LSTM-based short-term photovoltaic power prediction method and system

Abstract

The invention discloses a method for predicting short-term photovoltaic power generation power based on LSTM, which includes the following steps: S1, collecting historical power generation data of photovoltaic power plants and historical meteorological data of the area where they are located, and preprocessing them; S2, selecting a year The historical data of is used as the training set, and the input is the power generation and meteorological data of the 96 time points before the output point, which is expressed in the form of , where m ₁ is the number of samples, m ₂ is the number of time steps in the input layer, m ₃ is the dimension of the input layer, the number of time steps is selected as 96, and the dimension of the input layer is 8; S3, establish a prediction model based on LSTM, and input the training set into LSTM for training, and the training method adopts the backpropagation BPTT algorithm; S4, adopts Point-by-point forecasting is performed in an iterative manner, and the output power of the next day can be predicted in the form of a single output.

Description

technical field [0001] The present invention relates to the field of machine learning and new energy, in particular to a method and system for predicting photovoltaic power based on a long-short-term memory network. Background technique [0002] Compared with developed countries, my country started relatively late in the photovoltaic power generation industry. However, since the vigorous promotion of photovoltaic power generation in 2009, the development momentum has been very rapid. From 2009 to 2013, the annual growth rate of installed capacity was not less than 100%, from 3,000 kilowatts in 2009 to 125.79 million kilowatts in 2017, ranking first in the world. In recent years, the distribution of newly installed photovoltaic capacity has shifted significantly, and it has begun to develop in the central and eastern regions. From January to November 2017, the proportion of newly installed photovoltaic capacity in Northwest China decreased by 17 percentage points year-on-ye...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to address the deficiencies in the prior art, and propose a photovoltaic power prediction method based on long-short-term memory network, which solves the problem of low accuracy of prediction results in the case of large weather fluctuations, and can effectively predict photovoltaic power Effective ultra-short-term forecasting of power generation; at the same time, it has fewer considerations, low computational complexity, and high forecasting efficiency

Images