Research On Short Term Photovoltaic Power Prediction Based On Deep Learning

Energy is one of the important materials that can not be ignored in the process of the development of human social civilization.With our excessive attention to the speed of economic development,we ignore the energy crisis caused by the continuous consumption of traditional energy and the continuous deterioration of the ecological environment on which we rely for survival.The limitations of traditional energy can not meet the needs of society.Solar energy is widely used for its unique advantages of safety,efficiency and wide distribution.However,photovoltaic power varies with complex and changeable meteorological conditions,which makes it obviously nonlinear and unstable,and brings great impact and challenge to the safe and stable operation of power system.When large-scale photovoltaic grid connected power generation,its frequency will exceed the critical value of large power grid,resulting in the collapse of power grid.Accurate short-term photovoltaic power prediction can provide an important basis for the power system to make accurate scheduling plan,ensure that the power system can maximize the consumption of photovoltaic power generation,and improve the stability of the safe operation of the power system.Photovoltaic power generation has the characteristics of strong uncontrollability and intermittence,so how to effectively improve the accuracy of photovoltaic power prediction has a strong practical significance for the stable operation of power system.In this paper,according to the current situation of photovoltaic power prediction at home and abroad,a short-term photovoltaic power prediction model of VMD-ISSA-GRU based on deep learning algorithm is proposed to further improve the accuracy of short-term photovoltaic power prediction.First of all,through Pearson and Spearman correlation analysis,the main influencing factors of photovoltaic output power are determined to reduce the dimension of high-dimensional meteorological data and improve the operation efficiency and prediction accuracy of the model.Secondly,in view of the instability and intermittence of photovoltaic power,the historical data of photovoltaic power is decomposed by variational mode decomposition.The optimal number of decomposition submodes is determined by using the correlation between the decomposed residual error and the original photovoltaic power data,which can fully decompose the historical photovoltaic power data and effectively reduce the nonlinearity and instability of historical photovoltaic power.Then,the GRU neural network is constructed,and the structural parameters of the GRU neural network are optimized by improved sparrow search algorithm(ISSA)to improve the prediction accuracy of the model.Finally,the VMD-ISSA-GRU prediction model is established based on the actual photovoltaic power data.The results show that the root mean square error and the mean absolute percentage error of VMD-ISSA-GRU prediction model are smaller than other traditional prediction models,whether the weather conditions are stable or volatile,which verifies the effectiveness and superiority of the model.