Photovoltaic irradiance forecast method and system based on xgboost

Abstract

The present invention proposes a method and system for forecasting photovoltaic irradiance based on XGBoost. The data preparation and preprocessing include the data set R of the historical observation irradiance data of photovoltaic power plants, and the NWP forecast data of the same time period as the data set R Set W; calculate the clear sky irradiance at the same time as the data set R according to the longitude and latitude of the photovoltaic power station, and obtain the data set RT; multiply and divide all the elements in the data set W and RT respectively to obtain several columns of new elements for cross processing as the data set W2; use the XGBoost algorithm to establish a model M with the data set W2 as the input and the data set R as the target; verify and predict the short-term irradiance of the photovoltaic power station according to the model. The invention effectively expands the feature space of the NWP forecast by combining various meteorological elements, deeply excavates the effective information in the NWP forecast, and improves the accuracy of the irradiance forecast.

Description

technical field [0001] The invention belongs to the field of photovoltaic power generation, and in particular relates to a method and system for forecasting photovoltaic irradiance based on XGBoost. Background technique [0002] As a clean and renewable energy source, solar energy has attracted more and more attention. With the continuous expansion of photovoltaic grid-connected scale, photovoltaic has become the third largest electric energy source. However, the intensity of sunlight received by the ground is affected by weather conditions, cloud movement, etc. The impact is huge. When the photovoltaic penetration power is high, it will have an impact on the power grid. Therefore, in order to ensure the safe and reliable operation of the power grid and reduce the curtailment of solar power, it is very important to accurately predict the irradiance and photovoltaic power. According to the needs of the power sector to arrange dispatching plans, model forecast irradiance predi...

AI Technical Summary

Problems solved by technology

The time series method is based on a linear model, which has defects in dealing with problems in the multidimensional nonlinear field; SVM has good generalization ability, but the model training time is long, and it is difficult to meet the requirements in short-term forecasting; BP has good learning ability and non-linear Linear expression ability, but it is easy to fall into local optimum, and the convergence speed cannot meet the requirements

[0005] In addition, the method based on MOS correction can greatly reduce the system error, but it does not have the ability to improve the forecast of intraday cloudy, sudden weather changes, etc., and cannot reflect the details of weather changes at a higher time resolution; statistical methods such as Kalman filtering and neural networks are used to analyze The correction of NWP irradiance can also improve the accuracy of short-term irradiance forecast, but it has not combined with the law of photovoltaic power generation to make targeted corrections to NWP irradiance, which cannot meet the accuracy requirements of the industry for NWP forecasting at this stage

[0006] The above statistical forecasting methods and various machine learning methods do not consider the combination of various meteorological elements. In numerical weather forecasting, since the forecast error of the actual atmosphere has the characteristics of obvious changes with different weather situations, the closer the weather situation changes, the closer the weather situation is. The law of its forecast error is also closer. Therefore, there is an urgent need for a method that uses a combination of various elements as a sample to represent weather trends and give a more accurate forecast of the forecast error.

Images