光伏电池的建模综述;张纯杰 等;《电源技术》;20160430;第40卷(第4期);第927-930页 * |
| 基于通用分布的风电功率出力动态场景生成方法;曹慧秋 等;《电测与仪表》;20181225;第55卷(第24期);第6-10页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109767353A (en) | 2019-05-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109767353B (en) | Photovoltaic power generation power prediction method based on probability distribution function | |
| CN110321601B (en) | Advanced prediction method and system for dynamic current carrying capacity of overhead line | |
| CN110474339B (en) | Power grid reactive power control method based on deep power generation load prediction | |
| CN108063456B (en) | Distributed photovoltaic power generation grid-connected planning method and terminal equipment | |
| CN114792156B (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| CN111695736B (en) | Photovoltaic power generation short-term power prediction method based on multi-model fusion | |
| CN113592165B (en) | Sub-season climate prediction method and system based on ten-day trend and physical mode modeling | |
| CN112653198A (en) | Wind power output scene generation method and system based on prediction box | |
| CN110365053A (en) | Short-term wind power forecast method based on delay optimisation strategy | |
| CN112668806B (en) | Photovoltaic power ultra-short-term prediction method based on improved random forest | |
| Kanchana et al. | PV Power Forecasting with Holt-Winters Method | |
| CN115270640A (en) | Configuration optimization method for distributed photovoltaic power generation system of microgrid and terminal equipment | |
| CN108694475B (en) | Short-time-scale photovoltaic cell power generation capacity prediction method based on hybrid model | |
| CN113033136B (en) | Simplified photovoltaic cell physical parameter extraction optimization method and system | |
| CN111105072A (en) | Method for predicting health degree of photovoltaic power station based on entropy weight fuzzy synthesis method | |
| CN116054738B (en) | Solar power generation monitoring management system and method of Internet of things | |
| CN115575737A (en) | Current-carrying capacity detection method and system | |
| CN115936539A (en) | Static voltage stability rapid evaluation method considering wind power prediction error | |
| Yan et al. | A high concentrated photovoltaic output power predictive model based on Fuzzy Clustering and RBF neural network | |
| CN115173421A (en) | Probability optimal power flow calculation method based on progressive Latin hypercube sampling | |
| CN114301091A (en) | Photovoltaic array output power control method and device | |
| CN113780643A (en) | Photovoltaic power station short-term output prediction method based on case reasoning | |
| CN117937478B (en) | Photovoltaic power prediction method, device, computer equipment and storage medium | |
| WO2022054954A1 (en) | Method, system, and program for estimating power generation characteristics of solar cell. | |
| CN117663503B (en) | Method and system for intelligently adjusting molten salt heat storage rate |
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 |