CN-122026326-A - Ultra-short-term photovoltaic power prediction method for self-adaptive double-dimensional modeling and closed-loop deviation correction

CN122026326ACN 122026326 ACN122026326 ACN 122026326ACN-122026326-A

Abstract

The invention belongs to the technical field of photovoltaic power prediction methods, and particularly relates to an ultra-short-term photovoltaic power prediction method with self-adaptive double-dimensional modeling and closed loop deviation correction, wherein multidimensional data acquisition and characteristic preprocessing are carried out, and static parameters and historical operation data of a photovoltaic power station are obtained and normalized; constructing a time period, an irradiation two-dimensional dynamic K coefficient matrix, dividing the time period and the irradiation intensity clustering interval, calculating a corresponding K coefficient, calculating initial predicted power generation power of a predicted day based on the K coefficient matrix, introducing proportional-integral closed loop deviation compensation, calculating deviation amount, accumulated deviation amount and unit compensation coefficient of a day before the predicted day, combining the unit compensation coefficient and the initial predicted power generation power, and applying non-physical constraint to obtain final predicted power generation.

Inventors

HU CHAORAN
ZHANG XISHUAI
SHEN HONGYU

Assignees

江苏云储聚合科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (9)

1. The ultra-short-term photovoltaic power prediction method for self-adaptive double-dimensional modeling and closed-loop deviation correction is characterized by comprising the following steps of: Step one, multidimensional data acquisition and characteristic preprocessing are carried out, static parameters and historical operation data of a photovoltaic power station are obtained, and normalization processing is carried out; Step two, constructing a time period and irradiation two-dimensional dynamic K coefficient matrix, dividing a time period and irradiation intensity clustering interval and calculating a corresponding K coefficient; calculating initial predicted power generation of a predicted day based on the K coefficient matrix; Step four, introducing proportional-integral closed loop deviation compensation, and calculating the deviation amount, accumulated deviation amount and unit compensation coefficient of the day before the predicted day; and fifthly, combining the unit compensation coefficient and the initial predicted power generation power, and obtaining the final predicted power generation power after applying non-physical constraint.
2. The method for ultra-short term photovoltaic power prediction using adaptive bi-dimensional modeling and closed loop rectification as defined in claim 1, wherein the normalization process in step one comprises calculating the actual generated power per unit scale The calculation formula is as follows; ; Wherein, the Absolute value of actual generated power; is the direct current side installed capacity.
3. The adaptive bi-dimensional modeling and closed-loop rectification ultra-short-term photovoltaic power prediction method according to claim 1 is characterized in that the irradiation intensity clustering interval in the second step is divided into 4 continuous intervals, wherein the 4 continuous intervals are respectively: [0,500]W/m2、[501,1000]W/m2、[1001,1500]W/m2、[1501,2000]W/m2。
4. The adaptive bi-dimensional modeling and closed-loop rectification ultra-short term photovoltaic power prediction method according to claim 1 is characterized in that the K coefficient calculation formula in the second step is as follows: ; Wherein, the I.e. only statistics of historical data of irradiation intensities falling within the corresponding interval; The sum of the actual power generated by the historical unit scale under the scene is calculated; Is the sum of the historic measured irradiation intensities under the scene.
5. The method for predicting ultra-short term photovoltaic power by self-adaptive bi-dimensional modeling and closed loop rectification as claimed in claim 1, wherein the deviation amount of the fourth step is as follows The calculation formula is as follows: ; Wherein, the The average value of the actual power generation power of all minutes in the irradiation interval scene in the period of D-1 day; The average value of the predicted power generation power after correction for all minutes in the scene is calculated, and the data belong to the corresponding irradiation interval 。
6. The method for predicting ultra-short term photovoltaic power by adaptive bi-dimensional modeling and closed loop rectification as claimed in claim 5, wherein the accumulated deviation amount of step four is The accumulated deviation is the accumulation of historical deviation and is used for capturing the persistence characteristic of the prediction error, and the calculation formula is as follows: ; Wherein, the The accumulated deviation amount of the D-2 day corresponding period and the irradiation interval scene is shown; is the deviation amount in the scene on day D-1.
7. The method for predicting ultra-short term photovoltaic power by self-adaptive bi-dimensional modeling and closed loop rectification as claimed in claim 1, wherein the unit compensation coefficient of the fourth step is Combining a proportional term and an integral term, calculating a unit compensation coefficient under the scene, and realizing double correction of short-term fluctuation and long-term system error, wherein the calculation formula is as follows: ; Wherein the data belongs to the interval R, The correction coefficient is a preset proportion and integral correction coefficient, and can be finely adjusted according to the characteristics of the power station; the average value of the power generated is initially predicted for all minute steps in the scene on day D-1.
8. The ultra-short term photovoltaic power prediction method for self-adaptive bi-dimensional modeling and closed loop correction according to claim 1, wherein the unit scale corrected predicted generated power calculation formula in the fifth step is as follows: 。
9. The ultra-short term photovoltaic power prediction method for self-adaptive bi-dimensional modeling and closed loop rectification according to claim 1, wherein the final prediction generated power calculation formula in the fifth step is as follows: 。

Description

Ultra-short-term photovoltaic power prediction method for self-adaptive double-dimensional modeling and closed-loop deviation correction Technical Field The invention belongs to the technical field of photovoltaic power prediction methods, and particularly relates to an ultra-short-term photovoltaic power prediction method with self-adaptive double-dimensional modeling and closed loop deviation correction. Background The ultra-short term photovoltaic power prediction is a core support technology for grid-connected scheduling of photovoltaic power stations, electric market transaction and new energy consumption, and the prediction accuracy directly influences the stability and economy of power grid operation. The photovoltaic output is influenced by multiple environmental factors such as solar irradiation intensity, ambient temperature, solar incidence angle and the like in a coupling way, has the characteristics of strong volatility, randomness and intermittence, and particularly in an ultra-short-term scale (in minutes), the rapid change of the irradiation intensity can cause the severe fluctuation of the photovoltaic power, so that the method brings great challenges to accurate prediction. The existing prediction method is mainly based on a single dimension to construct a model, or only focuses on time interval division, or considers irradiation intensity influence alone, coupling association characteristics of the irradiation intensity and the time interval cannot be fully excavated, so that the model is insufficient in adaptability to complex and changeable environmental scenes, key parameters such as K coefficients are difficult to dynamically match with actual running states, and prediction errors are large. The traditional prediction method generally lacks an effective closed-loop deviation compensation mechanism, static modeling is only carried out by relying on historical data, and systematic errors and random fluctuation errors generated in the prediction process cannot be corrected in real time. Although deviation correction is introduced into a part of the method, a single proportion adjustment or simple historical deviation averaging mode is adopted, both quick response of short-term fluctuation errors and accumulation and offset of long-term system errors are difficult to achieve, influence of installation scale differences of different power stations on universality of a prediction model is not fully considered, and a normalization processing mode is imperfect, so that mobility of the model among photovoltaic power stations with different specifications is poor, and comprehensive requirements of prediction precision, adaptability and universality in actual engineering are difficult to meet. Disclosure of Invention The invention aims to provide an ultra-short-term photovoltaic power prediction method with self-adaptive two-dimensional modeling and closed-loop deviation correction aiming at the technical problems. In view of the above, the invention provides an ultra-short term photovoltaic power prediction method for self-adaptive bi-dimensional modeling and closed loop correction, which comprises the following steps: Step one, multidimensional data acquisition and characteristic preprocessing are carried out, static parameters and historical operation data of a photovoltaic power station are obtained, and normalization processing is carried out; Step two, constructing a time period and irradiation two-dimensional dynamic K coefficient matrix, dividing a time period and irradiation intensity clustering interval and calculating a corresponding K coefficient; calculating initial predicted power generation of a predicted day based on the K coefficient matrix; Step four, introducing proportional-integral closed loop deviation compensation, and calculating the deviation amount, accumulated deviation amount and unit compensation coefficient of the day before the predicted day; and fifthly, combining the unit compensation coefficient and the initial predicted power generation power, and obtaining the final predicted power generation power after applying non-physical constraint. Preferably, the normalization in the first step includes calculating the actual power generation per unit scaleThe calculation formula is as follows; ; Wherein, the Absolute value of actual generated power; is the direct current side installed capacity. Preferably, the irradiation intensity clustering interval in the second step is divided into 4 continuous intervals, which are respectively: [0,500]W/m2、[501,1000]W/m2、[1001,1500]W/m2、[1501,2000]W/m2。 The adaptive bi-dimensional modeling and closed-loop rectification ultra-short term photovoltaic power prediction method according to claim 1 is characterized in that the K coefficient calculation formula in the second step is as follows: ; Wherein, the I.e. only statistics of historical data of irradiation intensities falling within the corresponding interval; The sum of the ac