Solar radiation prediction method based on depth BISLTM

Abstract

The invention relates to the technical field of solar radiation prediction, and discloses a solar radiation prediction method based on depth BISLTM, and the method comprises the steps: employing CEEMDAN to decompose an original solar radiation sequence into a sub-component set; constructing an independent bidirectional long-short-term memory neural network sub-model for each component; optimizingthe BISLTM model of each sub-component by adopting an SCA algorithm, and outputting an optimal parameter of the BISLTM model; calculating prediction valued of a target variable by using the trained sub-component SCA-BiLSTM model; and aggregating the prediction values of the SCA-BiLSTM model of the sub-components to obtain a final prediction result of the CEN-SCA-BiLSTM model. Compared with the prior art, the prediction method provided by the invention can obtain a high-precision solar radiation prediction value, and plays an important role in the utilization efficiency of photovoltaic resources and the safe and stable operation of a power grid.

Description

technical field [0001] The invention relates to the technical field of solar radiation forecasting, in particular to a deep BISLTM-based solar radiation forecasting method. Background technique [0002] Solar energy is the radiant light and heat from the sun and is an important source of renewable energy. According to the "Renewable Energy 2019-Global Status Report", in 2018, the newly installed capacity of renewable energy in the world was 181GW, of which the installed capacity of solar photovoltaic power generation was 100GW, accounting for 55% of the newly installed capacity of renewable energy, followed by wind power 51GW (28%), followed by hydropower with 20GW (11%). The enormous potential of solar energy makes it one of the most attractive sources of energy for generating electricity and heating. In addition, efficient solar energy storage technologies, such as photovoltaic cells, make the use of solar energy cost-effective and economically competitive. Photovoltaic...

AI Technical Summary

Problems solved by technology

However, in the above-mentioned types of solar radiation prediction models, the input data of remote sensing inversion, radiative transfer model and numerical prediction model are difficult to obtain, and the calculation cost is high

Empirical models and time-series models have low computational costs, but due to the intermittent nature and high sampling frequency of solar radiation sequences, empirical models, time-series models, and traditional machine learning models have limited learning capabilities for complex nonlinear sequences, limiting radiation forecasting Therefore, a new method is urgently needed to meet the demand for high-precision solar radiation prediction data for large-scale photovoltaic grid-connected

Images