The Short-term Photovoltaic Power Prediction Based On Grey Model And Machine Learning

With the development of modern industry,the vigorous exploitation of fossil energy has brought great pressure to the natural environment,and the problem of energy shortage is becoming more and more prominent,and the development of renewable energy is imminent.Photovoltaic power generation,as one of the power generation methods with simple mining requirements and high commercial value,has been paid more and more attention by more and more countries.Although photovoltaic power generation has many advantages,in the actual mining process it is influenced by the environmental temperature,relative humidity and solar irradiance and many other factors,which not only increase the difficulty of grid-connected power generation,but also add a lot of pressures to power dispatching personnel.So accurately predicting the power of photovoltaic power generation is of great significance in practical use.Photovoltaic data has the characteristics of large volatility and randomness,this paper has carried out the following researches on the short-term photovoltaic power prediction problem: 1)Analyzing and determining the main factors affecting the photovoltaic power,making a simulation to analyze the impact of physical factors on power.The meteorological factors affecting short-term photovoltaic power prediction are analyzed by regression equation fitting and grey relational degree value.2)Based on grey model and machine learning method,the short-term photovoltaic power prediction of machine learning(GM,BP,SVM,GVM)and hybrid model(GM-ML)is realized respectively under the condition of sunny and cloudy days,The experimental results show that the prediction accuracy of GM-GVM is higher.3)The weights and thresholds of GM-ML model are optimized by particle swarm optimization algorithm,and a PSO-GM-ML model is established,and the results show that PSO-GM-ML is better than GM-ML prediction effect by comparison and analysis.4)The Sobol algorithm is used to analyze the uncertainty of the input factors of the model.By calculating the first-order sensitivity index and the total-order sensitivity index,the effects of single factor and combination factor on the short-term PV power prediction at different time periods are studied.The results show that the irradiance is the main meteorological factor affecting power,but the interaction of irradiance and temperature has a greater influence on power with time.