CN106251023B - Photovoltaic power short-term prediction method suitable for small sample - Google Patents
Photovoltaic power short-term prediction method suitable for small sample Download PDFInfo
- Publication number
- CN106251023B CN106251023B CN201610645331.XA CN201610645331A CN106251023B CN 106251023 B CN106251023 B CN 106251023B CN 201610645331 A CN201610645331 A CN 201610645331A CN 106251023 B CN106251023 B CN 106251023B
- Authority
- CN
- China
- Prior art keywords
- solar radiation
- time period
- surface solar
- radiation value
- weather
- 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
-
- 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
Abstract
The invention provides a short-term photovoltaic power prediction method suitable for a small sample, which comprises the following steps: screening historical data; carrying out statistics on penetration rate; transforming input quantity; training a neural network model; and (4) predicting a neural network model. According to the decoupling characteristic of influence factors of all links of photovoltaic power generation, the traditional prediction model is split, so that the network structure of all parts is simplified; and effectively integrating the cloud cover factors into the model input quantity through statistical analysis of the corresponding relation between the weather type and the cloud cover degree. The invention simplifies the input and output relations of the neural network prediction model, reduces the complexity of the relation between input and output, and reduces the requirements on training samples.
Description
Technical Field
The invention belongs to the technical field of new energy photovoltaic power generation, and particularly relates to a short-term photovoltaic power prediction method suitable for a small sample.
Background
In recent years, with the development of socioeconomic, the problems of energy shortage and environmental pollution are increasingly highlighted, and the development and utilization of renewable energy sources become effective ways to solve the problems of energy sources and environment. In new energy power generation, photovoltaic power generation is rapidly developed due to the advantages of safety, reliability, less regional limitation, short construction period and the like, and has a larger industrial scale at present. However, due to the influence of factors such as weather, the photovoltaic output power has great uncertainty in a short time scale, and an accurate and effective photovoltaic prediction technology has important significance for safe and stable operation of a system and effective utilization of photovoltaic energy.
The photovoltaic short-term prediction method one day ahead generally needs sufficient historical data support, the requirement on data accumulation time is different from 3 months to 1 year, if the photovoltaic power station is in the initial operation stage, the historical data accumulation is insufficient, and the application of the conventional prediction method is greatly limited, so that the conventional method needs to be improved aiming at a small sample situation, and the power prediction method suitable for the photovoltaic power generation system in the initial operation stage is obtained.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a photovoltaic power short-term prediction method suitable for small samples, which is characterized in that a prediction model is split by utilizing the natural decoupling characteristic of each influence factor, and cloud shading factors are effectively integrated into model input quantity through statistical analysis of the corresponding relation between weather types and cloud shading degrees, so that the network structure is simplified, the complexity of the relation between input and output is reduced, and the demand of the prediction method on historical data is reduced.
In order to achieve the purpose of the invention, the invention adopts the following technical scheme:
a photovoltaic power short-term prediction method suitable for small samples is provided, and the method comprises the following steps:
the first step is as follows: historical data screening
Removing parts with poor air quality and high air humidity from historical data to obtain a sample of which the surface solar radiation value is mainly influenced by cloud cover;
the second step is that: penetration statistics
Calculating the i-th time interval penetration rate k of the j-th day by using the earth surface solar radiation values of time intervals of historical adjacent daysi,jCombining historical measured weather types w in the ith time period, and counting to obtain expected penetration rates corresponding to the weather types in the ith time period;
on the basis of the single-period penetration rate statistics, the penetration rate statistics considering the weather types of the previous and subsequent periods is carried out to obtain the expected penetration rate in the ith day period of the jth day considering the weather types of the previous and subsequent periods
The third step: input quantity conversion
According to the weather types of the time period to be predicted and the time periods before and after the time period to be predicted, combining the statistical result in the second step to obtain the expected penetration rate of the time period to be predicted, combining the average earth surface solar radiation value of the corresponding time period in the cloud-free weather to obtain the earth surface solar radiation value after the cloud cover, and completing the conversion of the input quantity from the weather type to the earth surface solar radiation value after the cloud cover;
the fourth step: neural network model training
Carrying out neural network model training by using historical data, and carrying out model training by using a cloud-covered ground surface solar radiation value, a historical measured air quality index and a historical measured environment temperature corresponding to a historical measured weather type as input quantities and a historical measured photovoltaic output as output quantities, wherein the time interval of all data is 1 h;
the fifth step: neural network model prediction
And obtaining a corresponding cloud-covered ground surface solar radiation value according to the weather type forecast information of each time period of the day to be predicted, taking the cloud-covered ground surface solar radiation value as an input item of weather type influence factors, and obtaining a photovoltaic output power prediction result of each time period of the day to be predicted by combining the air temperature forecast information and the air quality index forecast information, wherein the time interval of all data is 1 h.
The penetration rate k of the j day and the i period in the second stepi,jThe calculation formula of (2) is as follows:
ki,j＝Ri,j/R0ii＝0,1,...,23，j＝1,2,...,n
in the formula, Ri,jMean surface solar radiation value, R, at day j and time i0iThe average earth surface solar radiation value of the corresponding time period under the cloudless weather is represented, and the average earth surface solar radiation value is approximately calculated by the maximum earth surface solar radiation value in the historical data of the corresponding time period of the n adjacent days:
R0i≈max{Ri,1,Ri,2,…,Ri,j,…,Ri,n}j＝1,2,…,n。
in the third step, the calculation formula of the ground surface solar radiation value after cloud shading is as follows:
in the formula (I), the compound is shown in the specification,
representing the surface solar radiation value of the i time period of the j day after cloud cover reduction,
and (4) the expected penetration rate corresponding to the weather type of the ith time period and the adjacent time period on the jth day is shown.
The neural network model used in the fourth step is divided into two parts which respectively represent a solar radiation reduction process and a photoelectric conversion process, wherein the input quantity of the neural network of the first part is a ground surface solar radiation value after cloud shading in a day-to-be-predicted time period
Air quality index Ai,jThe output quantity is the earth surface solar radiation value S of the time period to be predictedi,j(ii) a The input quantity of the second part of the neural network is the earth surface solar radiation value S of the time period to be predictedi,jAmbient temperature Ti,jThe output quantity is the photovoltaic power P in the period of waiting for predictionfi,j。
In the first step, the parts with poor air quality and high air humidity in the historical data are removed by removing the parts with the air quality index larger than 100 and the air humidity larger than 80%.
The power prediction method is an improvement on the conventional method aiming at the situation of small samples, and is suitable for the photovoltaic power generation system at the initial stage of operation.
Drawings
FIG. 1 is a schematic diagram of a prediction flow;
fig. 2 is a schematic diagram of input and output relationships of a neural network.
FIG. 3 is a graph of the predicted results of the comparison method under different numbers of samples.
FIG. 4 is a graph of predicted results for different numbers of samples according to the method of the present invention.
FIG. 5 is a plot of the root mean square error of a comparative method and the method of the present invention.
Detailed Description
The technical solution of the present patent will be described in further detail with reference to the following embodiments.
Referring to fig. 1, a method for short-term photovoltaic power prediction for a small sample includes the following steps:
the first step is as follows: historical data screening
And eliminating parts with the air quality index being more than 100 and the air humidity being more than 80% at the corresponding moment from the historical surface solar radiation data and the historical photovoltaic power data to obtain a sample with the surface solar radiation value mainly influenced by the cloud cover.
The second step is that: penetration statistics
Obtaining historical weather types from weather service websites, and determining weather type division such as sunny, cloudy, rainy and snowy according to the historical weather types recorded in the websites.
Calculating the penetration rate of each time period in the historical data by using the earth surface solar radiation value of each time period of the historical adjacent days, and recording the penetration rate of the ith time period of the jth day as ki,jAnd combining the historical measured weather types w in the ith time period to obtain penetration rate distribution corresponding to various weather types in the ith time period, and calculating according to the penetration rate distribution to obtain penetration rate expectation corresponding to various weather types in the ith time period.
And on the basis of the single-period penetration rate statistics, performing penetration rate statistics considering weather types of the previous and subsequent periods to obtain the penetration rate expectation considering the weather types of the previous and subsequent periods. For example, if the weather type in the current period i is cloudy, the weather type in the period i-1 is clear, and the weather type in the period i +1 is cloudy, the weather type in the current period is recorded as (clear) -cloudy- (cloudy). After the weather types of the form considering the weather conditions of the previous and subsequent time periods are obtained, the penetration rate distribution of the weather types of the previous and subsequent time periods can be obtained by combining the penetration rate of the single time period, and the expected penetration rate of the jth day and the ith time period considering the weather types of the previous and subsequent time periods is obtained by calculation according to the penetration rate distribution
The third step: input quantity conversion
And according to the weather types of the time period to be predicted and the time periods before and after the time period, obtaining the expected penetration rate of the time period to be predicted by searching the statistical result in the second step, integrating the time information into the penetration rate to obtain the earth surface solar radiation value after the cloud shelter, and completing the conversion of the input quantity from the weather type to the earth surface solar radiation value after the cloud shelter.
The fourth step: neural network model training
And training a neural network model by using historical data, wherein the model is trained by using a cloud-covered ground surface solar radiation value, a historical measured air quality index and a historical measured environment temperature corresponding to a historical measured weather type as input quantities and a historical measured photovoltaic output as output quantities, and the time interval of all data is 1 h.
The fifth step: neural network model prediction
And obtaining corresponding ground surface solar radiation values after cloud shading according to the weather type forecast information of each time period of the day to be predicted, and taking the values as input items of weather type influence factors. And (4) obtaining a photovoltaic output power prediction result of each time period of the day to be predicted by combining the air temperature prediction information and the air quality index prediction information, wherein the time interval of all data is 1 h.
The penetration rate k in the ith period of the j day in the second stepi,jThe calculation formula of (2) is as follows:
ki,j＝Ri,j/R0ii＝0,1,...,23，j＝1,2,...,n
in the formula, Ri,jMean surface solar radiation value, R, at day j and time i0iThe average earth surface solar radiation value of the corresponding time period under the cloudless weather is represented, and the average earth surface solar radiation value is approximately calculated by the maximum earth surface solar radiation value in the historical data of the corresponding time period of the n adjacent days:
R0i≈max{Ri,1,Ri,2,…,Ri,j,…,Ri,n}j＝1,2,…,n
in the third step, the calculation formula of the ground surface solar radiation value after cloud shading is as follows:
in the formula (I), the compound is shown in the specification,
representing the surface solar radiation value of the i time period of the j day after cloud cover reduction,
and (4) the expected penetration rate corresponding to the weather type of the ith time period and the adjacent time period on the jth day is shown.
The neural network model used in the fourth step is divided into two parts, which respectively represent the earth surface solar radiation reduction process and the photoelectric conversion process. Wherein, the input quantity of the first part of neural network is the surface solar radiation value after the cloud cover in the time period to be predicted for the day
Air quality index Ai,jThe output quantity is the earth surface solar radiation value S of the time period to be predictedi,j(ii) a The input quantity of the second part of the neural network is the earth surface solar radiation value S of the time period to be predictedi,jAmbient temperature Ti,jThe output quantity is the photovoltaic power P in the period of waiting for predictionfi,jPlease refer to fig. 2.
The traditional neural network multi-step prediction method without input quantity transformation is used as a comparison method, and the prediction results under different numbers of training samples are shown in fig. 3. The prediction results of the method of the present invention under different numbers of training samples are shown in FIG. 4. The prediction effect of the method is better than that of the comparison method under the condition of small samples, the prediction effect of the comparison method is gradually improved along with the increase of the number of the samples, and the prediction effects of the two methods are close to each other.
The mean of the root mean square errors of all the prediction days under different numbers of training samples in the comparison method and the method of the present invention are shown in fig. 5. The dependence of the comparison method on the training samples is larger, and the prediction error is gradually reduced along with the increase of the number of the samples. The method has good adaptability to small samples, and the prediction errors are not changed greatly and are all kept at a low level in the process of increasing the number of samples from 10 to 40.
Claims (4)
1. A short-term photovoltaic power prediction method suitable for small samples is characterized by comprising the following steps:
the first step is as follows: historical data screening
Removing parts with poor air quality and high air humidity from historical data to obtain a sample of which the surface solar radiation value is mainly influenced by cloud cover;
the second step is that: penetration statistics
Calculating the i-th time interval penetration rate k of the j-th day by using the earth surface solar radiation values of time intervals of historical adjacent daysi,jCombining historical measured weather types w in the ith time period, and counting to obtain expected penetration rates corresponding to the weather types in the ith time period;
on the basis of the single-period penetration rate statistics, the penetration rate statistics considering the weather types of the previous and subsequent periods is carried out to obtain the expected penetration rate in the ith day period of the jth day considering the weather types of the previous and subsequent periods
The third step: input quantity conversion
According to the weather types of the time period to be predicted and the time periods before and after the time period to be predicted, combining the statistical result in the second step to obtain the expected penetration rate of the time period to be predicted, combining the average earth surface solar radiation value of the corresponding time period in the cloud-free weather to obtain the earth surface solar radiation value after the cloud cover, and completing the conversion of the input quantity from the weather type to the earth surface solar radiation value after the cloud cover;
the fourth step: neural network model training
Carrying out neural network model training by using historical data, and carrying out model training by using a cloud-covered ground surface solar radiation value, a historical measured air quality index and a historical measured environment temperature corresponding to a historical measured weather type as input quantities and a historical measured photovoltaic output as output quantities, wherein the time interval of all data is 1 h;
the neural network model used in the fourth step is divided into two parts which respectively represent a solar radiation reduction process and a photoelectric conversion process, wherein the input quantity of the neural network of the first part is a ground surface solar radiation value after cloud shading in a day-to-be-predicted time period
Air quality index Ai,jThe output quantity is the earth surface solar radiation value S of the time period to be predictedi,j(ii) a The input quantity of the second part of the neural network is the earth surface solar radiation value S of the time period to be predictedi,jAmbient temperature Ti,jThe output quantity is the photovoltaic power P in the period of waiting for predictionfi,j；
The fifth step: neural network model prediction
And obtaining a corresponding cloud-covered ground surface solar radiation value according to the weather type forecast information of each time period of the day to be predicted, taking the cloud-covered ground surface solar radiation value as an input item of weather type influence factors, and obtaining a photovoltaic output power prediction result of each time period of the day to be predicted by combining the air temperature forecast information and the air quality index forecast information, wherein the time interval of all data is 1 h.
2. The method for short-term prediction of photovoltaic power as claimed in claim 1, characterized in that: the penetration rate k of the j day and the i period in the second stepi,jThe calculation formula of (2) is as follows:
ki,j＝Ri,j/R0ii＝0,1,...,23，j＝1,2,...,n
in the formula, Ri,jMean surface solar radiation value, R, at day j and time i0iThe average earth surface solar radiation value of the corresponding time period under the cloudless weather is represented, and the average earth surface solar radiation value is approximately calculated by the maximum earth surface solar radiation value in the historical data of the corresponding time period of the n adjacent days:
R0i≈max{Ri,1,Ri,2,…,Ri,j,…,Ri,n} j＝1,2,…,n。
3. the method for short-term prediction of photovoltaic power as claimed in claim 1, characterized in that: in the third step, the calculation formula of the ground surface solar radiation value after cloud shading is as follows:
in the formula (I), the compound is shown in the specification,
representing the surface solar radiation value R of the i-th time period of the j day after cloud cover reduction0iRepresenting an average surface solar radiation value of a corresponding time period in the cloud-free weather;and (4) the expected penetration rate corresponding to the weather type of the ith time period and the adjacent time period on the jth day is shown.
4. The method for short-term prediction of photovoltaic power as claimed in claim 1, characterized in that: in the first step, the parts with poor air quality and high air humidity in the historical data are removed by removing the parts with the air quality index larger than 100 and the air humidity larger than 80%.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610645331.XA CN106251023B (en) | 2016-08-05 | 2016-08-05 | Photovoltaic power short-term prediction method suitable for small sample |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610645331.XA CN106251023B (en) | 2016-08-05 | 2016-08-05 | Photovoltaic power short-term prediction method suitable for small sample |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106251023A CN106251023A (en) | 2016-12-21 |
| CN106251023B true CN106251023B (en) | 2020-01-14 |
Family
ID=58078850
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610645331.XA Active CN106251023B (en) | 2016-08-05 | 2016-08-05 | Photovoltaic power short-term prediction method suitable for small sample |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106251023B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105512760A (en) * | 2015-12-04 | 2016-04-20 | 北京国电通网络技术有限公司 | Neural network-based calculation method and calculation system for calculating power-generating capacity of photovoltaic station |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9322951B2 (en) * | 2007-02-12 | 2016-04-26 | Locus Energy, Inc. | Weather and satellite model for estimating solar irradiance |
| CN102281016A (en) * | 2011-06-09 | 2011-12-14 | 国网电力科学研究院 | Clear-sky photovoltaic super-short-term power forecast method based on real-time radiation acquisition technology |
| CN103971169B (en) * | 2014-04-14 | 2017-09-01 | 国家电网公司 | A kind of Forecasting Methodology for the photovoltaic ultra-short term generated output simulated based on cloud amount |
| CN103996082B (en) * | 2014-06-03 | 2017-06-20 | 华北电力大学（保定） | A kind of intensity of solar radiation Forecasting Methodology theoretical based on dual random |
| CN105303254B (en) * | 2015-10-26 | 2019-05-31 | 国网浙江省电力公司电力科学研究院 | A kind of pair of photovoltaic plant receives the method and device that radiation is predicted |
-
2016
- 2016-08-05 CN CN201610645331.XA patent/CN106251023B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105512760A (en) * | 2015-12-04 | 2016-04-20 | 北京国电通网络技术有限公司 | Neural network-based calculation method and calculation system for calculating power-generating capacity of photovoltaic station |
Non-Patent Citations (2)
| Title |
|---|
| 基于BP神经网络-马尔科夫链的光伏发电预测;吴雪莲等;《电工电气》;20141231;第22-27页 * |
| 考虑云量和气溶胶不确定性的太阳辐射值预测;赵书强等;《电工电能新技术》;20150531;第34卷(第5期);第41-46,75页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106251023A (en) | 2016-12-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108388956B (en) | Photovoltaic power prediction method considering radiation attenuation | |
| CN109524993B (en) | Wind power photovoltaic typical cycle output scene generation method for medium-long term optimization scheduling | |
| CN107194495B (en) | Photovoltaic power longitudinal prediction method based on historical data mining | |
| CN112257941B (en) | Photovoltaic power station short-term power prediction method based on improved Bi-LSTM | |
| CN113496311A (en) | Photovoltaic power station generated power prediction method and system | |
| CN104021427A (en) | Method for predicting daily generating capacity of grid-connected photovoltaic power station based on factor analysis | |
| CN111260126B (en) | Short-term photovoltaic power generation prediction method considering correlation degree of weather and meteorological factors | |
| CN106600060B (en) | Solar irradiation amount prediction method based on similar day sunny coefficient correction | |
| CN110334870B (en) | Photovoltaic power station short-term power prediction method based on gated cyclic unit network | |
| CN103473621A (en) | Wind power station short-term power prediction method | |
| CN114792156A (en) | Photovoltaic output power prediction method and system based on curve characteristic index clustering | |
| CN111626473A (en) | Two-stage photovoltaic power prediction method considering error correction | |
| CN112070311A (en) | Day-ahead light power prediction method based on similar day clustering and meteorological factor weighting | |
| CN114511132A (en) | Photovoltaic output short-term prediction method and prediction system | |
| WO2024051524A1 (en) | Joint prediction method and apparatus for hydraulic, wind and photovoltaic generation power | |
| CN113837426A (en) | Weather-typing-based photovoltaic power prediction method | |
| CN115441447B (en) | New energy power generation power prediction method | |
| CN115936177A (en) | Photovoltaic output power prediction method and system based on neural network | |
| CN110852492A (en) | Photovoltaic power ultra-short-term prediction method for finding similarity based on Mahalanobis distance | |
| CN113991711A (en) | Capacity configuration method for energy storage system of photovoltaic power station | |
| CN104915727A (en) | Multi-dimensional isomorphic heterogeneous BP neural network optical power ultrashort-term prediction method | |
| Matushkin et al. | Analysis of factors for forecasting electric power generation by solar power plants | |
| CN104834981B (en) | The main cause of short-term power station photovoltaic power implies type prediction technique | |
| Guo et al. | A new approach for interval forecasting of photovoltaic power based on generalized weather classification | |
| CN116581755A (en) | Power prediction method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |