Short-term Prediction Of Photovoltaic Power Generation Based On Variational Mode Decomposition And Neural Network

Energy is one of the basic materials on which the earth depends.With the development of human society,the demand for energy is increasing day by day,and the use of photovoltaics by humans is also increasing.Photovoltaic power generation has its own limitations,and the meteorological conditions of disorderly changes make the photovoltaic power generation random and volatility.Accurately predicting the photovoltaic power generation has important application value for the regulation of photovoltaic grid-connected power grid and the scale development of photovoltaic industry.This paper proposes a short-term prediction method of photovoltaic power generation based on BP neural network.The photovoltaic power generation is related to the photovoltaic power system factors and meteorological factors.The photovoltaic power plant system factors include the components and the types of components and line length loss.Once the photovoltaic power station is built,the system factors will not change much,so the photovoltaic power generation mainly depends on meteorological factors.related.This paper analyzes the relationship between photovoltaic power generation and solar radiation,ambient temperature,humidity,atmospheric pressure,wind speed,wind direction,rainfall,weather type and season.The main work and innovations of this paper are as follows:Photovoltaic power is related to the type of weather,but weather types cannot be directly used to analyze the quantitative relationship between power generation and photovoltaic power.Sample Entropy(SE)is a method of matching its own data and measuring its timing complexity,which is often used for quantitative classification of sequences.In this paper,the actual weather types are classified into four types:sunny,cloudy,hazy,rain and snow.The sample entropy method is used to quantify the daily weather type into different numerical data by combining different daily photovoltaic power values.It is then studied together with other meteorological factors as an influencing factor.According to the characteristics of photovoltaic power volatility and randomness,the variational mode decomposition(VMD)is used to decompose the historical data of photovoltaic power generation to obtain multiple stable sub-modes.If the number of sub-modals is set too much,modal aliasing will occur,causing serious information interference.If the number of sub-modals is set too small,information leakage will occur,and the significance of using variational mode decomposition will be weakened.In order to select the appropriate number of sub-modes,the permutation entropy method is used to determine the sub-modal number.In this paper,the photovoltaic power data is decomposed into six sub-modes.The BP neural network method is a commonly used method in the prediction of photovoltaic power generation.However,the single BP neural network has a slow convergence rate and is easy to fall into the local extremum in the network training for the photovoltaic power generation prediction.The genetic algorithm is the mathematical optimization in solving the optimization problem.The search algorithm has the characteristics of high convergence and fast and random operation.This paper chooses genetic algorithm(GA)to improve the weight threshold of BP neural network and improve the prediction accuracy.This paper selects the sampling data of a 200kW photovoltaic power station in Alice Springs,Australia,every 5 minutes in 2017,and establishes a GA-BP prediction model based on the improved output of SE-VMD.The data of the data from January 1st to January 30th were used to select the photovoltaic power generation in sunny and cloudy days,cloudy,hazy,rain and snow in summer and winter as verification data.The prediction results show that the mean square error and average percentage error of the four weather types in the SE-VMD-GA-BP prediction model are smaller than the BP prediction model and the GA-BP prediction model,which provides a more accurate short-term prediction of photovoltaic power generation,effective method.The hardware principle of the PV array power prediction system is introduced,and the principle of its wireless transmission system and its platform ideas are introduced.