New energy photovoltaic power generation power prediction method

Abstract

The invention belongs to the technical field of electric power, and particularly relates to a new energy photovoltaic power generation power prediction method. The method comprises the following steps: S1, obtaining photovoltaic power station historical data containing weather mutation sensitive factors as a basic training sample set A; S2, according to the basic training sample set A, screening deformable basic samples to form a deformable basic sample set Deformable; S3, selecting a deformed seed Seedm (m = 1, 2,..., M) according to the deformable basic sample set Deformable, and generating a deformed seed data set Seed by the M groups of deformed seeds; S4, sequentially performing deformation according to the first (M-1) groups of data Seedm (m = 1, 2,..., M-1) of the deformation seed data set Seed, generating (M-1) groups of virtual deformation samples Virtualm (m = 1, 2,..., M-1), and forming a virtual deformation sample set Virtual; S5, performing dual neural network training according to samples in the basic training sample set A and the virtual deformation sample set Virtual; S6, performing ultra-short-term power prediction by using the trained neural network; according to the invention, the stability of the power grid is improved while the power prediction accuracy under the severe weather condition is improved.

Description

Technical field [0001] The present invention belongs to the technical field of power, particularly relates to a new energy photovoltaic power prediction method. Background technique [0002] In recent years, the proportion of new energy power generation is increasing year by year, photovoltaic and wind power generation is steadily developing and advancing. But no matter photovoltaic power generation or wind power generation due to the characteristics of randomness and instability, its power into the grid would endanger the stability of the grid. So the need for new energy generation ahead power forecasting, scheduling, planning in advance in order to guarantee grid stability. Currently, the neural network forecasting method, we can make a good prediction of the ultra-short-term power. But the basic is better predict when or mild sunny weather, predict when sudden storms or other severe weather effects are often larger error. The reason is usually used when predicting a particular...

AI Technical Summary

Problems solved by technology

The reason is that cloud cover, wind speed, wind direction, air temperature, air pressure, humidity and other reference quantities at a specific time are usually used for ultra-short-term power forecasting, which is less sensitive to sudden changes in weather.

Moreover, when using prediction methods such as neural networks for sample training, it is difficult for the samples to provide more data on sudden weather, which makes it difficult for the trained neural network to predict the power under sudden weather.

Images